| import gradio as gr | |
| import transformers | |
| from torch import bfloat16 | |
| # from dotenv import load_dotenv # if you wanted to adapt this for a repo that uses auth | |
| from threading import Thread | |
| from gradio.themes.utils.colors import Color | |
| #HF_AUTH = os.getenv('HF_AUTH') | |
| #model_id = "stabilityai/StableBeluga2" # 70B parm model based off Llama 2 70B |
| from passlib.context import CryptContext | |
| from jose import jwt | |
| import datetime | |
| SECRET_KEY = "YOUR-SECRET-KEY" # Replace this with your secret key | |
| ALGORITHM = "HS256" | |
| ACCESS_TOKEN_EXPIRE_MINUTES = 30 | |
| pwd_context = CryptContext(schemes=["bcrypt"], deprecated="auto") | |
| class User: |
| import arxiv | |
| import ast | |
| import concurrent | |
| from csv import writer | |
| import openai | |
| import os | |
| import pandas as pd | |
| from PyPDF2 import PdfReader | |
| import requests | |
| from scipy import spatial |
| OUTPUT : | |
| 0.0: | |
| Summary: | |
| { | |
| "subject": "a man", | |
| "characters": ["a man", "a silhouetted figure", "an old friend"], | |
| "locations": ["a dark alley in the heart of the city"], |
| base_model: tiiuae/falcon-7b | |
| base_model_config: tiiuae/falcon-7b | |
| trust_remote_code: true | |
| model_type: AutoModelForCausalLM | |
| tokenizer_type: AutoTokenizer | |
| load_in_8bit: false | |
| load_in_4bit: true | |
| gptq: false | |
| strict: false | |
| push_dataset_to_hub: |
workspace/llm-playground/notebooks/axolotl/runpod/axolotl-falcon-7b-qlora-gsm8k.ipynb
Steps to reproduce :
1 ) Copy config from #4 run-16: 40*2 + xformer into examples/falcon/qlora.yml
2 ) Run cells 1 & 2
3 ) Run !accelerate launch scripts/finetune.py examples/falcon/qlora.yml
| #!/bin/bash | |
| # | |
| # Container source: https://github.com/OpenAccess-AI-Collective/axolotl/blob/main/docker/Dockerfile-runpod | |
| # | |
| # | |
| # To run this in RunPod with `winglian/axolotl-runpod:main-cu118-2.0.0`, set | |
| # Expose HTTP Ports (Max 10): 7860,8888 | |
| # docker command: `bash -c "curl -H 'Cache-Control: no-cache' https://raw.githubusercontent.com/utensil/llm-playground/main/scripts/entry/prepare_ax.sh -sSf | bash"` | |
| # JUPYTER_PASSWORD change to your secret | |
| # HUGGINGFACE_TOKEN change to your token from https://huggingface.co/settings/tokens |
| base_model: openlm-research/open_llama_3b_600bt_preview | |
| base_model_config: openlm-research/open_llama_3b_600bt_preview | |
| model_type: LlamaForCausalLM | |
| tokenizer_type: LlamaTokenizer | |
| load_in_8bit: false | |
| load_in_4bit: true | |
| strict: false | |
| push_dataset_to_hub: | |
| datasets: | |
| - path: teknium/GPT4-LLM-Cleaned |
| # Changing this would change the base model used for training | |
| base_model: huggyllama/llama-7b | |
| # Changing this would change the configuration used for the base model | |
| base_model_config: huggyllama/llama-7b | |
| # Changing this would change the type of model used for training | |
| model_type: LlamaForCausalLM | |
| # Changing this would change the type of tokenizer used for tokenizing text data | |
| tokenizer_type: LlamaTokenizer | |
| # Changing this to false would prevent the model from being loaded in 8-bit precision | |
| load_in_8bit: false |