---
library_name: peft
base_model: TheBloke/Llama-2-7b-Chat-GPTQ
pipeline_tag: text-generation
inference: false
license: openrail
language:
- en
datasets:
- flytech/python-codes-25k
tags:
- text2code
- LoRA
- GPTQ
- Llama-2-7B-Chat
- text2python
- instruction2code
---

# Llama-2-7b-Chat-GPTQ fine-tuned on PYTHON-CODES-25K

Generate Python code that accomplishes the task instructed.


## LoRA Adpater Head

### Description

Parameter Efficient Finetuning(PEFT) a 4bit quantized Llama-2-7b-Chat from TheBloke/Llama-2-7b-Chat-GPTQ on flytech/python-codes-25k dataset.

- **Language(s) (NLP):** English
- **License:** openrail
- **Qunatization:** GPTQ 4bit
- **PEFT:** LoRA
- **Finetuned from model [TheBloke/Llama-2-7b-Chat-GPTQ](https://huggingface.co/TheBloke/Llama-2-7B-Chat-GPTQ)**
- **Dataset:** [flytech/python-codes-25k](https://huggingface.co/datasets/flytech/python-codes-25k)

## Intended uses & limitations

Addressing the efficay of Quantization and PEFT. Implemented as a personal Project.

### How to use

```
The quantized model is finetuned as PEFT. We have the trained Adapter.
Merging LoRA adapater with GPTQ quantized model is not yet supported.
So instead of loading a single finetuned model, we need to load the base
model and merge the finetuned adapter on top.
```

```python
instruction = """model_input = "Help me set up my daily to-do list!""""
```
```python
from peft import PeftModel, PeftConfig
from transformers import AutoModelForCausalLM

config = PeftConfig.from_pretrained("SwastikM/Llama-2-7B-Chat-text2code")
model = AutoModelForCausalLM.from_pretrained("TheBloke/Llama-2-7b-Chat-GPTQ")
model = PeftModel.from_pretrained(model, "SwastikM/Llama-2-7B-Chat-text2code")
tokenizer = AutoTokenizer.from_pretrained("SwastikM/Llama-2-7B-Chat-text2code")

inputs = tokenizer(instruction, return_tensors="pt").input_ids.to('cuda')
outputs = model.generate(inputs, max_new_tokens=500, do_sample=False, num_beams=1)
code = tokenizer.decode(outputs[0], skip_special_tokens=True)

print(code)
```

### Size Comparison

The table shows comparison VRAM requirements for loading and training
of FP16 Base Model and 4bit GPTQ quantized model with PEFT.
The value for base model referenced from [Model Memory Calculator](https://huggingface.co/docs/accelerate/main/en/usage_guides/model_size_estimator)
from HuggingFace




| Model                   | Total Size  | Training Using Adam |
| ------------------------|-------------| --------------------| 
| Base Model              | 12.37 GB    | 49.48 GP            |
| 4bitQuantized+PEFT      | 3.90 GB     | 11 GB               |


## Training Details

### Training Data

[gretelai/synthetic_text_to_sql](https://huggingface.co/datasets/gretelai/synthetic_text_to_sql)

  **Train Set Size:** 20000 shuffled randomly

### Training Procedure

HuggingFace Accelerate with Training Loop.


#### Training Hyperparameters

- **Optimizer:** AdamW
- **lr:** 2e-5
- **decay:** linear
- **batch_size:** 4
- **gradient_accumulation_steps:** 8
- **global_step:** 625

 LoraConfig
 - ***r:*** 8
 - ***lora_alpha:*** 32
 - ***target_modules:***  ["k_proj","o_proj","q_proj","v_proj"]
 - ***lora_dropout:*** 0.05


#### Hardware

- **GPU:** P100


## Additional Information

- ***Github:*** [Repository]()
- ***Intro to quantization:*** [Blog](https://huggingface.co/blog/merve/quantization)
- ***Emergent Feature:*** [Academic](https://timdettmers.com/2022/08/17/llm-int8-and-emergent-features)
- ***GPTQ Paper:*** [GPTQ](https://arxiv.org/pdf/2210.17323)
- ***BITSANDBYTES and further*** [LLM.int8()](https://arxiv.org/pdf/2208.07339)

## Acknowledgment

Thanks to [@AMerve Noyan](https://huggingface.co/blog/merve/quantization) for precise intro.
Thanks to [@HuggungFace Team](https://colab.research.google.com/drive/1_TIrmuKOFhuRRiTWN94iLKUFu6ZX4ceb?usp=sharing#scrollTo=vT0XjNc2jYKy) for the notebook on gptq.


## Model Card Authors

Swastik Maiti