| { |
| "source_repo": "yjb6/UniSpace", |
| "source_revision": "0f0fdf941329f570be65e6c6549b8452e1f967b1", |
| "source_file": "encoders/pr-dinov2-tokenizer.pt", |
| "selected": { |
| "ema.encoder.sem_embeddings.cls_token": { |
| "shape": [ |
| 1, |
| 1, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 768, |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2015309760, |
| "contiguous": true |
| }, |
| "ema.encoder.sem_embeddings.mask_token": { |
| "shape": [ |
| 1, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2015312960, |
| "contiguous": true |
| }, |
| "ema.encoder.sem_embeddings.register_tokens": { |
| "shape": [ |
| 1, |
| 4, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 3072, |
| 768, |
| 1 |
| ], |
| "numel": 3072, |
| "bytes": 12288, |
| "storage_offset": 0, |
| "checkpoint_offset": 2015316160, |
| "contiguous": true |
| }, |
| "ema.encoder.sem_embeddings.position_embeddings": { |
| "shape": [ |
| 1, |
| 1370, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1052160, |
| 768, |
| 1 |
| ], |
| "numel": 1052160, |
| "bytes": 4208640, |
| "storage_offset": 0, |
| "checkpoint_offset": 2015328576, |
| "contiguous": true |
| }, |
| "ema.encoder.sem_embeddings.patch_embeddings.projection.weight": { |
| "shape": [ |
| 768, |
| 3, |
| 14, |
| 14 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 588, |
| 196, |
| 14, |
| 1 |
| ], |
| "numel": 451584, |
| "bytes": 1806336, |
| "storage_offset": 0, |
| "checkpoint_offset": 2019537344, |
| "contiguous": true |
| }, |
| "ema.encoder.sem_embeddings.patch_embeddings.projection.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2021343808, |
| "contiguous": true |
| }, |
| "ema.encoder.recon_embeddings.cls_token": { |
| "shape": [ |
| 1, |
| 1, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 768, |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2021347008, |
| "contiguous": true |
| }, |
| "ema.encoder.recon_embeddings.register_tokens": { |
| "shape": [ |
| 1, |
| 4, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 3072, |
| 768, |
| 1 |
| ], |
| "numel": 3072, |
| "bytes": 12288, |
| "storage_offset": 0, |
| "checkpoint_offset": 2021350208, |
| "contiguous": true |
| }, |
| "ema.encoder.recon_embeddings.position_embeddings": { |
| "shape": [ |
| 1, |
| 1370, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1052160, |
| 768, |
| 1 |
| ], |
| "numel": 1052160, |
| "bytes": 4208640, |
| "storage_offset": 0, |
| "checkpoint_offset": 2021362624, |
| "contiguous": true |
| }, |
| "ema.encoder.recon_embeddings.patch_embeddings.projection.weight": { |
| "shape": [ |
| 768, |
| 3, |
| 14, |
| 14 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 588, |
| 196, |
| 14, |
| 1 |
| ], |
| "numel": 451584, |
| "bytes": 1806336, |
| "storage_offset": 0, |
| "checkpoint_offset": 2025571392, |
| "contiguous": true |
| }, |
| "ema.encoder.recon_embeddings.patch_embeddings.projection.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2027377856, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.norm1.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2027381056, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.norm1.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2027384256, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.attention.attention.query.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2027387456, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.attention.attention.query.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2029746880, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.attention.attention.key.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2029750080, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.attention.attention.key.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2032109504, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.attention.attention.value.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2032112704, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.attention.attention.value.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2034472128, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.attention.output.dense.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2034475328, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.attention.output.dense.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2036834752, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.layer_scale1.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2036837952, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.norm2.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2036841152, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.norm2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2036844352, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.mlp.fc1.weight": { |
| "shape": [ |
| 3072, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2036847552, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.mlp.fc1.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 3072, |
| "bytes": 12288, |
| "storage_offset": 0, |
| "checkpoint_offset": 2046284864, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.mlp.fc2.weight": { |
| "shape": [ |
| 768, |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 3072, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2046297280, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.mlp.fc2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2055734592, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.0.layer_scale2.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2055737792, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.norm1.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2055740992, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.norm1.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2055744192, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.attention.attention.query.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2055747392, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.attention.attention.query.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2058106816, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.attention.attention.key.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2058110016, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.attention.attention.key.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2060469440, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.attention.attention.value.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2060472640, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.attention.attention.value.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2062832064, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.attention.output.dense.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2062835264, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.attention.output.dense.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2065194688, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.layer_scale1.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2065197888, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.norm2.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2065201088, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.norm2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2065204288, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.mlp.fc1.weight": { |
| "shape": [ |
| 3072, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2065207488, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.mlp.fc1.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 3072, |
| "bytes": 12288, |
| "storage_offset": 0, |
| "checkpoint_offset": 2074644800, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.mlp.fc2.weight": { |
| "shape": [ |
| 768, |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 3072, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2074657216, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.mlp.fc2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2084094528, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.1.layer_scale2.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2084097728, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.norm1.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2084100928, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.norm1.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2084104128, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.attention.attention.query.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2084107328, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.attention.attention.query.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2086466752, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.attention.attention.key.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2086469952, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.attention.attention.key.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2088829376, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.attention.attention.value.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2088832576, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.attention.attention.value.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2091192000, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.attention.output.dense.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2091195200, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.attention.output.dense.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2093554624, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.layer_scale1.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2093557824, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.norm2.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2093561024, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.norm2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2093564224, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.mlp.fc1.weight": { |
| "shape": [ |
| 3072, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2093567424, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.mlp.fc1.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 3072, |
| "bytes": 12288, |
| "storage_offset": 0, |
| "checkpoint_offset": 2103004736, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.mlp.fc2.weight": { |
| "shape": [ |
| 768, |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 3072, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2103017152, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.mlp.fc2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2112454464, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.2.layer_scale2.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2112457664, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.norm1.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2112460864, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.norm1.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2112464064, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.attention.attention.query.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2112467264, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.attention.attention.query.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2114826688, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.attention.attention.key.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2114829888, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.attention.attention.key.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2117189312, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.attention.attention.value.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2117192512, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.attention.attention.value.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2119551936, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.attention.output.dense.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2119555136, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.attention.output.dense.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2121914560, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.layer_scale1.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2121917760, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.norm2.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2121920960, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.norm2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2121924160, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.mlp.fc1.weight": { |
| "shape": [ |
| 3072, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2121927360, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.mlp.fc1.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 3072, |
| "bytes": 12288, |
| "storage_offset": 0, |
| "checkpoint_offset": 2131364672, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.mlp.fc2.weight": { |
| "shape": [ |
