graykode / commit-autosuggestions

Go to a project
Toggle navigation Toggle navigation pinning
graykode
4a080ca9 1 parent 3ba668ba

(add) jupyter notebook file in goolge colab for people who dont have nvidia gpu

Inline Side-by-side
Showing 1 changed file with 444 additions and 0 deletions
1 +{
2 + "nbformat": 4,
3 + "nbformat_minor": 0,
4 + "metadata": {
5 + "colab": {
6 + "name": "commit-autosuggestions.ipynb",
7 + "provenance": [],
8 + "collapsed_sections": [],
9 + "toc_visible": true
10 + },
11 + "kernelspec": {
12 + "name": "python3",
13 + "display_name": "Python 3"
14 + },
15 + "accelerator": "GPU"
16 + },
17 + "cells": [
18 + {
19 + "cell_type": "markdown",
20 + "metadata": {
21 + "id": "DZ7rFp2gzuNS"
22 + },
23 + "source": [
24 + "## Start commit-autosuggestions server\n",
25 + "Running flask app server in Google Colab for people without GPU"
26 + ]
27 + },
28 + {
29 + "cell_type": "markdown",
30 + "metadata": {
31 + "id": "d8Lyin2I3wHq"
32 + },
33 + "source": [
34 + "#### Clone github repository"
35 + ]
36 + },
37 + {
38 + "cell_type": "code",
39 + "metadata": {
40 + "id": "e_cu9igvzjcs"
41 + },
42 + "source": [
43 + "!git clone https://github.com/graykode/commit-autosuggestions.git\n",
44 + "%cd commit-autosuggestions\n",
45 + "!pip install -r requirements.txt"
46 + ],
47 + "execution_count": null,
48 + "outputs": []
49 + },
50 + {
51 + "cell_type": "markdown",
52 + "metadata": {
53 + "id": "PFKn5QZr0dQx"
54 + },
55 + "source": [
56 + "#### Download model weights\n",
57 + "\n",
58 + "Download the two weights of model from the google drive through the gdown module.\n",
59 + "1. [Added model](https://drive.google.com/uc?id=1YrkwfM-0VBCJaa9NYaXUQPODdGPsmQY4) : A model trained Code2NL on Python using pre-trained CodeBERT (Feng at al, 2020).\n",
60 + "2. [Diff model](https://drive.google.com/uc?id=1--gcVVix92_Fp75A-mWH0pJS0ahlni5m) : A model retrained by initializing with the weight of model (1), adding embedding of the added and deleted parts(`patch_ids_embedding`) of the code."
61 + ]
62 + },
63 + {
64 + "cell_type": "code",
65 + "metadata": {
66 + "id": "P9-EBpxt0Dp0"
67 + },
68 + "source": [
69 + "!pip install gdown \\\n",
70 + " && gdown \"https://drive.google.com/uc?id=1YrkwfM-0VBCJaa9NYaXUQPODdGPsmQY4\" -O weight/added/pytorch_model.bin \\\n",
71 + " && gdown \"https://drive.google.com/uc?id=1--gcVVix92_Fp75A-mWH0pJS0ahlni5m\" -O weight/diff/pytorch_model.bin"
72 + ],
73 + "execution_count": null,
74 + "outputs": []
75 + },
76 + {
77 + "cell_type": "markdown",
78 + "metadata": {
79 + "id": "org4Gqdv3iUu"
80 + },
81 + "source": [
82 + "#### ngrok setting with flask\n",
83 + "\n",
84 + "Before starting the server, you need to configure ngrok to open this notebook to the outside. I have referred [this jupyter notebook](https://github.com/alievk/avatarify/blob/master/avatarify.ipynb) in detail."
