|
--- |
|
language: en |
|
tags: |
|
- phi-2 |
|
- openassistant |
|
- conversational |
|
license: mit |
|
--- |
|
|
|
# Phi-2 Fine-tuned on OpenAssistant |
|
|
|
This model is a fine-tuned version of Microsoft's Phi-2 model, trained on the OpenAssistant dataset using QLoRA techniques. |
|
|
|
## Model Description |
|
|
|
- **Base Model:** Microsoft Phi-2 |
|
- **Training Data:** OpenAssistant Conversations Dataset |
|
- **Training Method:** QLoRA (Quantized Low-Rank Adaptation) |
|
- **Use Case:** Conversational AI and text generation |
|
|
|
## Usage |
|
|
|
```python |
|
from transformers import AutoModelForCausalLM, AutoTokenizer |
|
|
|
model = AutoModelForCausalLM.from_pretrained("your-username/phi2-finetuned-openassistant") |
|
tokenizer = AutoTokenizer.from_pretrained("your-username/phi2-finetuned-openassistant") |
|
|
|
# Generate text |
|
input_text = "Hello, how are you?" |
|
inputs = tokenizer(input_text, return_tensors="pt") |
|
outputs = model.generate(**inputs, max_length=100) |
|
response = tokenizer.decode(outputs[0], skip_special_tokens=True) |
|
print(response) |
|
``` |
|
|
|
## Training Details |
|
|
|
- Fine-tuned for 1 epoch |
|
- Used 4-bit quantization for efficient training |
|
- Implemented gradient checkpointing and mixed precision training |
|
|
|
## Limitations |
|
|
|
- The model inherits limitations from both Phi-2 and the OpenAssistant dataset |
|
- May produce incorrect or biased information |
|
- Should be used with appropriate content filtering and moderation |
|
|
|
## License |
|
|
|
This model is released under the MIT License. |
|
|