| 768, |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 3072, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2131377088, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.mlp.fc2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2140814400, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.3.layer_scale2.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2140817600, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.norm1.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2140820800, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.norm1.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2140824000, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.attention.attention.query.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2140827200, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.attention.attention.query.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2143186624, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.attention.attention.key.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2143189824, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.attention.attention.key.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2145549248, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.attention.attention.value.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2145552448, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.attention.attention.value.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2147911872, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.attention.output.dense.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2147915072, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.attention.output.dense.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2150274496, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.layer_scale1.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2150277696, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.norm2.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2150280896, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.norm2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2150284096, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.mlp.fc1.weight": { |
| "shape": [ |
| 3072, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2150287296, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.mlp.fc1.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 3072, |
| "bytes": 12288, |
| "storage_offset": 0, |
| "checkpoint_offset": 2159724608, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.mlp.fc2.weight": { |
| "shape": [ |
| 768, |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 3072, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2159737024, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.mlp.fc2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2169174336, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.4.layer_scale2.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2169177536, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.norm1.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2169180736, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.norm1.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2169183936, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.attention.attention.query.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2169187136, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.attention.attention.query.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2171546560, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.attention.attention.key.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2171549760, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.attention.attention.key.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2173909184, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.attention.attention.value.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2173912384, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.attention.attention.value.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2176271808, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.attention.output.dense.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2176275008, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.attention.output.dense.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2178634432, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.layer_scale1.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2178637632, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.norm2.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2178640832, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.norm2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2178644032, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.mlp.fc1.weight": { |
| "shape": [ |
| 3072, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2178647232, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.mlp.fc1.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 3072, |
| "bytes": 12288, |
| "storage_offset": 0, |
| "checkpoint_offset": 2188084544, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.mlp.fc2.weight": { |
| "shape": [ |
| 768, |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 3072, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2188096960, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.mlp.fc2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2197534272, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.5.layer_scale2.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2197537472, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.norm1.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2197540672, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.norm1.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2197543872, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.attention.attention.query.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2197547072, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.attention.attention.query.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2199906496, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.attention.attention.key.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2199909696, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.attention.attention.key.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2202269120, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.attention.attention.value.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2202272320, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.attention.attention.value.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2204631744, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.attention.output.dense.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2204634944, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.attention.output.dense.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2206994368, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.layer_scale1.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2206997568, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.norm2.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2207000768, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.norm2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2207003968, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.mlp.fc1.weight": { |
| "shape": [ |
| 3072, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2207007168, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.mlp.fc1.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 3072, |
| "bytes": 12288, |
| "storage_offset": 0, |
| "checkpoint_offset": 2216444480, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.mlp.fc2.weight": { |
| "shape": [ |
| 768, |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 3072, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2216456896, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.mlp.fc2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2225894208, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.6.layer_scale2.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2225897408, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.norm1.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2225900608, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.norm1.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2225903808, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.attention.attention.query.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2225907008, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.attention.attention.query.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2228266432, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.attention.attention.key.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2228269632, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.attention.attention.key.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2230629056, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.attention.attention.value.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2230632256, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.attention.attention.value.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2232991680, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.attention.output.dense.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2232994880, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.attention.output.dense.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2235354304, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.layer_scale1.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2235357504, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.norm2.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2235360704, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.norm2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2235363904, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.mlp.fc1.weight": { |
| "shape": [ |
| 3072, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2235367104, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.mlp.fc1.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 3072, |
| "bytes": 12288, |
| "storage_offset": 0, |
| "checkpoint_offset": 2244804416, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.mlp.fc2.weight": { |
| "shape": [ |
| 768, |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 3072, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2244816832, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.mlp.fc2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2254254144, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.7.layer_scale2.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2254257344, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.norm1.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2254260544, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.norm1.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2254263744, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.attention.attention.query.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2254266944, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.attention.attention.query.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2256626368, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.attention.attention.key.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2256629568, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.attention.attention.key.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2258988992, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.attention.attention.value.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2258992192, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.attention.attention.value.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2261351616, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.attention.output.dense.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2261354816, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.attention.output.dense.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2263714240, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.layer_scale1.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2263717440, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.norm2.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2263720640, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.norm2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2263723840, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.mlp.fc1.weight": { |
| "shape": [ |
| 3072, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2263727040, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.mlp.fc1.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 3072, |
| "bytes": 12288, |
| "storage_offset": 0, |
| "checkpoint_offset": 2273164352, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.mlp.fc2.weight": { |
| "shape": [ |
