import sys import json from torch.utils.data import DataLoader from sentence_transformers import SentenceTransformer, LoggingHandler, util, models, evaluation, losses, InputExample import logging from datetime import datetime import gzip import os import tarfile from collections import defaultdict from torch.utils.data import IterableDataset import tqdm from torch.utils.data import Dataset import random from shutil import copyfile import argparse parser = argparse.ArgumentParser() parser.add_argument("--train_batch_size", default=64, type=int) parser.add_argument("--max_seq_length", default=300, type=int) parser.add_argument("--model_name", required=True) parser.add_argument("--max_passages", default=0, type=int) parser.add_argument("--epochs", default=10, type=int) parser.add_argument("--pooling", default="cls") parser.add_argument("--negs_to_use", default=None, help="From which systems should negatives be used? Multiple systems seperated by comma. None = all") parser.add_argument("--warmup_steps", default=1000, type=int) parser.add_argument("--lr", default=2e-5, type=float) parser.add_argument("--name", default='') parser.add_argument("--num_negs_per_system", default=5, type=int) parser.add_argument("--use_pre_trained_model", default=False, action="store_true") parser.add_argument("--use_all_queries", default=False, action="store_true") args = parser.parse_args() print(args) #### Just some code to print debug information to stdout logging.basicConfig(format='%(asctime)s - %(message)s', datefmt='%Y-%m-%d %H:%M:%S', level=logging.INFO, handlers=[LoggingHandler()]) #### /print debug information to stdout # The model we want to fine-tune train_batch_size = args.train_batch_size #Increasing the train batch size improves the model performance, but requires more GPU memory model_name = args.model_name max_passages = args.max_passages max_seq_length = args.max_seq_length #Max length for passages. Increasing it, requires more GPU memory num_negs_per_system = args.num_negs_per_system # We used different systems to mine hard negatives. Number of hard negatives to add from each system num_epochs = args.epochs # Number of epochs we want to train # We construct the SentenceTransformer bi-encoder from scratch if args.use_pre_trained_model: print("use pretrained SBERT model") model = SentenceTransformer(model_name) model.max_seq_length = max_seq_length else: print("Create new SBERT model") word_embedding_model = models.Transformer(model_name, max_seq_length=max_seq_length) pooling_model = models.Pooling(word_embedding_model.get_word_embedding_dimension(), args.pooling) model = SentenceTransformer(modules=[word_embedding_model, pooling_model]) model_save_path = f'output/train_bi-encoder-margin_mse_en-{args.name}-{model_name.replace("/", "-")}-batch_size_{train_batch_size}-{datetime.now().strftime("%Y-%m-%d_%H-%M-%S")}' # Write self to path os.makedirs(model_save_path, exist_ok=True) train_script_path = os.path.join(model_save_path, 'train_script.py') copyfile(__file__, train_script_path) with open(train_script_path, 'a') as fOut: fOut.write("\n\n# Script was called via:\n#python " + " ".join(sys.argv)) ### Now we read the MS Marco dataset data_folder = 'msmarco-data' #### Read the corpus files, that contain all the passages. Store them in the corpus dict corpus = {} #dict in the format: passage_id -> passage. Stores all existent passages collection_filepath = os.path.join(data_folder, 'collection.tsv') if not os.path.exists(collection_filepath): tar_filepath = os.path.join(data_folder, 'collection.tar.gz') if not os.path.exists(tar_filepath): logging.info("Download collection.tar.gz") util.http_get('https://msmarco.blob.core.windows.net/msmarcoranking/collection.tar.gz', tar_filepath) with tarfile.open(tar_filepath, "r:gz") as tar: tar.extractall(path=data_folder) logging.info("Read corpus: collection.tsv") with open(collection_filepath, 'r', encoding='utf8') as fIn: for line in fIn: pid, passage = line.strip().split("\t") corpus[pid] = passage ### Read the train queries, store in queries dict queries = {} #dict in the format: query_id -> query. Stores all training queries queries_filepath = os.path.join(data_folder, 'queries.train.tsv') if not os.path.exists(queries_filepath): tar_filepath = os.path.join(data_folder, 'queries.tar.gz') if not os.path.exists(tar_filepath): logging.info("Download queries.tar.gz") util.http_get('https://msmarco.blob.core.windows.net/msmarcoranking/queries.tar.gz', tar_filepath) with