Delete Hugging Face Model Card.txt
Browse files- Hugging Face Model Card.txt +0 -99
Hugging Face Model Card.txt
DELETED
|
@@ -1,99 +0,0 @@
|
|
| 1 |
-
# Model Card for `my-chatbot-llama2-7b`
|
| 2 |
-
|
| 3 |
-
## Model Details
|
| 4 |
-
|
| 5 |
-
* **Model Name**: my-chatbot-llama2-7b
|
| 6 |
-
* **Version**: 1.0.0
|
| 7 |
-
* **License**: MIT
|
| 8 |
-
* **Base Model**: LLaMA 2 7B (Meta)
|
| 9 |
-
* **Languages**: English, Chinese
|
| 10 |
-
* **Author**: Open Community
|
| 11 |
-
* **Repository**: [https://huggingface.co/your-username/my-chatbot-llama2-7b](https://huggingface.co/your-username/my-chatbot-llama2-7b)
|
| 12 |
-
* **Library**: llama.cpp + FastAPI
|
| 13 |
-
|
| 14 |
-
## Model Description
|
| 15 |
-
|
| 16 |
-
`my-chatbot-llama2-7b` is a bilingual chatbot model based on Meta's LLaMA 2 7B, fine-tuned for English and Chinese conversational tasks. It has been trained on large-scale web dialogue and domain-specific Q\&A, emphasizing fluency, informativeness, and adaptability in both languages.
|
| 17 |
-
|
| 18 |
-
* Supports up to 2048 tokens
|
| 19 |
-
* Good at casual chat, multi-turn Q\&A, and domain reasoning
|
| 20 |
-
* Optimized for self-hosting and low-latency environments
|
| 21 |
-
|
| 22 |
-
## Intended Use
|
| 23 |
-
|
| 24 |
-
### Use Cases
|
| 25 |
-
|
| 26 |
-
* Customer support chatbot (web/app)
|
| 27 |
-
* Domain-specific Q\&A (IT, medical, finance)
|
| 28 |
-
* Language-learning assistant
|
| 29 |
-
* Private, local deployment for enterprises
|
| 30 |
-
|
| 31 |
-
### Not Recommended for
|
| 32 |
-
|
| 33 |
-
* Generating harmful, illegal, or deceptive content
|
| 34 |
-
* Unmoderated medical/legal decision-making
|
| 35 |
-
|
| 36 |
-
## How to Use
|
| 37 |
-
|
| 38 |
-
```bash
|
| 39 |
-
# Install llama.cpp and dependencies
|
| 40 |
-
git clone https://github.com/ggerganov/llama.cpp
|
| 41 |
-
cd llama.cpp && make
|
| 42 |
-
pip install fastapi uvicorn
|
| 43 |
-
|
| 44 |
-
# Convert weights
|
| 45 |
-
python convert-llama2-to-gguf.py /path/to/llama2-7b /models/llama2-7b.gguf
|
| 46 |
-
|
| 47 |
-
# Start API server
|
| 48 |
-
uvicorn app:app --host 0.0.0.0 --port 8000 --reload
|
| 49 |
-
|
| 50 |
-
# Test with curl
|
| 51 |
-
curl -X POST http://localhost:8000/generate \
|
| 52 |
-
-H "Content-Type: application/json" \
|
| 53 |
-
-d '{"prompt": "你好，世界！", "token": "your_token_here"}'
|
| 54 |
-
```
|
| 55 |
-
|
| 56 |
-
## Training Data
|
| 57 |
-
|
| 58 |
-
| Source | Type | Samples |
|
| 59 |
-
| ----------------- | --------------------- | ------- |
|
| 60 |
-
| Zhihu, Weibo | Chinese Conversations | 200,000 |
|
| 61 |
-
| Reddit, StackExch | English Conversations | 150,000 |
|
| 62 |
-
| Industry QA | IT, Health, Finance | 50,000 |
|
| 63 |
-
|
| 64 |
-
All datasets are cleaned, normalized, deduplicated, and filtered for profanity.
|
| 65 |
-
|
| 66 |
-
## Evaluation Results
|
| 67 |
-
|
| 68 |
-
| Benchmark | Metric | Score | Notes |
|
| 69 |
-
| ------------- | ------ | ----- | ------------------------------------ |
|
| 70 |
-
| C-Eval | EM | 68.3 | Chinese QA accuracy |
|
| 71 |
-
| GPT4Bot-Bench | F1 | 72.1 | Conversational F1 score |
|
| 72 |
-
| SelfChat Sim | Sim | 0.87 | Self-consistency & diversity measure |
|
| 73 |
-
|
| 74 |
-
## Limitations
|
| 75 |
-
|
| 76 |
-
* May hallucinate facts or generate outdated information (cutoff: Sep 2023)
|
| 77 |
-
* Sensitive to input phrasing
|
| 78 |
-
* No knowledge of post-2023 world events
|
| 79 |
-
|
| 80 |
-
## Ethical Considerations
|
| 81 |
-
|
| 82 |
-
* No user-identifiable data used in training
|
| 83 |
-
* Content filters and moderation are strongly recommended
|
| 84 |
-
* Biases from web data may persist
|
| 85 |
-
|
| 86 |
-
## Citation
|
| 87 |
-
|
| 88 |
-
```bibtex
|
| 89 |
-
@misc{mychatbot2025,
|
| 90 |
-
title = {my-chatbot-llama2-7b: A Self-Hosted Conversational AI},
|
| 91 |
-
author = {Open Community},
|
| 92 |
-
year = {2025},
|
| 93 |
-
howpublished = {\url{https://huggingface.co/your-username/my-chatbot-llama2-7b}}
|
| 94 |
-
}
|
| 95 |
-
```
|
| 96 |
-
|
| 97 |
-
## Changelog
|
| 98 |
-
|
| 99 |
-
* **v1.0.0 (June 2025)**: Initial bilingual release with domain QA support
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|