| 768, |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 3072, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2273176768, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.mlp.fc2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2282614080, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.8.layer_scale2.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2282617280, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.norm1.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2282620480, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.norm1.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2282623680, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.attention.attention.query.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2282626880, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.attention.attention.query.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2284986304, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.attention.attention.key.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2284989504, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.attention.attention.key.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2287348928, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.attention.attention.value.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2287352128, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.attention.attention.value.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2289711552, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.attention.output.dense.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2289714752, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.attention.output.dense.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2292074176, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.layer_scale1.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2292077376, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.norm2.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2292080576, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.norm2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2292083776, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.mlp.fc1.weight": { |
| "shape": [ |
| 3072, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2292086976, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.mlp.fc1.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 3072, |
| "bytes": 12288, |
| "storage_offset": 0, |
| "checkpoint_offset": 2301524288, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.mlp.fc2.weight": { |
| "shape": [ |
| 768, |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 3072, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2301536704, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.mlp.fc2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2310974016, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.9.layer_scale2.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2310977216, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.norm1.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2310980416, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.norm1.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2310983616, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.attention.attention.query.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2310986816, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.attention.attention.query.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2313346240, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.attention.attention.key.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2313349440, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.attention.attention.key.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2315708864, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.attention.attention.value.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2315712064, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.attention.attention.value.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2318071488, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.attention.output.dense.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2318074688, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.attention.output.dense.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2320434112, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.layer_scale1.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2320437312, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.norm2.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2320440512, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.norm2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2320443712, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.mlp.fc1.weight": { |
| "shape": [ |
| 3072, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2320446912, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.mlp.fc1.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 3072, |
| "bytes": 12288, |
| "storage_offset": 0, |
| "checkpoint_offset": 2329884224, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.mlp.fc2.weight": { |
| "shape": [ |
| 768, |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 3072, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2329896640, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.mlp.fc2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2339333952, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.10.layer_scale2.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2339337152, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.norm1.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2339340352, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.norm1.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2339343552, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.attention.attention.query.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2339346752, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.attention.attention.query.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2341706176, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.attention.attention.key.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2341709376, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.attention.attention.key.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2344068800, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.attention.attention.value.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2344072000, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.attention.attention.value.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2346431424, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.attention.output.dense.weight": { |
| "shape": [ |
| 768, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 589824, |
| "bytes": 2359296, |
| "storage_offset": 0, |
| "checkpoint_offset": 2346434624, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.attention.output.dense.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2348794048, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.layer_scale1.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2348797248, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.norm2.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2348800448, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.norm2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2348803648, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.mlp.fc1.weight": { |
| "shape": [ |
| 3072, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2348806848, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.mlp.fc1.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 3072, |
| "bytes": 12288, |
| "storage_offset": 0, |
| "checkpoint_offset": 2358244160, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.mlp.fc2.weight": { |
| "shape": [ |
| 768, |
| 3072 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 3072, |
| 1 |
| ], |
| "numel": 2359296, |
| "bytes": 9437184, |
| "storage_offset": 0, |
| "checkpoint_offset": 2358256576, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.mlp.fc2.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2367693888, |
| "contiguous": true |
| }, |
| "ema.encoder.encoder.layer.11.layer_scale2.lambda1": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2367697088, |
| "contiguous": true |
| }, |
| "ema.encoder.layernorm.weight": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2367700288, |
| "contiguous": true |
| }, |
| "ema.encoder.layernorm.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 2367703488, |
| "contiguous": true |
| }, |
| "ema.encoder.recon_proj.weight": { |
| "shape": [ |
| 128, |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 768, |
| 1 |
| ], |
| "numel": 98304, |
| "bytes": 393216, |
| "storage_offset": 0, |
| "checkpoint_offset": 2367706688, |
| "contiguous": true |
| }, |
| "ema.encoder.recon_proj.bias": { |
| "shape": [ |
| 128 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 128, |
| "bytes": 512, |
| "storage_offset": 0, |
| "checkpoint_offset": 2368100032, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_pos_embed": { |
| "shape": [ |
| 1, |
| 257, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 296064, |
| 1152, |
| 1 |
| ], |
| "numel": 296064, |
| "bytes": 1184256, |
| "storage_offset": 0, |
| "checkpoint_offset": 2368100672, |
| "contiguous": true |
| }, |
| "ema.decoder.trainable_cls_token": { |
| "shape": [ |
| 1, |
| 1, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1152, |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2369285056, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_embed.weight": { |
| "shape": [ |
| 1152, |
| 896 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 896, |
| 1 |
| ], |
| "numel": 1032192, |
| "bytes": 4128768, |
| "storage_offset": 0, |
| "checkpoint_offset": 2369289792, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_embed.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2373418688, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.0.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2373423424, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.0.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2378731968, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.0.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2378736704, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.0.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2384045248, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.0.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2384049984, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.0.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2389358528, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.0.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2389363264, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.0.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2394671808, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.0.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2394676544, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.0.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 2413551040, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.0.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2413567552, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.0.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2432442048, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.0.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2432446784, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.0.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2432451520, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.0.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2432456256, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.0.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2432460992, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.1.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2432465728, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.1.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2437774272, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.1.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2437779008, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.1.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2443087552, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.1.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2443092288, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.1.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2448400832, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.1.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2448405568, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.1.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2453714112, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.1.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2453718848, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.1.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 2472593344, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.1.