85 + ]
86 + },
87 + {
88 + "cell_type": "code",
89 + "metadata": {
90 + "id": "lZA3kuuG1Crj"
91 + },
92 + "source": [
93 + "!pip install flask-ngrok"
94 + ],
95 + "execution_count": null,
96 + "outputs": []
97 + },
98 + {
99 + "cell_type": "markdown",
100 + "metadata": {
101 + "id": "hR78FRCMcqrZ"
102 + },
103 + "source": [
104 + "Go to https://dashboard.ngrok.com/auth/your-authtoken (sign up if required), copy your authtoken and put it below.\n",
105 + "\n"
106 + ]
107 + },
108 + {
109 + "cell_type": "code",
110 + "metadata": {
111 + "id": "L_mInbOKcoc2"
112 + },
113 + "source": [
114 + "# Paste your authtoken here in quotes\n",
115 + "authtoken = \"21KfrFEW1BptdPPM4SS_7s1Z4HwozyXX9NP2fHC12\""
116 + ],
117 + "execution_count": null,
118 + "outputs": []
119 + },
120 + {
121 + "cell_type": "markdown",
122 + "metadata": {
123 + "id": "QwCN4YFUc0M8"
124 + },
125 + "source": [
126 + "Set your region\n",
127 + "\n",
128 + "Code | Region\n",
129 + "--- | ---\n",
130 + "us | United States\n",
131 + "eu | Europe\n",
132 + "ap | Asia/Pacific\n",
133 + "au | Australia\n",
134 + "sa | South America\n",
135 + "jp | Japan\n",
136 + "in | India"
137 + ]
138 + },
139 + {
140 + "cell_type": "code",
141 + "metadata": {
142 + "id": "p4LSNN2xc0dQ"
143 + },
144 + "source": [
145 + "# Set your region here in quotes\n",
146 + "region = \"jp\"\n",
147 + "\n",
148 + "# Input and output ports for communication\n",
149 + "local_in_port = 5000\n",
150 + "local_out_port = 5000"
151 + ],
152 + "execution_count": null,
153 + "outputs": []
154 + },
155 + {
156 + "cell_type": "code",
157 + "metadata": {
158 + "id": "kg56PVrOdhi1"
159 + },
160 + "source": [
161 + "config =\\\n",
162 + "f\"\"\"\n",
163 + "authtoken: {authtoken}\n",
164 + "region: {region}\n",
165 + "console_ui: False\n",
166 + "tunnels:\n",
167 + " input:\n",
168 + " addr: {local_in_port}\n",
169 + " proto: http \n",
170 + " output:\n",
171 + " addr: {local_out_port}\n",
172 + " proto: http\n",
173 + "\"\"\"\n",
174 + "\n",
175 + "with open('ngrok.conf', 'w') as f:\n",
176 + " f.write(config)"
177 + ],
178 + "execution_count": null,
179 + "outputs": []
180 + },
181 + {
182 + "cell_type": "code",
183 + "metadata": {
184 + "id": "hrWDrw_YdjIy"
185 + },
186 + "source": [
187 + "import time\n",
188 + "from subprocess import Popen, PIPE\n",
189 + "\n",
190 + "# (Re)Open tunnel\n",
191 + "ps = Popen('./scripts/open_tunnel_ngrok.sh', stdout=PIPE, stderr=PIPE)\n",
192 + "time.sleep(3)"
193 + ],
194 + "execution_count": null,
195 + "outputs": []
196 + },
197 + {
198 + "cell_type": "code",
199 + "metadata": {
200 + "id": "pJgdFr0Fdjoq",
201 + "outputId": "3948f70b-d4f3-4ed8-a864-fe5c6df50809",
202 + "colab": {
203 + "base_uri": "https://localhost:8080/"
204 + }
205 + },
206 + "source": [
207 + "# Get tunnel addresses\n",
208 + "try:\n",
209 + " in_addr, out_addr = get_tunnel_adresses()\n",
210 + " print(\"Tunnel opened\")\n",
211 + "except Exception as e:\n",
212 + " [print(l.decode(), end='') for l in ps.stdout.readlines()]\n",
213 + " print(\"Something went wrong, reopen the tunnel\")"
214 + ],
215 + "execution_count": null,
216 + "outputs": [
217 + {
218 + "output_type": "stream",
219 + "text": [
220 + "Opening tunnel\n",
221 + "Something went wrong, reopen the tunnel\n"
222 + ],
223 + "name": "stdout"
224 + }
225 + ]
226 + },
227 + {
228 + "cell_type": "markdown",
229 + "metadata": {
230 + "id": "cEZ-O0wz74OJ"
231 + },
232 + "source": [
233 + "#### Run you server!"
