File size: 2,078 Bytes
96a8e58
 
2e3505c
 
 
 
 
 
 
 
7aa76f9
96a8e58
2e3505c
 
 
 
 
 
 
 
 
 
39409e1
2e3505c
 
 
 
 
 
 
 
8a8a74e
2e3505c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: other
language:
- en
pipeline_tag: text2text-generation
tags:
- alpaca
- llama
- chat
- gpt4
inference: false
---

This is a 4bit 128g GPTQ of [chansung's gpt4-alpaca-lora-13b](https://huggingface.co/chansung/gpt4-alpaca-lora-13b).

More details will be put in this README tomorrow.  Until then, please see one of my other GPTQ repos for more instructions.

Command to create was:
```
cd gptq-safe && CUDA_VISIBLE_DEVICES=0 python3 llama.py /content/gpt4-alpaca-lora-13B-HF c4 --wbits 4 --true-sequential --act-order --groupsize 128 --save_safetensors /content/gpt4-alpaca-lora-13B-GPTQ-4bit-128g.safetensors 
```

Note that  as `--act-order` was used, this will not work with ooba's fork of GPTQ. You must use the qwopqwop repo as of April 13th.

Command to clone the correct GPTQ-for-LLaMa repo for inference using `llama_inference.py`, or in `text-generation-webui`:
```
git clone -n  https://github.com/qwopqwop200/GPTQ-for-LLaMa gptq-safe
cd gptq-safe
git checkout 58c8ab4c7aaccc50f507fd08cce941976affe5e0
```

There is also a `no-act-order.safetensors` file which will work with oobabooga's fork of GPTQ-for-LLaMa; it does not require the latest GPTQ code.

# Original model card is below

This repository comes with LoRA checkpoint to make LLaMA into a chatbot like language model. The checkpoint is the output of instruction following fine-tuning process with the following settings on 8xA100(40G) DGX system.
- Training script: borrowed from the official [Alpaca-LoRA](https://github.com/tloen/alpaca-lora) implementation
- Training script:
```shell
python finetune.py \
    --base_model='decapoda-research/llama-30b-hf' \
    --data_path='alpaca_data_gpt4.json' \
    --num_epochs=10 \
    --cutoff_len=512 \
    --group_by_length \
    --output_dir='./gpt4-alpaca-lora-30b' \
    --lora_target_modules='[q_proj,k_proj,v_proj,o_proj]' \
    --lora_r=16 \
    --batch_size=... \
    --micro_batch_size=...
```

You can find how the training went from W&B report [here](https://wandb.ai/chansung18/gpt4_alpaca_lora/runs/w3syd157?workspace=user-chansung18).