Instructions to use optimum-intel-internal-testing/tiny-random-qwen38-flash-next with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use optimum-intel-internal-testing/tiny-random-qwen38-flash-next with Transformers:

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

pipe = pipeline("image-text-to-text", model="optimum-intel-internal-testing/tiny-random-qwen38-flash-next")
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 AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("optimum-intel-internal-testing/tiny-random-qwen38-flash-next")
model = AutoModelForMultimodalLM.from_pretrained("optimum-intel-internal-testing/tiny-random-qwen38-flash-next", 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=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use optimum-intel-internal-testing/tiny-random-qwen38-flash-next with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "optimum-intel-internal-testing/tiny-random-qwen38-flash-next"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "optimum-intel-internal-testing/tiny-random-qwen38-flash-next",
		"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/optimum-intel-internal-testing/tiny-random-qwen38-flash-next

SGLang

How to use optimum-intel-internal-testing/tiny-random-qwen38-flash-next 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 "optimum-intel-internal-testing/tiny-random-qwen38-flash-next" \
    --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": "optimum-intel-internal-testing/tiny-random-qwen38-flash-next",
		"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 "optimum-intel-internal-testing/tiny-random-qwen38-flash-next" \
        --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": "optimum-intel-internal-testing/tiny-random-qwen38-flash-next",
		"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 optimum-intel-internal-testing/tiny-random-qwen38-flash-next with Docker Model Runner:
```
docker model run hf.co/optimum-intel-internal-testing/tiny-random-qwen38-flash-next
```

Hidden size and expert sizes 128 for fused MoE on OpenVINO GPU

by rkazants - opened 13 days ago

base: refs/heads/main

←

from: refs/pr/2

Discussion Files changed

+18

-13

rkazants

Optimum Intel Internal Testing org 13 days ago

Increases the hidden size and the expert sizes from 16 to 128 (hidden_size, moe_intermediate_size, shared_expert_intermediate_size, and the vision out_hidden_size that follows the hidden size). Nothing else in the architecture changes.

Why: with expert input dimensions of 16, a 4-bit export with the default group size (128) leaves the experts uncompressed, so OpenVINO GPU does not fuse them into its MoE operation (moe_3gemm_fused_compressed). Smaller group sizes do not help either: the GPU MoE kernel needs a compression group of at least 64. With 128, the default int4 export compresses all experts and they are fused on both CPU (GatherMatmul) and GPU.

Same generator script, same initialization scheme; the layer std is 0.09 (about 1 / sqrt(hidden size)) instead of 0.25.
Same file layout: fused 3D experts, n-gram table shards, MTP head under mtp..
Health check on text, image and video prompts: logits std about 11, median greedy top-2 margins above 1.3, all experts used on 70+ token prompts, QSA drops 23–93% of visible tokens.
The model card is updated with the new sizes and checks.
The model grows from 8M to 71M parameters (33 MB → 289 MB checkpoint), mostly the two 248320 × 128 embedding tables.
An fp32 OpenVINO export of this checkpoint matches transformers token for token on CPU (text and image).

🤖 Generated with Claude Code

Hidden size and expert sizes 128 for fused MoE on OpenVINO GPU0166bfc8

rkazants changed pull request status to merged 13 days ago

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment