File size: 3,121 Bytes
c653d96 b796196 c653d96 4bbd8a4 c653d96 4bbd8a4 c653d96 4bbd8a4 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 |
---
inference: false
language: en
license: llama2
model_type: llama
datasets:
- mlabonne/CodeLlama-2-20k
pipeline_tag: text-generation
tags:
- llama-2
---
# CRIA v1.3
💡 [Article](https://walterteng.com/cria) |
💻 [Github](https://github.com/davzoku/cria) |
📔 Colab [1](https://colab.research.google.com/drive/1rYTs3qWJerrYwihf1j0f00cnzzcpAfYe),[2](https://colab.research.google.com/drive/1Wjs2I1VHjs6zT_GE42iEXsLtYh6VqiJU)
## What is CRIA?
> krē-ə plural crias. : a baby llama, alpaca, vicuña, or guanaco.
<p align="center">
<img src="https://raw.githubusercontent.com/davzoku/cria/main/assets/icon-512x512.png" width="300" height="300" alt="Cria Logo"> <br>
<i>or what ChatGPT suggests, <b>"Crafting a Rapid prototype of an Intelligent llm App using open source resources"</b>.</i>
</p>
The initial objective of the CRIA project is to develop a comprehensive end-to-end chatbot system, starting from the instruction-tuning of a large language model and extending to its deployment on the web using frameworks such as Next.js.
Specifically, we have fine-tuned the `llama-2-7b-chat-hf` model with QLoRA (4-bit precision) using the [mlabonne/CodeLlama-2-20k](https://huggingface.co/datasets/mlabonne/CodeLlama-2-20k) dataset. This fine-tuned model serves as the backbone for the [CRIA chat](https://chat.walterteng.com) platform.
## 📦 Model Release
CRIA v1.3 comes with several variants.
- [davzoku/cria-llama2-7b-v1.3](https://huggingface.co/davzoku/cria-llama2-7b-v1.3): Merged Model
- [davzoku/cria-llama2-7b-v1.3-GGML](https://huggingface.co/davzoku/cria-llama2-7b-v1.3-GGML): Quantized Merged Model
- [davzoku/cria-llama2-7b-v1.3_peft](https://huggingface.co/davzoku/cria-llama2-7b-v1.3_peft): PEFT adapter
## 🔧 Training
It was trained on a Google Colab notebook with a T4 GPU and high RAM.
### Training procedure
The following `bitsandbytes` quantization config was used during training:
- load_in_8bit: False
- load_in_4bit: True
- llm_int8_threshold: 6.0
- llm_int8_skip_modules: None
- llm_int8_enable_fp32_cpu_offload: False
- llm_int8_has_fp16_weight: False
- bnb_4bit_quant_type: nf4
- bnb_4bit_use_double_quant: False
- bnb_4bit_compute_dtype: float16
### Framework versions
- PEFT 0.4.0
## 💻 Usage
```python
# pip install transformers accelerate
from transformers import AutoTokenizer
import transformers
import torch
model = "davzoku/cria-llama2-7b-v1.3"
prompt = "What is a cria?"
tokenizer = AutoTokenizer.from_pretrained(model)
pipeline = transformers.pipeline(
"text-generation",
model=model,
torch_dtype=torch.float16,
device_map="auto",
)
sequences = pipeline(
f'<s>[INST] {prompt} [/INST]',
do_sample=True,
top_k=10,
num_return_sequences=1,
eos_token_id=tokenizer.eos_token_id,
max_length=200,
)
for seq in sequences:
print(f"Result: {seq['generated_text']}")
```
## References
We'd like to thank:
- [mlabonne](https://huggingface.co/mlabonne) for his article and resources on implementation of instruction tuning
- [TheBloke](https://huggingface.co/TheBloke) for his script for LLM quantization.
|