Instructions to use GroundingPI/GroundAnything-VLM with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use GroundingPI/GroundAnything-VLM with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="GroundingPI/GroundAnything-VLM", trust_remote_code=True)
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("GroundingPI/GroundAnything-VLM", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use GroundingPI/GroundAnything-VLM with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "GroundingPI/GroundAnything-VLM"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "GroundingPI/GroundAnything-VLM",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/GroundingPI/GroundAnything-VLM

SGLang

How to use GroundingPI/GroundAnything-VLM with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "GroundingPI/GroundAnything-VLM" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "GroundingPI/GroundAnything-VLM",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "GroundingPI/GroundAnything-VLM" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "GroundingPI/GroundAnything-VLM",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use GroundingPI/GroundAnything-VLM with Docker Model Runner:
```
docker model run hf.co/GroundingPI/GroundAnything-VLM
```

Skywalker0410 commited on Oct 2

Commit

e0a1aad

verified ·

1 Parent(s): ca15e4c

Document architecture, inference, evaluation, and paper results

Browse files

Files changed (2) hide show

README.md +6 -6
checksums.sha256 +1 -1

README.md CHANGED Viewed

@@ -16,7 +16,7 @@ tags:
 inference: false
 ---
-<p align="center"><img src="assets/logo.png" width="150" alt="GroundAnything logo" /></p>
 # GroundAnything-VLM: Reconciling Parallel Decoding with Precise Visual Grounding at Flash Speed
@@ -24,7 +24,7 @@ inference: false
 **This repository contains the autoregressive GroundAnything-VLM checkpoint.** The family overview and figures below are shared with the DLM page; use the VLM serving recipe for these weights.
-<p align="center"><img src="assets/fig1-teaser.png" width="100%" alt="GroundAnything Figure 1: broad visual grounding and parallel visual evidence extraction" /></p>
 ## 🔗 Quick Links
@@ -103,11 +103,11 @@ Global.
 - **Spatial vocabulary:** 1,000 coordinate tokens shared with semantic labels and protocol markers.
 - **DLM conversion:** the shared decoder and vocabulary head support both causal prediction and bidirectional response-block denoising. A mask token is added for diffusion generation.
-<p align="center"><img src="assets/fig2-architecture.png" width="100%" alt="GroundAnything Figure 2: vision-language architecture and autoregressive-to-diffusion conversion" /></p>
 *Figure 2. Shared model architecture and the conversion to parallel grounding.*
-<p align="center"><img src="assets/fig4-attention-mask.png" width="640" alt="GroundAnything Figure 4: clean-stream causal and noisy-response block attention" /></p>
 *Figure 4. The attention mask used during diffusion conversion. The clean stream uses causal attention. A noisy response block attends bidirectionally within its block and reads the clean conditioning context and strictly preceding clean response blocks. Conversion uses B=32; the illustration uses two-token blocks.*
@@ -190,7 +190,7 @@ External baseline weights and model services are separate from the evaluation to
 ## Quantitative Evaluation Benchmarks
-<p align="center"><img src="assets/fig7-grounding-performance.png" width="100%" alt="GroundAnything Figure 7: GroundAnything and GroundAnything-VLM benchmark overview" /></p>
 *Figure 7. Paper-reported capability overview. Detailed task-level scores, baseline settings, and evaluation protocols are provided in the paper and supplementary material.*
@@ -297,7 +297,7 @@ After a block is complete, a causal forward reconstructs its authoritative KV ca
 The model uses its own shared weights to draft tokens with bidirectional attention and verify them with causal attention. Verification accepts the **longest consecutive matching prefix**, stops at the first mismatch, applies the causal correction, and discards the rejected suffix cache states. The shipped speculative route uses greedy verification; it is not a general stochastic speculative sampler.
-<p align="center"><img src="assets/fig6-self-speculative-decoding.png" width="100%" alt="GroundAnything Figure 6: linear and quadratic self-speculative schedules with shared model weights" /></p>
 *Figure 6. The paper studies two schedules: linear drafting and verification use two model forwards and 2B query tokens per round; quadratic fusion uses one forward after initialization with B(B + 1) query tokens. These counts describe queries and model calls, not total Transformer FLOPs.*

