Instructions to use play0718/EgoAction-LLaVA-OV2-8B-SFT-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use play0718/EgoAction-LLaVA-OV2-8B-SFT-v1 with Transformers:

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

pipe = pipeline("image-text-to-text", model="play0718/EgoAction-LLaVA-OV2-8B-SFT-v1", 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 AutoModelForImageTextToText
model = AutoModelForImageTextToText.from_pretrained("play0718/EgoAction-LLaVA-OV2-8B-SFT-v1", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use play0718/EgoAction-LLaVA-OV2-8B-SFT-v1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "play0718/EgoAction-LLaVA-OV2-8B-SFT-v1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "play0718/EgoAction-LLaVA-OV2-8B-SFT-v1",
		"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/play0718/EgoAction-LLaVA-OV2-8B-SFT-v1

SGLang

How to use play0718/EgoAction-LLaVA-OV2-8B-SFT-v1 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 "play0718/EgoAction-LLaVA-OV2-8B-SFT-v1" \
    --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": "play0718/EgoAction-LLaVA-OV2-8B-SFT-v1",
		"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 "play0718/EgoAction-LLaVA-OV2-8B-SFT-v1" \
        --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": "play0718/EgoAction-LLaVA-OV2-8B-SFT-v1",
		"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 play0718/EgoAction-LLaVA-OV2-8B-SFT-v1 with Docker Model Runner:
```
docker model run hf.co/play0718/EgoAction-LLaVA-OV2-8B-SFT-v1
```

You need to agree to share your Hugging Face information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

EgoAction LLaVA-OV2-8B SFT v1

Private inference export of the completed plain EgoAction SFT run, checkpoint-322 (epoch 2; lowest validation loss 0.206548). The selected checkpoint passed the training run's finite-weight audit.

Base: lmms-lab-encoder/LLaVA-OneVision-2-8B-Instruct, revision 878cde7e04fd962a6082951d8f83083a73fea588. This is the successfully retrained single-node r2 checkpoint. Vision encoder frozen; language model and vision merger fully fine-tuned. No GRPO or auxiliary CareEgo attention loss.

Training reused the EgoAction SFT v1 dataset (5,146 training / 208 validation examples), with 10 FPS, at most 512 frames, and 65,536–131,072 requested pixels/frame. Targets use start_s/end_s and include placeholder segments. The training messages are in training_prompt.json; supply your own video and use the native OV2 processor's target_fps=10, max_frames=512 options. Qwen-specific template options in that reference file are provenance, not OV2 processor arguments.

Original weights, configurations, tokenizer, chat template and custom model/processor Python files are preserved byte-for-byte. SHA-256 hashes and source identity are in export_manifest.json. Optimizer states and training examples are excluded.

Loading

The training runtime used Transformers 5.7.0, huggingface-hub 1.5.0, tokenizers 0.23.0rc0 and decord 0.6.0. Loading requires the included custom code; review it before enabling trust_remote_code.

from transformers import AutoModelForImageTextToText, AutoProcessor
repo = "play0718/EgoAction-LLaVA-OV2-8B-SFT-v1"
processor = AutoProcessor.from_pretrained(repo, trust_remote_code=True, token=True)
model = AutoModelForImageTextToText.from_pretrained(
    repo, trust_remote_code=True, dtype="auto", device_map="auto", token=True
)
model.config.use_cache = True
model.config.text_config.use_cache = True

Access requires authorization to this private repository. Configuration and processor loading were checked locally; the upload does not constitute a new GPU inference evaluation.

Downloads last month: -

Safetensors

Model size

9B params

Tensor type

BF16

Model tree for play0718/EgoAction-LLaVA-OV2-8B-SFT-v1

Base model

lmms-lab-encoder/LLaVA-OneVision-2-8B-Instruct

Finetuned

(5)

this model