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2472609856, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.1.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2491484352, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.1.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2491489088, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.1.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2491493824, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.1.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2491498560, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.1.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2491503296, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.2.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2491508032, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.2.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2496816576, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.2.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2496821312, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.2.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2502129856, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.2.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2502134592, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.2.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2507443136, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.2.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2507447872, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.2.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2512756416, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.2.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2512761152, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.2.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 2531635648, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.2.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2531652160, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.2.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2550526656, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.2.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2550531392, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.2.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2550536128, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.2.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2550540864, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.2.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2550545600, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.3.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2550550336, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.3.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2555858880, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.3.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2555863616, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.3.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2561172160, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.3.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2561176896, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.3.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2566485440, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.3.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2566490176, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.3.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2571798720, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.3.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2571803456, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.3.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 2590677952, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.3.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2590694464, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.3.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2609568960, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.3.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2609573696, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.3.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2609578432, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.3.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2609583168, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.3.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2609587904, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.4.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2609592640, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.4.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2614901184, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.4.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2614905920, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.4.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2620214464, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.4.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2620219200, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.4.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2625527744, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.4.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2625532480, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.4.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2630841024, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.4.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2630845760, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.4.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 2649720256, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.4.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2649736768, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.4.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2668611264, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.4.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2668616000, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.4.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2668620736, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.4.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2668625472, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.4.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2668630208, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.5.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2668634944, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.5.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2673943488, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.5.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2673948224, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.5.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2679256768, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.5.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2679261504, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.5.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2684570048, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.5.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2684574784, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.5.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2689883328, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.5.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2689888064, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.5.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 2708762560, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.5.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2708779072, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.5.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2727653568, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.5.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2727658304, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.5.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2727663040, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.5.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2727667776, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.5.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2727672512, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.6.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2727677248, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.6.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2732985792, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.6.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2732990528, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.6.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2738299072, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.6.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2738303808, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.6.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2743612352, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.6.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2743617088, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.6.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2748925632, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.6.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2748930368, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.6.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 2767804864, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.6.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2767821376, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.6.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2786695872, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.6.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2786700608, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.6.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2786705344, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.6.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2786710080, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.6.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2786714816, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.7.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2786719552, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.7.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2792028096, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.7.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2792032832, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.7.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2797341376, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.7.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2797346112, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.7.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2802654656, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.7.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2802659392, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.7.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2807967936, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.7.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2807972672, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.7.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 2826847168, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.7.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2826863680, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.7.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2845738176, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.7.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2845742912, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.7.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2845747648, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.7.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2845752384, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.7.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2845757120, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.8.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2845761856, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.8.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2851070400, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.8.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2851075136, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.8.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2856383680, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.8.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2856388416, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.8.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2861696960, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.8.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2861701696, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.8.