234 + ]
235 + },
236 + {
237 + "cell_type": "code",
238 + "metadata": {
239 + "id": "7PRkeYTL8Y_6"
240 + },
241 + "source": [
242 + "import os\n",
243 + "import torch\n",
244 + "import argparse\n",
245 + "from tqdm import tqdm\n",
246 + "import torch.nn as nn\n",
247 + "from torch.utils.data import TensorDataset, DataLoader, SequentialSampler\n",
248 + "from transformers import (RobertaConfig, RobertaTokenizer)\n",
249 + "\n",
250 + "from commit.model import Seq2Seq\n",
251 + "from commit.utils import (Example, convert_examples_to_features)\n",
252 + "from commit.model.diff_roberta import RobertaModel\n",
253 + "\n",
254 + "from flask import Flask, jsonify, request\n",
255 + "\n",
256 + "MODEL_CLASSES = {'roberta': (RobertaConfig, RobertaModel, RobertaTokenizer)}"
257 + ],
258 + "execution_count": null,
259 + "outputs": []
260 + },
261 + {
262 + "cell_type": "code",
263 + "metadata": {
264 + "id": "CiJKucX17qb4"
265 + },
266 + "source": [
267 + "def get_model(model_class, config, tokenizer, mode):\n",
268 + " encoder = model_class(config=config)\n",
269 + " decoder_layer = nn.TransformerDecoderLayer(\n",
270 + " d_model=config.hidden_size, nhead=config.num_attention_heads\n",
271 + " )\n",
272 + " decoder = nn.TransformerDecoder(decoder_layer, num_layers=6)\n",
273 + " model = Seq2Seq(encoder=encoder, decoder=decoder, config=config,\n",
274 + " beam_size=args.beam_size, max_length=args.max_target_length,\n",
275 + " sos_id=tokenizer.cls_token_id, eos_id=tokenizer.sep_token_id)\n",
276 + "\n",
277 + " assert args.load_model_path\n",
278 + " assert os.path.exists(os.path.join(args.load_model_path, mode, 'pytorch_model.bin'))\n",
279 + "\n",
280 + " model.load_state_dict(\n",
281 + " torch.load(\n",
282 + " os.path.join(args.load_model_path, mode, 'pytorch_model.bin'),\n",
283 + " map_location=torch.device('cpu')\n",
284 + " ),\n",
285 + " strict=False\n",
286 + " )\n",
287 + " return model\n",
288 + "\n",
289 + "def get_features(examples):\n",
290 + " features = convert_examples_to_features(examples, args.tokenizer, args, stage='test')\n",
291 + " all_source_ids = torch.tensor(\n",
292 + " [f.source_ids[:args.max_source_length] for f in features], dtype=torch.long\n",
293 + " )\n",
294 + " all_source_mask = torch.tensor(\n",
295 + " [f.source_mask[:args.max_source_length] for f in features], dtype=torch.long\n",
296 + " )\n",
297 + " all_patch_ids = torch.tensor(\n",
298 + " [f.patch_ids[:args.max_source_length] for f in features], dtype=torch.long\n",
299 + " )\n",
300 + " return TensorDataset(all_source_ids, all_source_mask, all_patch_ids)\n",
301 + "\n",
302 + "def create_app():\n",
303 + " @app.route('/')\n",
304 + " def index():\n",
305 + " return jsonify(hello=\"world\")\n",
306 + "\n",
307 + " @app.route('/added', methods=['POST'])\n",
308 + " def added():\n",
309 + " if request.method == 'POST':\n",
310 + " payload = request.get_json()\n",
311 + " example = [\n",
312 + " Example(\n",
313 + " idx=payload['idx'],\n",
314 + " added=payload['added'],\n",
315 + " deleted=payload['deleted'],\n",
316 + " target=None\n",
317 + " )\n",
318 + " ]\n",
319 + " message = inference(model=args.added_model, data=get_features(example))\n",
320 + " return jsonify(idx=payload['idx'], message=message)\n",
321 + "\n",
322 + " @app.route('/diff', methods=['POST'])\n",
323 + " def diff():\n",
324 + " if request.method == 'POST':\n",
325 + " payload = request.get_json()\n",
326 + " example = [\n",
327 + " Example(\n",
328 + " idx=payload['idx'],\n",
329 + " added=payload['added'],\n",
330 + " deleted=payload['deleted'],\n",
331 + " target=None\n",
332 + " )\n",
333 + " ]\n",
