149 lines
6.4 KiB
Text
149 lines
6.4 KiB
Text
{
|
|
"cells": [
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 3,
|
|
"metadata": {},
|
|
"outputs": [
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\u001b[38;5;226m2022-07-01 14:28:43,377 | WARNING | Environment variable ENVIRONMENT is not set, defaulting to development mode ‼️\u001b[0m\n",
|
|
"\u001b[38;5;39m2022-07-01 14:28:43,378 | INFO | Found credentials file (/data/projects/scoutinscience/platform/projects/highlights-service2/mongo.ini), initialising MongoDbDriver\u001b[0m\n",
|
|
"\u001b[38;5;39m2022-07-01 14:28:43,379 | INFO | Found credentials file (/data/projects/scoutinscience/platform/projects/highlights-service2/s3.ini), initialising LargeFileS3\u001b[0m\n",
|
|
"\u001b[38;5;39m2022-07-01 14:28:43,380 | INFO | Settings: configured ✅\u001b[0m\n",
|
|
"\u001b[38;5;39m2022-07-01 14:28:43,380 | INFO | 🔩 tracing_database: MongoDbDriver\u001b[0m\n",
|
|
"\u001b[38;5;39m2022-07-01 14:28:43,381 | INFO | 🔩 large_file_implementation: LargeFileS3\u001b[0m\n",
|
|
"\u001b[38;5;39m2022-07-01 14:28:43,381 | INFO | 🔩 is_production: False\u001b[0m\n",
|
|
"\u001b[38;5;39m2022-07-01 14:28:43,382 | INFO | 🔩 should_log_exception_stack: True\u001b[0m\n",
|
|
"\u001b[38;5;39m2022-07-01 14:28:43,382 | INFO | 🔩 prediction_cache_size: 512\u001b[0m\n",
|
|
"\u001b[38;5;39m2022-07-01 14:28:43,383 | INFO | 🔩 dashboard_table_size: 50\u001b[0m\n",
|
|
"\u001b[38;5;226m2022-07-01 14:28:43,384 | WARNING | You still need to check whether you follow all best practices before trusting your deployment.\u001b[0m\n",
|
|
"\u001b[38;5;226m2022-07-01 14:28:43,385 | WARNING | > Find out more at https://se-ml.github.io/practices/\u001b[0m\n",
|
|
"\u001b[38;5;39m2022-07-01 14:28:44,082 | INFO | Latest version of scibert-highlights is 0 (from versions: 0)\u001b[0m\n",
|
|
"\u001b[38;5;39m2022-07-01 14:28:44,083 | INFO | File scibert-highlights-0 found in cache\u001b[0m\n",
|
|
"\u001b[38;5;39m2022-07-01 14:28:44,084 | INFO | File scibert-highlights-0 found in cache\u001b[0m\n"
|
|
]
|
|
}
|
|
],
|
|
"source": [
|
|
"from great_ai import GreatAI, use_model, MongoDbDriver, configure\n",
|
|
"from great_ai.utilities import clean\n",
|
|
"from pathlib import Path\n",
|
|
"from transformers import (\n",
|
|
" AutoConfig,\n",
|
|
" AutoModelForSequenceClassification,\n",
|
|
" AutoTokenizer,\n",
|
|
")\n",
|
|
"import re\n",
|
|
"import numpy as np\n",
|
|
"import torch\n",
|
|
"from transformers.modeling_outputs import SequenceClassifierOutput\n",
|
|
"from transformers import (\n",
|
|
" PreTrainedModel,\n",
|
|
" PreTrainedTokenizer,\n",
|
|
")\n",
|
|
"from views import EvaluatedSentence, Match\n",
|
|
"from great_ai.large_file import LargeFileS3\n",
|
|
"\n",
|
|
"LargeFileS3.configure_credentials_from_file(\"config.ini\")\n",
|
|
"MongoDbDriver.configure_credentials_from_file(\"config.ini\")\n",
|
|
"configure(dashboard_table_size=100)\n",