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2867010240, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.8.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2867014976, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.8.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 2885889472, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.8.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2885905984, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.8.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2904780480, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.8.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2904785216, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.8.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2904789952, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.8.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2904794688, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.8.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2904799424, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.9.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2904804160, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.9.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2910112704, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.9.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2910117440, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.9.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2915425984, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.9.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2915430720, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.9.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2920739264, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.9.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2920744000, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.9.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2926052544, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.9.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2926057280, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.9.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 2944931776, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.9.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2944948288, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.9.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2963822784, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.9.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2963827520, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.9.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2963832256, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.9.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2963836992, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.9.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2963841728, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.10.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2963846464, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.10.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2969155008, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.10.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2969159744, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.10.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2974468288, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.10.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2974473024, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.10.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2979781568, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.10.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 2979786304, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.10.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 2985094848, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.10.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 2985099584, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.10.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3003974080, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.10.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3003990592, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.10.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3022865088, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.10.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3022869824, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.10.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3022874560, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.10.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3022879296, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.10.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3022884032, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.11.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3022888768, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.11.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3028197312, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.11.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3028202048, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.11.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3033510592, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.11.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3033515328, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.11.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3038823872, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.11.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3038828608, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.11.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3044137152, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.11.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3044141888, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.11.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3063016384, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.11.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3063032896, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.11.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3081907392, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.11.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3081912128, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.11.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3081916864, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.11.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3081921600, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.11.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3081926336, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.12.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3081931072, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.12.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3087239616, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.12.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3087244352, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.12.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3092552896, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.12.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3092557632, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.12.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3097866176, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.12.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3097870912, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.12.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3103179456, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.12.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3103184192, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.12.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3122058688, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.12.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3122075200, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.12.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3140949696, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.12.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3140954432, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.12.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3140959168, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.12.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3140963904, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.12.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3140968640, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.13.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3140973376, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.13.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3146281920, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.13.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3146286656, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.13.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3151595200, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.13.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3151599936, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.13.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3156908480, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.13.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3156913216, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.13.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3162221760, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.13.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3162226496, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.13.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3181100992, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.13.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3181117504, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.13.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3199992000, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.13.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3199996736, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.13.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3200001472, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.13.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3200006208, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.13.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3200010944, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.14.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3200015680, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.14.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3205324224, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.14.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3205328960, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.14.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3210637504, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.14.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3210642240, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.14.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3215950784, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.14.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3215955520, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.14.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3221264064, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.14.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3221268800, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.14.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3240143296, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.14.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3240159808, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.14.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3259034304, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.14.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3259039040, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.14.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3259043776, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.14.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3259048512, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.14.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3259053248, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.15.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3259057984, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.15.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3264366528, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.15.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3264371264, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.15.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3269679808, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.15.