Instructions to use tugstugi/Qwen2.5-Coder-0.5B-QwQ-draft with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tugstugi/Qwen2.5-Coder-0.5B-QwQ-draft with Transformers:

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

pipe = pipeline("text-generation", model="tugstugi/Qwen2.5-Coder-0.5B-QwQ-draft")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tugstugi/Qwen2.5-Coder-0.5B-QwQ-draft")
model = AutoModelForCausalLM.from_pretrained("tugstugi/Qwen2.5-Coder-0.5B-QwQ-draft")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use tugstugi/Qwen2.5-Coder-0.5B-QwQ-draft with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tugstugi/Qwen2.5-Coder-0.5B-QwQ-draft"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tugstugi/Qwen2.5-Coder-0.5B-QwQ-draft",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tugstugi/Qwen2.5-Coder-0.5B-QwQ-draft

SGLang

How to use tugstugi/Qwen2.5-Coder-0.5B-QwQ-draft 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 "tugstugi/Qwen2.5-Coder-0.5B-QwQ-draft" \
    --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": "tugstugi/Qwen2.5-Coder-0.5B-QwQ-draft",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "tugstugi/Qwen2.5-Coder-0.5B-QwQ-draft" \
        --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": "tugstugi/Qwen2.5-Coder-0.5B-QwQ-draft",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tugstugi/Qwen2.5-Coder-0.5B-QwQ-draft with Docker Model Runner:
```
docker model run hf.co/tugstugi/Qwen2.5-Coder-0.5B-QwQ-draft
```

Tokenizer Details

by qingy2024 - opened Jan 23, 2025

Discussion

qingy2024

Jan 23, 2025

Great work on this!

I was wondering, how did you get the tokenizer to have the same vocab size as QwQ 32B Preview? I would like to do this for some other models too!

If you have a script or just a set of steps to do this, I'd appreciate if you could share it :)

tugstugi

Owner Jan 23, 2025

The tokenizer is actually the same, you only need to change the embedding layer size.

model.resize_token_embeddings(152064)

tugstugi changed discussion status to closed Jan 23, 2025

qingy2024

Jan 31, 2025

So, one more thing – when you do this: model.resize_token_embeddings(152064), does it affect the model's performance in any way, or is the vocabulary just filled up with pad tokens?

Just wondering since I got this warning while running it myself:

The new embeddings will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False`

djuna

Jan 31, 2025

@qingy2024 As long as it can be divided by 64, it should not be a problem right?

qingy2024

Jan 31, 2025

•

edited Jan 31, 2025

@djuna I don't know too much about this and how the model architecture uses the vocab; please enlighten me :)

djuna

Jan 31, 2025

•

edited Jan 31, 2025

@qingy2024 I can only find this
https://x.com/karpathy/status/1621578354024677377

tugstugi

Owner Jan 31, 2025

•

edited Jan 31, 2025

@qingy2024 If the tokenizer is not changed, resize_token_embeddings should have no effect. Here is a test for the embedding:

import torch

with torch.no_grad():
    embed_dim = 5
    old_num_embed, new_num_embed = 5, 8
    old_embeddings = torch.nn.Embedding(old_num_embed, embed_dim)
    new_embeddings = torch.nn.Embedding(new_num_embed, embed_dim, device=old_embeddings.weight.device, dtype=old_embeddings.weight.dtype)
    new_embeddings.weight.data[:old_num_embed, :] = old_embeddings.weight.data[:old_num_embed, :]

    out1 = new_embeddings(torch.LongTensor([range(old_num_embed)]))
    out2 = old_embeddings(torch.LongTensor([range(old_num_embed)]))
    assert torch.allclose(out1, out2)

qingy2024

Jan 31, 2025

Oh, very interesting; thanks for clarifying!

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