qwenredux

qwenredux is a custom Qwen3Redux language-model checkpoint. Its configuration identifies Qwen/Qwen3-0.6B as the backbone and adds symbolic-layer, memory, and iterative-reasoning configuration. It is not a standard Qwen3 checkpoint and should not be treated as a drop-in replacement for one.

This card documents the published repository artifacts. Architecture names and configuration fields describe the implementation, not demonstrated reasoning ability or benchmark performance.

Architecture and configuration

Values below are recorded in config.json. A configured sequence limit is not evidence of tested long-context quality; a configured dtype is not a report of training precision.

Setting Published value
Architecture Qwen3Redux
Model type qwen3_redux
Backbone Qwen/Qwen3-0.6B
Backbone revision c1899de289a04d12100db370d81485cdf75e47ca
Transformer layers 28
Hidden size 1024
Intermediate size 3072
Attention heads 16
Key/value heads 8
Symbolic layers 2
Symbolic dimension 256
Symbolic heads 4
Memory size 256
Thought depth 3
Reasoning steps 3
Discrepancy samples 8
Intent count 6
Maximum position embeddings 40960
Vocabulary size 151936
Configured dtype float32
Recorded Transformers version 4.57.6
Use cache true

This export configures 3 reasoning steps. It is a separate repository from Qwen3_Redux-0.6B; a shared backbone identifier does not establish identical weights, training history, or capabilities.

Repository contents and loading

The repository provides checkpoint weights, configuration, tokenizer assets, and a chat template. The inspected file listing does not include the custom Qwen3Redux implementation, and config.json has no auto_map entry.

Loading requires the matching custom model/configuration classes and their dependencies. A stock AutoModelForCausalLM.from_pretrained(...) example is therefore not provided as a supported recipe. Replacing model_type with qwen3 would not reproduce the custom architecture. Obtain the matching implementation before attempting inference or further training.

Training and evaluation

The inspected repository does not document a complete training recipe or reproducible evaluation report. Training datasets and their licenses, data preprocessing, training duration/token count, optimizer settings, hardware, and checkpoint-selection criteria are not specified here. No benchmark scores, comparative improvements, or production-readiness claims are asserted.

Uses and limitations

This artifact can support research into the published architecture and checkpoint once its runtime requirements are satisfied. Before using generated output in an application, evaluate accuracy, bias, memorization, and safety for that application. Language-model output may be incorrect, biased, or unsafe. Do not rely on it without independent verification for consequential decisions.

License

No repository-wide model license is declared in the existing card metadata. This card does not assign a new license or infer one from a backbone, framework, or dependency. Confirm applicable terms with the repository owner before redistribution or deployment.

Source artifacts

Downloads last month
370
Safetensors
Model size
0.6B params
Tensor type
I64
路
F32
路
BOOL
路
Inference Providers NEW
This model isn't deployed by any Inference Provider. 馃檵 Ask for provider support