|
|
"\n",
|
|
"ORIGINAL_MODEL = \"allenai/scibert_scivocab_uncased\"\n",
|
|
"\n",
|
|
"loaded_model: PreTrainedModel = None\n",
|
|
"tokenizer: PreTrainedTokenizer = None\n",
|
|
"\n",
|
|
"\n",
|
|
"@GreatAI.create\n",
|
|
"@use_model(\"scibert-highlights\", version=\"latest\")\n",
|
|
"def find_highlights(sentence: str, model: Path) -> EvaluatedSentence:\n",
|
|
" global loaded_model, tokenizer\n",
|
|
"\n",
|
|
" if loaded_model is None:\n",
|
|
" config = AutoConfig.from_pretrained(\n",
|
|
" model, output_hidden_states=True, output_attentions=True\n",
|
|
" )\n",
|
|
" loaded_model = AutoModelForSequenceClassification.from_pretrained(\n",
|
|
" model, config=config\n",
|
|
" )\n",
|
|
" if tokenizer is None:\n",
|
|
" tokenizer = AutoTokenizer.from_pretrained(ORIGINAL_MODEL)\n",
|
|
"\n",
|
|
" sentence = clean(sentence, convert_to_ascii=True, remove_brackets=True)\n",
|
|
"\n",
|
|
" return evaluate_sentence(sentence=sentence)\n",
|
|
"\n",
|
|
"\n",
|
|
"def evaluate_sentence(sentence: str) -> EvaluatedSentence:\n",
|
|
" tensors = tokenizer(sentence, return_tensors=\"pt\", truncation=True, max_length=512)\n",
|
|
"\n",
|
|
" with torch.no_grad():\n",
|
|
" result: SequenceClassifierOutput = loaded_model(**tensors)\n",
|
|
" positive_likelihood = torch.nn.Softmax(dim=1)(result.logits)[0][1]\n",
|
|
" tokens = tensors[\"input_ids\"][0]\n",
|
|
"\n",
|
|
" attentions = np.sum(result.attentions[-1].numpy()[0], axis=0)[0][\n",
|
|
" 1:-1\n",
|
|
" ] # Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length, sequence_length)`.\n",
|
|
" matches = []\n",
|
|
"\n",
|
|
" token_attentions = list(zip(attentions, tokens[1:-1]))\n",
|
|
" for token in re.split(r\"([ .,])\", sentence):\n",
|
|
" token = token.strip()\n",
|
|
" if not token:\n",
|
|
" continue\n",
|
|
" bert_tokens = tokenizer(\n",
|
|
" token, return_tensors=\"pt\", truncation=True, max_length=512\n",
|
|
" )[\"input_ids\"][0][\n",
|
|
" 1:-1\n",
|
|
" ] # truncation=True needed to fix RuntimeError: Already borrowed\n",
|
|
" score = 0\n",
|
|
" for t1 in bert_tokens:\n",
|
|
" if not token_attentions:\n",
|
|
" break\n",
|
|
" a, t2 = token_attentions.pop(0)\n",
|
|
" assert t1 == t2, sentence\n",
|
|
" score += a\n",
|
|
" matches.append(\n",
|
|
" Match(phrase=token if token in \".,\" else \" \" + token, score=round(score, 4))\n",
|
|
" )\n",
|
|
" if not token_attentions:\n",
|
|
" break\n",
|
|
"\n",
|
|
" return EvaluatedSentence(\n",
|
|
" score=positive_likelihood, text=sentence, explanation=matches\n",
|
|
" )"
|
|
]
|
|
}
|
|
],
|
|
"metadata": {
|
|
"kernelspec": {
|
|
"display_name": "Python 3.10.4 ('.env': venv)",
|
|
"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": "3.10.4"
|
|
},
|
|
"orig_nbformat": 4,
|
|
"vscode": {
|
|
"interpreter": {
|
|
"hash": "02dd6d3afbfa9fbbe1037d64ad9014965528a1ccad21929d6e72f466389a68ad"
|
|
}
|
|
}
|
|
},
|
|
"nbformat": 4,
|
|
"nbformat_minor": 2
|
|
}
|