Instructions to use prithivMLmods/Sombrero-QwQ-32B-Elite11 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use prithivMLmods/Sombrero-QwQ-32B-Elite11 with Transformers:

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

pipe = pipeline("text-generation", model="prithivMLmods/Sombrero-QwQ-32B-Elite11")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("prithivMLmods/Sombrero-QwQ-32B-Elite11")
model = AutoModelForCausalLM.from_pretrained("prithivMLmods/Sombrero-QwQ-32B-Elite11", device_map="auto")
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/Sombrero-QwQ-32B-Elite11 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "prithivMLmods/Sombrero-QwQ-32B-Elite11"
# 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/Sombrero-QwQ-32B-Elite11",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/prithivMLmods/Sombrero-QwQ-32B-Elite11

SGLang

How to use prithivMLmods/Sombrero-QwQ-32B-Elite11 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/Sombrero-QwQ-32B-Elite11" \
    --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/Sombrero-QwQ-32B-Elite11",
		"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/Sombrero-QwQ-32B-Elite11" \
        --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/Sombrero-QwQ-32B-Elite11",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use prithivMLmods/Sombrero-QwQ-32B-Elite11 with Docker Model Runner:
```
docker model run hf.co/prithivMLmods/Sombrero-QwQ-32B-Elite11
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Sombrero-QwQ-32B-Elite11

Sombrero-QwQ-32B-Elite11 is based on the QwQ 32B architecture by Qwen, optimized for Streamlined Memory Optimization and enhanced explanatory, mathematical problem-solving, and reasoning capabilities. This model is particularly effective for coding purposes, avoiding unwanted textual token generation and ensuring efficiency in structured programming outputs.

Key Improvements

Optimized Memory Utilization: Designed to minimize computational overhead while maintaining high accuracy and response coherence.
Advanced Problem-Solving: Excels in mathematical reasoning, step-by-step solutions, and logical deductions.
Superior Coding Capabilities: Fine-tuned for various programming languages, assisting in debugging, generating code snippets, and optimizing algorithms.
Enhanced Explanatory Depth: Provides structured, well-organized explanations for complex queries across different domains.
Long-Context Processing: Supports up to 256K tokens for input and can generate up to 12K tokens in a single output, making it ideal for extensive documentation and detailed responses.
Multilingual Proficiency: Supports over 35 languages, including English, Chinese, French, Spanish, German, Russian, Japanese, Arabic, and more.

Quickstart with Transformers

Here is a code snippet demonstrating how to load the tokenizer and model for streamlined memory-efficient inference:

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "prithivMLmods/Sombrero-QwQ-32B-Elite11"

model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "Write an optimized Python function for matrix multiplication."
messages = [
    {"role": "system", "content": "You are an AI assistant specializing in coding and problem-solving."},
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=512
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]

Intended Use

Coding and Development Assistance:
- Generates optimized code snippets for multiple programming languages.
- Assists with debugging, refactoring, and explaining algorithms.
- Converts pseudocode to functional implementations efficiently.
Mathematical and Logical Problem-Solving:
- Excels in step-by-step explanations for complex mathematical problems.
- Generates proofs, formulas, and structured reasoning for numerical analysis.
Explanatory and Technical Writing:
- Ideal for generating technical documentation, research summaries, and structured reports.
- Provides detailed breakdowns of complex topics in an easy-to-understand manner.
AI-Powered Conversational Agents:
- Enhances chatbot interactions with accurate, structured, and contextually relevant responses.
- Adapts to different conversational styles while maintaining coherence.
Multilingual Applications:
- Supports multilingual responses for global usability.
- Capable of programming language translations and text-to-code conversions.
Long-Form Content Generation:
- Capable of generating extensive articles, research papers, and code documentation without losing coherence.

Limitations

High Computational Requirements:
- Requires high-memory GPUs or TPUs for optimal performance, especially with long-context processing.
Potential Bias in Outputs:
- Although optimized for neutrality, responses may reflect biases present in training data.
Sensitivity to Prompt Engineering:
- The quality of the response depends on how well the input query is structured.
Error Accumulation in Large Outputs:
- Minor inconsistencies in early responses can propagate through long-form content.
Limited Awareness of Real-Time Data:
- Lacks direct access to real-time updates, news, or dynamic internet data beyond its training cutoff.