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3269684544, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.15.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3274993088, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.15.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3274997824, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.15.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3280306368, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.15.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3280311104, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.15.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3299185600, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.15.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3299202112, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.15.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3318076608, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.15.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3318081344, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.15.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3318086080, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.15.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3318090816, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.15.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3318095552, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.16.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3318100288, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.16.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3323408832, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.16.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3323413568, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.16.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3328722112, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.16.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3328726848, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.16.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3334035392, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.16.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3334040128, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.16.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3339348672, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.16.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3339353408, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.16.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3358227904, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.16.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3358244416, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.16.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3377118912, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.16.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3377123648, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.16.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3377128384, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.16.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3377133120, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.16.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3377137856, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.17.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3377142592, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.17.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3382451136, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.17.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3382455872, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.17.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3387764416, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.17.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3387769152, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.17.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3393077696, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.17.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3393082432, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.17.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3398390976, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.17.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3398395712, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.17.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3417270208, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.17.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3417286720, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.17.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3436161216, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.17.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3436165952, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.17.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3436170688, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.17.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3436175424, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.17.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3436180160, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.18.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3436184896, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.18.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3441493440, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.18.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3441498176, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.18.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3446806720, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.18.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3446811456, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.18.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3452120000, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.18.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3452124736, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.18.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3457433280, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.18.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3457438016, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.18.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3476312512, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.18.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3476329024, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.18.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3495203520, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.18.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3495208256, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.18.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3495212992, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.18.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3495217728, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.18.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3495222464, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.19.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3495227200, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.19.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3500535744, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.19.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3500540480, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.19.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3505849024, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.19.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3505853760, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.19.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3511162304, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.19.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3511167040, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.19.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3516475584, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.19.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3516480320, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.19.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3535354816, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.19.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3535371328, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.19.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3554245824, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.19.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3554250560, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.19.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3554255296, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.19.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3554260032, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.19.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3554264768, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.20.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3554269504, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.20.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3559578048, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.20.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3559582784, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.20.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3564891328, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.20.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3564896064, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.20.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3570204608, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.20.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3570209344, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.20.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3575517888, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.20.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3575522624, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.20.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3594397120, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.20.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3594413632, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.20.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3613288128, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.20.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3613292864, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.20.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3613297600, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.20.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3613302336, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.20.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3613307072, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.21.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3613311808, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.21.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3618620352, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.21.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3618625088, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.21.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3623933632, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.21.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3623938368, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.21.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3629246912, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.21.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3629251648, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.21.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3634560192, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.21.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3634564928, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.21.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3653439424, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.21.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3653455936, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.21.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3672330432, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.21.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3672335168, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.21.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3672339904, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.21.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3672344640, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.21.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3672349376, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.22.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3672354112, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.22.