|
|
--- |
|
|
license: llama2 |
|
|
--- |
|
|
# Llama2 fine tuned in Intel Hardware using peft and Lora |
|
|
|
|
|
**Description :** Meta's Llama 2 is a transformer-based model tailored for converting natural language instructions into Python code snippets. This model has been optimized for efficient deployment on resource-constrained hardware through techniques such as LORA (Low-Rank Adaptation) and QLORA (Quantized Low-Rank Adaptation), enabling 4-bit quantization without sacrificing performance. Leveraging advanced optimization libraries, such as Intel's Accelerate and Extension for PyTorch, Meta's Llama 2 offers streamlined fine-tuning and inference on Intel Xeon Scalable processors. |
|
|
|
|
|
**Usage :** To utilize Meta's Llama 2 finetuned using the python code snippets, simply load the model using the Hugging Face Transformers library. Ensure compatibility with the prompt template structure: s [inst] instruction [\inst] answer s. Fine-tune the model using the Hugging Face Trainer class, specifying training configurations and leveraging Intel hardware and oneAPI optimization libraries for enhanced performance. |
|
|
|
|
|
**Use in Transformers** |
|
|
```python |
|
|
from transformers import AutoTokenizer, AutoModelForCausalLM |
|
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained("Smd-Arshad/Llama-python-finetuned") |
|
|
model = AutoModelForCausalLM.from_pretrained("Smd-Arshad/Llama-python-finetuned") |
|
|
|
|
|
``` |