334 + " message = inference(model=args.diff_model, data=get_features(example))\n",
335 + " return jsonify(idx=payload['idx'], message=message)\n",
336 + "\n",
337 + " @app.route('/tokenizer', methods=['POST'])\n",
338 + " def tokenizer():\n",
339 + " if request.method == 'POST':\n",
340 + " payload = request.get_json()\n",
341 + " tokens = args.tokenizer.tokenize(payload['code'])\n",
342 + " return jsonify(tokens=tokens)\n",
343 + "\n",
344 + " return app\n",
345 + "\n",
346 + "def inference(model, data):\n",
347 + " # Calculate bleu\n",
348 + " eval_sampler = SequentialSampler(data)\n",
349 + " eval_dataloader = DataLoader(data, sampler=eval_sampler, batch_size=len(data))\n",
350 + "\n",
351 + " model.eval()\n",
352 + " p=[]\n",
353 + " for batch in tqdm(eval_dataloader, total=len(eval_dataloader)):\n",
354 + " batch = tuple(t.to(args.device) for t in batch)\n",
355 + " source_ids, source_mask, patch_ids = batch\n",
356 + " with torch.no_grad():\n",
357 + " preds = model(source_ids=source_ids, source_mask=source_mask, patch_ids=patch_ids)\n",
358 + " for pred in preds:\n",
359 + " t = pred[0].cpu().numpy()\n",
360 + " t = list(t)\n",
361 + " if 0 in t:\n",
362 + " t = t[:t.index(0)]\n",
363 + " text = args.tokenizer.decode(t, clean_up_tokenization_spaces=False)\n",
364 + " p.append(text)\n",
365 + " return p"
366 + ],
367 + "execution_count": null,
368 + "outputs": []
369 + },
370 + {
371 + "cell_type": "markdown",
372 + "metadata": {
373 + "id": "Esf4r-Ai8cG3"
374 + },
375 + "source": [
376 + "**Set enviroment**"
377 + ]
378 + },
379 + {
380 + "cell_type": "code",
381 + "metadata": {
382 + "id": "mR7gVmSoSUoy"
383 + },
384 + "source": [
385 + "import easydict \n",
386 + "\n",
387 + "args = easydict.EasyDict({\n",
388 + " 'load_model_path': 'weight/', \n",
389 + " 'model_type': 'roberta',\n",
390 + " 'config_name' : 'microsoft/codebert-base',\n",
391 + " 'tokenizer_name' : 'microsoft/codebert-base',\n",
392 + " 'max_source_length' : 512,\n",
393 + " 'max_target_length' : 128,\n",
394 + " 'beam_size' : 10,\n",
395 + " 'do_lower_case' : False,\n",
396 + " 'device' : torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
397 + "})"
398 + ],
399 + "execution_count": null,
400 + "outputs": []
401 + },
402 + {
403 + "cell_type": "code",
404 + "metadata": {
405 + "id": "e8dk5RwvToOv"
406 + },
407 + "source": [
408 + "# flask_ngrok_example.py\n",
409 + "from flask_ngrok import run_with_ngrok\n",
410 + "\n",
411 + "app = Flask(__name__)\n",
412 + "run_with_ngrok(app) # Start ngrok when app is run\n",
413 + "\n",
414 + "config_class, model_class, tokenizer_class = MODEL_CLASSES[args.model_type]\n",
415 + "config = config_class.from_pretrained(args.config_name)\n",
416 + "args.tokenizer = tokenizer_class.from_pretrained(args.tokenizer_name, do_lower_case=args.do_lower_case)\n",
417 + "\n",
418 + "# budild model\n",
419 + "args.added_model =get_model(model_class=model_class, config=config,\n",
420 + " tokenizer=args.tokenizer, mode='added').to(args.device)\n",
421 + "args.diff_model = get_model(model_class=model_class, config=config,\n",
422 + " tokenizer=args.tokenizer, mode='diff').to(args.device)\n",
423 + "\n",
424 + "app = create_app()\n",
425 + "app.run()"
426 + ],
427 + "execution_count": null,
428 + "outputs": []
429 + },
430 + {
431 + "cell_type": "markdown",
432 + "metadata": {
433 + "id": "DXkBcO_sU_VN"
434 + },
435 + "source": [
436 + "## Set commit configure\n",
437 + "Now, set commit configure on your local computer.\n",
438 + "```shell\n",
439 + "$ commit configure --endpoint http://********.ngrok.io\n",
440 + "```"
441 + ]
442 + }
443 + ]
444 +}
...\ No newline at end of file ...\ No newline at end of file