Instructions to use LiquidAI/LFM2.5-VL-450M-Extract with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use LiquidAI/LFM2.5-VL-450M-Extract with Transformers:

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

pipe = pipeline("image-text-to-text", model="LiquidAI/LFM2.5-VL-450M-Extract")
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, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("LiquidAI/LFM2.5-VL-450M-Extract")
model = AutoModelForImageTextToText.from_pretrained("LiquidAI/LFM2.5-VL-450M-Extract")
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 LiquidAI/LFM2.5-VL-450M-Extract with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "LiquidAI/LFM2.5-VL-450M-Extract"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "LiquidAI/LFM2.5-VL-450M-Extract",
		"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/LiquidAI/LFM2.5-VL-450M-Extract

SGLang

How to use LiquidAI/LFM2.5-VL-450M-Extract 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 "LiquidAI/LFM2.5-VL-450M-Extract" \
    --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": "LiquidAI/LFM2.5-VL-450M-Extract",
		"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 "LiquidAI/LFM2.5-VL-450M-Extract" \
        --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": "LiquidAI/LFM2.5-VL-450M-Extract",
		"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 LiquidAI/LFM2.5-VL-450M-Extract with Docker Model Runner:
```
docker model run hf.co/LiquidAI/LFM2.5-VL-450M-Extract
```

LFM2.5-VL-450M-Extract / model_eval /run_eval.sh

zhkleciel

Super-squash branch 'main' using huggingface_hub

c013d55 4 days ago

raw

history blame contribute delete

2.82 kB

	#!/bin/bash
	# Example evaluation script — extraction on local GPU, judges via OpenRouter API.
	# bash run_eval.sh

	set -euo pipefail

	# --- vLLM env workarounds ----------------------------------------------------
	# Needed when EXTRACTION_BACKEND=vllm or "auto". No-op on systems without
	# vLLM / environment-modules.
	# 1. CUDA toolkit on LD_LIBRARY_PATH so flashinfer's GDN/Mamba kernels can
	# dlopen libcudart.so.12 (LFM2.5-VL is a hybrid arch and uses these).
	# 2. Skip flashinfer's sampling kernel — flashinfer 0.6 + CUDA 12.9 trigger
	# an NVCC stub bug (`__cudaLaunch` not declared). PyTorch-native sampler
	# is ~20% slower but works.
	# 3. Skip vLLM 0.21's DeepGEMM autotune warmup (~18 min for MoE/FP8 models).
	module load cuda12.9/toolkit/12.9.1 2>/dev/null \|\| true
	export VLLM_USE_FLASHINFER_SAMPLER="${VLLM_USE_FLASHINFER_SAMPLER:-0}"
	export VLLM_USE_DEEP_GEMM="${VLLM_USE_DEEP_GEMM:-0}"

	# --- OpenRouter API key ------------------------------------------------------
	# Required. Get one at https://openrouter.ai/keys, then either:
	# export OPENROUTER_API_KEY=...
	# in your shell, OR uncomment and set it here.
	#OPENROUTER_API_KEY="sk-or-v1-..."

	# --- Checkpoint --------------------------------------------------------------
	# HF id of the trained model, OR a local merged/LoRA checkpoint dir.
	CHECKPOINT="LiquidAI/LFM2.5-VL-450M-Extract"

	# --- Eval data ---------------------------------------------------------------
	# WDS tar / dir of tars / brace-glob.
	DATA_PATH="./eval_data"

	# Output JSON path.
	OUTPUT="./eval_result.json"

	# --- Sample count ------------------------------------------------------------
	# Number of samples to evaluate. Default 2000 runs the full shipped eval_data
	# (~30 min). Set to 50 for a quick smoke test (~5 min).
	NUM_SAMPLES=2000

	# --- Extraction (local GPU) --------------------------------------------------
	# "auto" tries vLLM first, falls back to HF transformers on init failure.
	EXTRACTION_BACKEND="auto"
	EXTRACTION_BATCH=8

	# --- Judge model (OpenRouter) ------------------------------------------------
	# Any image-capable OpenRouter model id works. Pricing:
	# https://openrouter.ai/models
	VLM_JUDGE_MODEL="qwen/qwen3.5-35b-a3b"

	# Concurrent OpenRouter calls. Lower if you hit rate limits.
	JUDGE_CONCURRENCY=16

	# --- Run ---------------------------------------------------------------------
	LOG_FILE="${LOG_FILE:-./eval_run.log}"
	echo "Logging to: ${LOG_FILE}"
	python run_eval.py \
	--checkpoint-path "${CHECKPOINT}" \
	--data-path "${DATA_PATH}" \
	--output-path "${OUTPUT}" \
	--num-samples "${NUM_SAMPLES}" \
	--extraction-backend "${EXTRACTION_BACKEND}" \
	--extraction-batch "${EXTRACTION_BATCH}" \
	--vlm-judge --vlm-judge-model "${VLM_JUDGE_MODEL}" \
	--judge-concurrency "${JUDGE_CONCURRENCY}" \
	2>&1 \| tee "${LOG_FILE}"