Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

WIP: fine-tuning #52

Draft
wants to merge 4 commits into
base: main
Choose a base branch
from
+1,980 −0
Draft

WIP: fine-tuning #52

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
f0e8498
WIP: fine-tuning
haesleinhuepf Oct 29, 2023
0ab526c
WIP: fine tuning from manually curated lists of questions and answers
haesleinhuepf Oct 29, 2023
43198be
improved QA formulation
haesleinhuepf Oct 29, 2023
b8ffeb3
add pyclesperanto_prototype QA
haesleinhuepf Oct 29, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
206 changes: 206 additions & 0 deletions demo/fine-tuning/10_extract_from_notebooks.ipynb
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,206 @@
{
"cells": [
{
"cell_type": "markdown",
"id": "2e3da9ae-74e1-4083-a793-39e083a84a29",
"metadata": {},
"source": [
"# Extract text and code from notebooks\n",
"In this notebook we will semi-automatically iterate through Python Jupyter Notebooks and extract a list of text-code pairs that might be useful for LLM fine tuning.\n",
"\n",
"The resulting data is stored in a jsonl file."
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "2e12c7f0-e5ff-418c-bd4d-bf78a09ae588",
"metadata": {},
"outputs": [],
"source": [
"import json\n",
"import openai\n",
"import time\n",
"from bia_bob import bob\n",
"import os\n",
"from bia_bob._utilities import filter_out_blacklist, save_jsonl_file\n",
"import json\n",
"import ipywidgets as widgets\n",
"from IPython.display import display"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "b1d52773-2153-4f70-8c58-0a8aaf40b53e",
"metadata": {},
"outputs": [],
"source": [
"training_data = []"
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "b80ec64c-e1ec-46c6-88f4-eebe6042fb6a",
"metadata": {},
"outputs": [],
"source": [
"def parse_notebook(notebook_path:str):\n",
"\n",
" import nbformat\n",
" \n",
" # Reading the notebook\n",
" with open(notebook_path, 'r', encoding='utf-8') as f:\n",
" notebook = nbformat.read(f, as_version=4)\n",
" \n",
" metadata = notebook['metadata']\n",
" \n",
" last_text = \"\"\n",
" last_code = \"\"\n",
" \n",
" first_text = \"\"\n",
" first_code = \"\"\n",
" all_code = \"\"\n",
" \n",
" cells = notebook['cells']\n",
" for cell in cells:\n",
" cell_type = cell['cell_type']\n",
" if cell_type == 'code':\n",
" # Access code cell\n",
" last_code = last_code + \"\\n\\n\" + cell['source']\n",
" elif cell_type == 'markdown':\n",
" \n",
" if len(last_code) > 0:\n",
" # remove inital line breaks, tabs and spaces\n",
" while last_code[0] in [\"\\n\", \"\\t\", \" \"]:\n",
" last_code = last_code[1:]\n",
" if len(last_code) == 0:\n",
" break\n",
" \n",
" #print(\"----\", last_code[:6])\n",
" if not (last_code.startswith(\"from\") or last_code.startswith(\"import\")):\n",
" # we skip the first block which contains a lot of introductory text and import statements\n",
" training_data.append({\n",
" \"imports\":first_code,\n",
" \"text\":last_text, \n",
" \"code\":last_code\n",
" })\n",
" else:\n",
" # but we keep it for later\n",
" first_text = last_text\n",
" first_code = last_code\n",
" all_code = all_code + \"\\n\\n\" + last_code\n",
" last_code = \"\"\n",
" last_text = \"\"\n",
"\n",
" \n",
" # Access markdown cell\n",
" last_text = last_text + \"\\n\\n\" + cell['source']"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "4881eab3-ae03-4f5e-8ed2-50bdeed35823",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"'346 conversations extracted'"
]
},
"execution_count": 5,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"\n",
"\n",
"def parse_notebooks(directory, extension=\".ipynb\"):\n",
" for root, dirs, files in os.walk(directory):\n",
" for file in files:\n",
" if file.endswith(extension) and \".ipynb_checkpoints\" not in root:\n",
" # print(os.path.join(root, file))\n",
"\n",
" parse_notebook(os.path.join(root, file))\n",
"\n",
"parse_notebooks(r\"C:\\structure\\code\\pyclesperanto_prototype\\demo\")\n",
"\n",
"f\"{len(training_data)} conversations extracted\""
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "43d955be-02f6-4013-aa5d-668bb4df4b06",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"'198 conversations remaining'"
]
},
"execution_count": 6,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"\n",
"training_data = filter_out_blacklist(training_data, [\n",
" \"napari\",\n",
" \"nbscreenshot\",\n",
" \"def \",\n",
" \"print\",\n",
" \"openai\",\n",
" \"https://\"\n",
"])\n",
"\n",
"f\"{len(training_data)} conversations remaining\""
]
},
{
"cell_type": "code",
"execution_count": 7,
"id": "b088341b-36b6-4e16-b366-060cdedcbf64",
"metadata": {},
"outputs": [],
"source": [
"save_jsonl_file(training_data, \"cle_imports_text_code.jsonl\")"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "89df54c5-682a-424c-acbe-15b7e839e925",
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3 (ipykernel)",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.9.16"
}
},
"nbformat": 4,
"nbformat_minor": 5
}
Loading