tarfile.open(tar_filepath, "r:gz") as tar: tar.extractall(path=data_folder) with open(queries_filepath, 'r', encoding='utf8') as fIn: for line in fIn: qid, query = line.strip().split("\t") queries[qid] = query # Read our training file: msmarco-hard-negatives.jsonl.gz contains all queries and hard-negatives that were mined with different systems # For each positive and mined-hard negative passage, we have a Cross-Encoder score from the cross-encoder/ms-marco-MiniLM-L-6-v2 model # This Cross-Encoder score allows to de-noise our hard-negatives by requiring that their CE-score is below a certain treshold train_filepath = '/home/msmarco/data/hard-negatives/msmarco-hard-negatives-v6.jsonl.gz' #### Create our training data logging.info("Read train dataset") train_queries = {} ce_scores = {} negs_to_use = None with gzip.open(train_filepath, 'rt') as fIn: for line in tqdm.tqdm(fIn): if max_passages > 0 and len(train_queries) >= max_passages: break data = json.loads(line) if data['qid'] not in ce_scores: ce_scores[data['qid']] = {} # Add pos ce_scores for item in data['pos'] : ce_scores[data['qid']][item['pid']] = item['ce-score'] #Get the positive passage ids pos_pids = [item['pid'] for item in data['pos']] #Get the hard negatives neg_pids = set() if negs_to_use is None: if args.negs_to_use is not None: #Use specific system for negatives negs_to_use = args.negs_to_use.split(",") else: #Use all systems negs_to_use = list(data['neg'].keys()) print("Using negatives from the following systems:", negs_to_use) for system_name in negs_to_use: if system_name not in data['neg']: continue system_negs = data['neg'][system_name] negs_added = 0 for item in system_negs: #Add neg ce_scores ce_scores[data['qid']][item['pid']] = item['ce-score'] pid = item['pid'] if pid not in neg_pids: neg_pids.add(pid) negs_added += 1 if negs_added >= num_negs_per_system: break if args.use_all_queries or (len(pos_pids) > 0 and len(neg_pids) > 0): train_queries[data['qid']] = {'qid': data['qid'], 'query': queries[data['qid']], 'pos': pos_pids, 'neg': neg_pids} logging.info("Train queries: {}".format(len(train_queries))) # We create a custom MSMARCO dataset that returns triplets (query, positive, negative) # on-the-fly based on the information from the mined-hard-negatives jsonl file. class MSMARCODataset(Dataset): def __init__(self, queries, corpus, ce_scores): self.queries = queries self.queries_ids = list(queries.keys()) self.corpus = corpus self.ce_scores = ce_scores for qid in self.queries: self.queries[qid]['pos'] = list(self.queries[qid]['pos']) self.queries[qid]['neg'] = list(self.queries[qid]['neg']) random.shuffle(self.queries[qid]['neg']) def __getitem__(self, item): query = self.queries[self.queries_ids[item]] query_text = query['query'] qid = query['qid'] if len(query['pos']) > 0: pos_id = query['pos'].pop(0) #Pop positive and add at end pos_text = self.corpus[pos_id] query['pos'].append(pos_id) else: #We only have negatives, use two negs pos_id = query['neg'].pop(0) #Pop negative and add at end pos_text = self.corpus[pos_id] query['neg'].append(pos_id) #Get a negative passage neg_id = query['neg'].pop(0) #Pop negative and add at end neg_text = self.corpus[neg_id] query['neg'].append(neg_id) pos_score = self.ce_scores[qid][pos_id] neg_score = self.ce_scores[qid][neg_id] return InputExample(texts=[query_text, pos_text, neg_text], label=pos_score-neg_score) def __len__(self): return len(self.queries) # For training the SentenceTransformer model, we need a dataset, a dataloader, and a loss used for training. train_dataset = MSMARCODataset(queries=train_queries, corpus=corpus, ce_scores=ce_scores) train_dataloader = DataLoader(train_dataset, shuffle=True, batch_size=train_batch_size, drop_last=True) train_loss = losses.MarginMSELoss(model=model) # Train the model model.fit(train_objectives=[(train_dataloader, train_loss)], epochs=num_epochs, warmup_steps=args.warmup_steps, use_amp=True, checkpoint_path=model_save_path, checkpoint_save_steps=10000, checkpoint_save_total_limit = 0, optimizer_params = {'lr': args.lr}, ) # Train latest model model.save(model_save_path) # Script was called via: #python train_bi-encoder-margin_mse-en.py --model final-models/distilbert-margin_mse-sym_mnrl-mean-v1 --lr=1e-5 --warmup_steps=10000 --negs_to_use=distilbert-margin_mse-sym_mnrl-mean-v1 --num_negs_per_system=10 --epochs=30 --name=cnt_with_mined_negs_mean --use_pre_trained_model --train_batch_size 64