 inference: false
 ---
+<p align="center"><img src="https://huggingface.co/GroundingPI/GroundAnything-VLM/resolve/c367acf62cc8255782e3d19bd8c312f9b7ab35db/assets/logo.png" width="150" alt="GroundAnything logo" /></p>
 # GroundAnything-VLM: Reconciling Parallel Decoding with Precise Visual Grounding at Flash Speed
 **This repository contains the autoregressive GroundAnything-VLM checkpoint.** The family overview and figures below are shared with the DLM page; use the VLM serving recipe for these weights.
+<p align="center"><img src="https://huggingface.co/GroundingPI/GroundAnything-VLM/resolve/c367acf62cc8255782e3d19bd8c312f9b7ab35db/assets/fig1-teaser.png" width="100%" alt="GroundAnything Figure 1: broad visual grounding and parallel visual evidence extraction" /></p>
 ## 🔗 Quick Links
 - **Spatial vocabulary:** 1,000 coordinate tokens shared with semantic labels and protocol markers.
 - **DLM conversion:** the shared decoder and vocabulary head support both causal prediction and bidirectional response-block denoising. A mask token is added for diffusion generation.
+<p align="center"><img src="https://huggingface.co/GroundingPI/GroundAnything-VLM/resolve/c367acf62cc8255782e3d19bd8c312f9b7ab35db/assets/fig2-architecture.png" width="100%" alt="GroundAnything Figure 2: vision-language architecture and autoregressive-to-diffusion conversion" /></p>
 *Figure 2. Shared model architecture and the conversion to parallel grounding.*
+<p align="center"><img src="https://huggingface.co/GroundingPI/GroundAnything-VLM/resolve/c367acf62cc8255782e3d19bd8c312f9b7ab35db/assets/fig4-attention-mask.png" width="640" alt="GroundAnything Figure 4: clean-stream causal and noisy-response block attention" /></p>
 *Figure 4. The attention mask used during diffusion conversion. The clean stream uses causal attention. A noisy response block attends bidirectionally within its block and reads the clean conditioning context and strictly preceding clean response blocks. Conversion uses B=32; the illustration uses two-token blocks.*
 ## Quantitative Evaluation Benchmarks
+<p align="center"><img src="https://huggingface.co/GroundingPI/GroundAnything-VLM/resolve/c367acf62cc8255782e3d19bd8c312f9b7ab35db/assets/fig7-grounding-performance.png" width="100%" alt="GroundAnything Figure 7: GroundAnything and GroundAnything-VLM benchmark overview" /></p>
 *Figure 7. Paper-reported capability overview. Detailed task-level scores, baseline settings, and evaluation protocols are provided in the paper and supplementary material.*
 The model uses its own shared weights to draft tokens with bidirectional attention and verify them with causal attention. Verification accepts the **longest consecutive matching prefix**, stops at the first mismatch, applies the causal correction, and discards the rejected suffix cache states. The shipped speculative route uses greedy verification; it is not a general stochastic speculative sampler.
+<p align="center"><img src="https://huggingface.co/GroundingPI/GroundAnything-VLM/resolve/c367acf62cc8255782e3d19bd8c312f9b7ab35db/assets/fig6-self-speculative-decoding.png" width="100%" alt="GroundAnything Figure 6: linear and quadratic self-speculative schedules with shared model weights" /></p>
 *Figure 6. The paper studies two schedules: linear drafting and verification use two model forwards and 2B query tokens per round; quadratic fusion uses one forward after initialization with B(B + 1) query tokens. These counts describe queries and model calls, not total Transformer FLOPs.*

checksums.sha256 CHANGED Viewed

@@ -1,5 +1,5 @@
 20c797ce19af0c17de52c6afb144644768a591c521655f5ebf5712c9850f2887  LICENSE
-87d6a4d7dc208402f51849bfa0583ef64e742100ef5ad5ebe20c99735c68d2c9  README.md
 7316e325dde0dd407bcc80ed5d3d080223e54572a4685b2aa1ee52041e41d466  added_tokens.json
 a0bc6f6fc7a29a80017a433e8f03a1cc1236e838a944a2d034295a60c4f2fddb  chat_template.jinja
 2a23a7cf76128488c8866582bab06ffb1db387f7f7de82e06fafea65514ce10d  config.json

 20c797ce19af0c17de52c6afb144644768a591c521655f5ebf5712c9850f2887  LICENSE
+94a9cd2753a1eca3a19566727433f7d3aa72050e2b4eeb0b25b3280ea82e9f83  README.md
 7316e325dde0dd407bcc80ed5d3d080223e54572a4685b2aa1ee52041e41d466  added_tokens.json
 a0bc6f6fc7a29a80017a433e8f03a1cc1236e838a944a2d034295a60c4f2fddb  chat_template.jinja
 2a23a7cf76128488c8866582bab06ffb1db387f7f7de82e06fafea65514ce10d  config.json