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3677662656, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.22.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3677667392, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.22.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3682975936, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.22.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3682980672, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.22.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3688289216, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.22.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3688293952, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.22.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3693602496, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.22.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3693607232, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.22.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3712481728, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.22.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3712498240, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.22.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3731372736, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.22.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3731377472, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.22.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3731382208, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.22.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3731386944, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.22.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3731391680, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.23.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3731396416, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.23.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3736704960, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.23.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3736709696, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.23.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3742018240, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.23.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3742022976, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.23.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3747331520, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.23.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3747336256, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.23.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3752644800, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.23.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3752649536, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.23.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3771524032, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.23.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3771540544, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.23.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3790415040, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.23.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3790419776, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.23.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3790424512, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.23.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3790429248, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.23.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3790433984, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.24.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3790438720, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.24.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3795747264, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.24.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3795752000, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.24.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3801060544, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.24.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3801065280, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.24.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3806373824, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.24.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3806378560, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.24.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3811687104, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.24.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3811691840, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.24.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3830566336, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.24.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3830582848, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.24.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3849457344, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.24.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3849462080, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.24.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3849466816, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.24.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3849471552, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.24.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3849476288, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.25.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3849481024, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.25.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3854789568, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.25.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3854794304, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.25.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3860102848, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.25.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3860107584, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.25.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3865416128, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.25.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3865420864, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.25.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3870729408, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.25.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3870734144, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.25.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3889608640, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.25.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3889625152, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.25.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3908499648, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.25.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3908504384, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.25.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3908509120, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.25.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3908513856, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.25.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3908518592, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.26.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3908523328, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.26.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3913831872, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.26.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3913836608, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.26.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3919145152, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.26.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3919149888, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.26.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3924458432, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.26.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3924463168, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.26.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3929771712, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.26.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3929776448, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.26.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 3948650944, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.26.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3948667456, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.26.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3967541952, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.26.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3967546688, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.26.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3967551424, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.26.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3967556160, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.26.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3967560896, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.27.attention.attention.query.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3967565632, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.27.attention.attention.query.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3972874176, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.27.attention.attention.key.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3972878912, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.27.attention.attention.key.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3978187456, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.27.attention.attention.value.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3978192192, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.27.attention.attention.value.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3983500736, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.27.attention.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 1327104, |
| "bytes": 5308416, |
| "storage_offset": 0, |
| "checkpoint_offset": 3983505472, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.27.attention.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 3988814016, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.27.intermediate.dense.weight": { |
| "shape": [ |
| 4096, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 3988818752, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.27.intermediate.dense.bias": { |
| "shape": [ |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 4096, |
| "bytes": 16384, |
| "storage_offset": 0, |
| "checkpoint_offset": 4007693248, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.27.output.dense.weight": { |
| "shape": [ |
| 1152, |
| 4096 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 4096, |
| 1 |
| ], |
| "numel": 4718592, |
| "bytes": 18874368, |
| "storage_offset": 0, |
| "checkpoint_offset": 4007709760, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.27.output.dense.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 4026584256, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.27.layernorm_before.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 4026588992, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.27.layernorm_before.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 4026593728, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.27.layernorm_after.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 4026598464, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_layers.27.layernorm_after.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 4026603200, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_norm.weight": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 4026607936, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_norm.bias": { |
| "shape": [ |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 1152, |
| "bytes": 4608, |
| "storage_offset": 0, |
| "checkpoint_offset": 4026612672, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_pred.weight": { |
| "shape": [ |
| 768, |
| 1152 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1152, |
| 1 |
| ], |
| "numel": 884736, |
| "bytes": 3538944, |
| "storage_offset": 0, |
| "checkpoint_offset": 4026617408, |
| "contiguous": true |
| }, |
| "ema.decoder.decoder_pred.bias": { |
| "shape": [ |
| 768 |
| ], |
| "dtype": "float32", |
| "stride": [ |
| 1 |
| ], |
| "numel": 768, |
| "bytes": 3072, |
| "storage_offset": 0, |
| "checkpoint_offset": 4030156480, |
| "contiguous": true |
| } |
| } |
| } |
