114 lines
3.5 KiB
Markdown
114 lines
3.5 KiB
Markdown
---
|
|
license: creativeml-openrail-m
|
|
datasets:
|
|
- amphora/QwQ-LongCoT-130K
|
|
language:
|
|
- en
|
|
base_model:
|
|
- Qwen/Qwen2.5-7B-Instruct
|
|
pipeline_tag: text-generation
|
|
library_name: transformers
|
|
tags:
|
|
- Long-CoT
|
|
- Qwen2.5
|
|
- 7B
|
|
- safetensors
|
|
- text-generation-inference
|
|
- QwQ
|
|
- SFT
|
|
- Math
|
|
- Qwen with Questions
|
|
new_version: prithivMLmods/QwQ-LCoT2-7B-Instruct
|
|
---
|
|
|
|
# **QwQ-LCoT-7B-Instruct Model File**
|
|
|
|
The QwQ-LCoT-7B-Instruct is a fine-tuned language model designed for advanced reasoning and instruction-following tasks. It leverages the Qwen2.5-7B base model and has been fine-tuned on the amphora/QwQ-LongCoT-130K dataset, focusing on chain-of-thought (CoT) reasoning. This model is optimized for tasks requiring logical reasoning, detailed explanations, and multi-step problem-solving, making it ideal for applications such as instruction-following, text generation, and complex reasoning tasks.
|
|
|
|
## Quickstart with Transformers
|
|
|
|
Here provides a code snippet with `apply_chat_template` to show you how to load the tokenizer and model and how to generate contents.
|
|
|
|
```python
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer
|
|
|
|
model_name = "prithivMLmods/QwQ-LCoT-7B-Instruct"
|
|
|
|
model = AutoModelForCausalLM.from_pretrained(
|
|
model_name,
|
|
torch_dtype="auto",
|
|
device_map="auto"
|
|
)
|
|
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
|
|
|
prompt = "How many r in strawberry."
|
|
messages = [
|
|
{"role": "system", "content": "You are a helpful and harmless assistant. You are Qwen developed by Alibaba. You should think step-by-step."},
|
|
{"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]
|
|
```
|
|
|
|
### **Sample Long CoT:**
|
|
|
|
|
|
|
|
---
|
|
### **Key Features:**
|
|
|
|
1. **Model Size:**
|
|
- **7.62B parameters** (FP16 precision).
|
|
|
|
2. **Model Sharding:**
|
|
- The model weights are split into 4 shards (`safetensors`) for efficient storage and download:
|
|
- `model-00001-of-00004.safetensors` (4.88 GB)
|
|
- `model-00002-of-00004.safetensors` (4.93 GB)
|
|
- `model-00003-of-00004.safetensors` (4.33 GB)
|
|
- `model-00004-of-00004.safetensors` (1.09 GB)
|
|
|
|
3. **Tokenizer:**
|
|
- Byte-pair encoding (BPE) based.
|
|
- Files included:
|
|
- `vocab.json` (2.78 MB)
|
|
- `merges.txt` (1.82 MB)
|
|
- `tokenizer.json` (11.4 MB)
|
|
- Special tokens mapped in `special_tokens_map.json` (e.g., `<pad>`, `<eos>`).
|
|
|
|
4. **Configuration Files:**
|
|
- `config.json`: Defines model architecture and hyperparameters.
|
|
- `generation_config.json`: Settings for inference and text generation tasks.
|
|
|
|
---
|
|
|
|
### **Training Dataset:**
|
|
- **Dataset Name:** [amphora/QwQ-LongCoT-130K](https://huggingface.co/datasets/amphora/QwQ-LongCoT-130K)
|
|
- **Size:** 133k examples.
|
|
- **Focus:** Chain-of-Thought reasoning for complex tasks.
|
|
|
|
---
|
|
|
|
### **Use Cases:**
|
|
1. **Instruction Following:**
|
|
Handle user instructions effectively, even for multi-step tasks.
|
|
|
|
2. **Reasoning Tasks:**
|
|
Perform logical reasoning and generate detailed step-by-step solutions.
|
|
|
|
3. **Text Generation:**
|
|
Generate coherent, context-aware responses.
|
|
--- |