Instructions to use prithivMLmods/QwQ-Math-IO-500M with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use prithivMLmods/QwQ-Math-IO-500M with Transformers:

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

pipe = pipeline("text-generation", model="prithivMLmods/QwQ-Math-IO-500M")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("prithivMLmods/QwQ-Math-IO-500M")
model = AutoModelForCausalLM.from_pretrained("prithivMLmods/QwQ-Math-IO-500M")
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 Settings

vLLM

How to use prithivMLmods/QwQ-Math-IO-500M with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "prithivMLmods/QwQ-Math-IO-500M"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "prithivMLmods/QwQ-Math-IO-500M",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/prithivMLmods/QwQ-Math-IO-500M

SGLang

How to use prithivMLmods/QwQ-Math-IO-500M 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 "prithivMLmods/QwQ-Math-IO-500M" \
    --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": "prithivMLmods/QwQ-Math-IO-500M",
		"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 "prithivMLmods/QwQ-Math-IO-500M" \
        --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": "prithivMLmods/QwQ-Math-IO-500M",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use prithivMLmods/QwQ-Math-IO-500M with Docker Model Runner:
```
docker model run hf.co/prithivMLmods/QwQ-Math-IO-500M
```

prithivMLmods commited on Jan 11, 2025

Commit

79ebb17

verified ·

1 Parent(s): 81ab1a8

Update README.md

Browse files

Files changed (1) hide show

README.md +0 -17

README.md CHANGED Viewed

@@ -27,18 +27,6 @@ QwQ-Math-IO-500M is a fine-tuned variant of Qwen2.5-0.5B, specifically optimized
    - Multilingual capabilities (over 29 languages).
 4. **Optimized for Long Context**: Supports input contexts up to 128K tokens with generation capability up to 8K tokens.
----
-## **Datasets Used**
-The model was fine-tuned on high-quality datasets explicitly curated for Chain of Thought (CoT) reasoning, mathematical problem-solving, and long-context tasks. Notable datasets include:
-1. **[amphora/QwQ-LongCoT-130K](https://huggingface.co/datasets/amphora/QwQ-LongCoT-130K)**: 133k samples focused on complex CoT reasoning.
-2. **[qingy2024/QwQ-LongCoT-Verified-130K](https://huggingface.co/datasets/qingy2024/QwQ-LongCoT-Verified-130K)**: 467k verified samples emphasizing detailed step-by-step reasoning.
-3. **[gghfez/QwQ-LongCoT-130K-cleaned](https://huggingface.co/datasets/gghfez/QwQ-LongCoT-130K-cleaned)**: 125k cleaned samples for high-accuracy reasoning tasks.
----
 ## **Running the Model**
 To run the model using the Transformers library:
@@ -63,9 +51,6 @@ input_ids = tokenizer(input_text, return_tensors="pt").to("cuda")
 outputs = model.generate(**input_ids, max_new_tokens=100)
 print(tokenizer.decode(outputs[0], skip_special_tokens=True))
 ```
----
 ## **Limitations**
 1. **Bias and Fairness**: Despite fine-tuning efforts, biases from the training data may persist. Users should critically assess model outputs.
@@ -75,8 +60,6 @@ print(tokenizer.decode(outputs[0], skip_special_tokens=True))
 5. **Safety Considerations**: Safety alignment has been performed, but users should monitor outputs to avoid inappropriate content.
 6. **Resource Requirements**: Running the model efficiently requires a GPU with sufficient memory.
----
 ## **Intended Use Cases**
 1. **Mathematical Assistance**: Solving equations, performing calculations, and explaining mathematical concepts.

    - Multilingual capabilities (over 29 languages).
 4. **Optimized for Long Context**: Supports input contexts up to 128K tokens with generation capability up to 8K tokens.
 ## **Running the Model**
 To run the model using the Transformers library:
 outputs = model.generate(**input_ids, max_new_tokens=100)
 print(tokenizer.decode(outputs[0], skip_special_tokens=True))
 ```
 ## **Limitations**
 1. **Bias and Fairness**: Despite fine-tuning efforts, biases from the training data may persist. Users should critically assess model outputs.
 5. **Safety Considerations**: Safety alignment has been performed, but users should monitor outputs to avoid inappropriate content.
 6. **Resource Requirements**: Running the model efficiently requires a GPU with sufficient memory.
 ## **Intended Use Cases**
 1. **Mathematical Assistance**: Solving equations, performing calculations, and explaining mathematical concepts.