Instructions to use CewEhao/EAPO-SFT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use CewEhao/EAPO-SFT with Transformers:

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

pipe = pipeline("image-text-to-text", model="CewEhao/EAPO-SFT")
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)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("CewEhao/EAPO-SFT")
model = AutoModelForMultimodalLM.from_pretrained("CewEhao/EAPO-SFT", device_map="auto")
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?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use CewEhao/EAPO-SFT with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "CewEhao/EAPO-SFT"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "CewEhao/EAPO-SFT",
		"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/CewEhao/EAPO-SFT

SGLang

How to use CewEhao/EAPO-SFT 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 "CewEhao/EAPO-SFT" \
    --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": "CewEhao/EAPO-SFT",
		"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 "CewEhao/EAPO-SFT" \
        --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": "CewEhao/EAPO-SFT",
		"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 CewEhao/EAPO-SFT with Docker Model Runner:
```
docker model run hf.co/CewEhao/EAPO-SFT
```

EAPO-SFT checkpoint 500

This private model repository contains checkpoint-500 from the Distill-HQ cold-start SFT run used in the EAPO project.

Checkpoint identity

Base model: Qwen/Qwen2.5-VL-7B-Instruct
Training stage: Distill-HQ supervised cold start
Global step: 500
Recorded epoch: 0.4699248120300752
Format: Transformers-compatible full model, split across four safetensors shards

This is an intermediate checkpoint from the same cold-start run. It is not the final step-1064 checkpoint selected by the canonical Full EAPO launcher.

Loading

from transformers import AutoProcessor, Qwen2_5_VLForConditionalGeneration

model_id = "CewEhao/EAPO-SFT"
processor = AutoProcessor.from_pretrained(model_id)
model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
    model_id,
    torch_dtype="auto",
    device_map="auto",
)

Access requires authorization because this repository is private.

Included files

The repository contains model weights, their index, model and generation configuration, processor configuration, tokenizer files, and the chat template.

Training-only state is deliberately excluded: no trainer_state.json, training_args.bin, optimizer state, scheduler state, logs, datasets, credentials, or machine-local paths are published.

Reproducibility scope

This artifact reproduces the model state at step 500. Exact continuation of the original trainer state is not supported because trainer and optimizer state are intentionally excluded. EAPO training code is maintained separately at https://github.com/Echochef/EAPO.

Downloads last month: 16

Safetensors

Model size

849k params

Tensor type

BF16

Model tree for CewEhao/EAPO-SFT

Base model

Qwen/Qwen2.5-VL-7B-Instruct

Finetuned

(1215)

this model