RobustRanking/generic_training.py at main · dice-group/RobustRanking · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
import json
import pickle
from torch.utils.data import DataLoader
#from transformers import AutoTokenizer
import torch
from tqdm import tqdm, trange
from parameters import RankingParser
from models.E5 import E5Ranker
import random
import os
from pytorch_transformers.tokenization_bert import BertTokenizer
from optimizers import noise
import Trainer
#from blinkRanker.parameters import BlinkParser
#from blinkRanker import trainer as blinkTrainer ,evaluator as blinkEvaluator
import data_processing
import Evaluator
import logging
from data_processing import Aida_joint_el
logging.disable(logging.WARNING)
device = torch.device(
            "cuda" if torch.cuda.is_available() else "cpu")


max_candsize=5
add_gold_mention=True


def load_train_blink_Ranking_Model():
    parser = RankingParser(add_model_args=True)
    parser.add_training_args()
    parser.add_eval_args()

    # args = argparse.Namespace(**params)
    args = parser.parse_args()
    print(args)
    params = args.__dict__
    global device
    device=torch.device(
            "cuda:"+str(params["gpu_id"]) if torch.cuda.is_available() else "cpu")
    #for lc-quad
    if params["dataset"]=="lcquad":
        entities,documents,doc_to_ent=data_processing.process_lcquad_file("data/train/lcquad.json")
    #for mintaka
    elif params["dataset"]=="mintaka":
        entities, documents, doc_to_ent = data_processing.process_minitaka_file("data/mintaka/mintaka_train.json")

    #for E5
    if params["found_model"]=="e5":
        train_inst = Trainer.TrainerE5(params=params, evaluate_after_batch=params["eval_interval"], device=device)
    # for BiEncoder
    if params["found_model"] == "biencoder":
        train_inst = Trainer.TrainerRanker(params=params, evaluate_after_batch=params["eval_interval"], device=device)

    #for aida
    if params["dataset"] == "aida":
        dp=Aida_joint_el()
        tk = BertTokenizer.from_pretrained(params["bert_model"], do_lower_case=params["lowercase"])
        entities, documents, doc_to_ent=dp.read_ds_to_list("data/aida/wikidata/aida_train",tk)
    #entities, documents, doc_to_ent = dp.read_ds_to_list("data/aida/wikidata/aida_train",tk)

    '''
    queries= {}
    indexToNode={}
    for node in handbook:
        if "questions"in node["handbookData"]:
            indexToNode.update({str(node["nodeId"]["clusterId"])+"-"+str(node["nodeId"]["entityId"]):node})
            for question in node["handbookData"]["questions"]:
                if not question in queries:
                    queries.update({question:[]})
                queries[question].append(str(node["nodeId"]["clusterId"])+"-"+str(node["nodeId"]["entityId"]))
    '''
    '''
    train_dataloader = DataLoader(list(entities.keys()), shuffle=True, batch_size=1,
                                  )
    '''
    train_dataloader = DataLoader(list(entities.keys()), shuffle=True, batch_size=1,
                                  )
    #train_inst=Trainer.TrainerRanker(params=params, evaluate_after_batch=params["eval_interval"], device=device)

    optimizer, scheduler = train_inst.getOptimizerAndSheduler(len(entities))
    #for lcquad
    #evaluator_inst = Evaluator.IndexEvaluator(params=params, collator=train_inst.collator)
    #for minaka
    #evaluator_inst = Evaluator.IndexEvaluator(params=params, collator=train_inst.collator,use_lcquad=False)

    #for aida
    #eval_entities, eval_documents, eval_doc_to_ent = dp.read_ds_to_list("data/aida/wikidata/aida_testa",tk )
    '''
    evaluator_inst = Evaluator.IndexEvaluator(params=params,collator=train_inst.collator,entities=eval_entities
                                              ,doc_to_ent=eval_doc_to_ent,documents=eval_documents)

    '''
    #filehandler=data_processing.process_lcquad_file
    if params["dataset"] == "aida":
        tk = BertTokenizer.from_pretrained(params["bert_model"], do_lower_case=params["lowercase"])
        evaluator_inst = Evaluator.IndexEvaluator(params=params, collator=train_inst.collator,filehandler=dp.read_ds_to_list,file="data/aida/wikidata/aida_testa",tokenizer=tk)
    if params["dataset"] == "mintaka":
        evaluator_inst = Evaluator.IndexEvaluator(params=params, collator=train_inst.collator,
                                              filehandler=data_processing.process_minitaka_file, file="data/mintaka/mintaka_test.json")
    if params["dataset"] == "lcquad":
        evaluator_inst = Evaluator.IndexEvaluator(params=params, collator=train_inst.collator,
                                                  filehandler=data_processing.process_lcquad_file, file="data/test/lcquad.json")
    evaluator_inst.entities.extend(entities.keys())
    return train_inst,evaluator_inst, train_dataloader, optimizer,scheduler,entities,documents,doc_to_ent


def save_model(model, tokenizer, output_dir):
    """Saves the model and the tokenizer used in the output directory."""
    if not os.path.exists(output_dir):
        os.makedirs(output_dir)

    model_to_save = model.module if hasattr(model, "module") else model
    output_model_file = os.path.join(output_dir, "pytorch_model.bin")
    output_config_file = os.path.join(output_dir, "model_config")
    torch.save(model_to_save.state_dict(), output_model_file)
    #model_to_save.config.to_json_file(output_config_file)
    tokenizer.save_vocabulary(output_dir)

def encode_documents(documents,model,collator):
    documents=list(documents)
    data_loader = DataLoader(documents, shuffle=True, batch_size=100,
                             collate_fn=collator.collate_context)
    #iter_ = tqdm(data_loader, desc="Encode Train Documents")
    doc_encodings = []
    for step, batch in enumerate(data_loader):
        if not isinstance(model,E5Ranker):
            context_input = batch
            # candidate_input = batch["candidate_input"]
            # labels=[0 for i in range(batch["candidate_input"].size(0))]
            # label_input = batch[0]["label_idx"].to(device)
            # label_input = torch.LongTensor(torch.zeros(candidate_input.size(0),dtype=torch.int64)).to(device)
            # context_input, candidate_input, label_input = batch
            encodings = model.encode_context(context_input).tolist()
        else:
            encodings=model.encode_context(batch).tolist()
        doc_encodings.extend(encodings)
    encoding_map={}
    for i in range(len(doc_encodings)):
        encoding_map[documents[i]]=doc_encodings[i]
    return encoding_map

def train(epochs):
    #trainer,evaluator, train_dataloader, optimizer, scheduler = load_train_only_Graph_Model(device)
    trainer, evaluator, train_dataloader, optimizer, scheduler,entities,documents,doc_to_ent = load_train_blink_Ranking_Model()
    trainer.model.train()
    #print(evaluator.evaluate(trainer.model))
    #index,results=evaluator.evaluate(trainer.model)
    index, results, mrr = evaluator.evaluate(trainer.model)
    #index, mrr = evaluator.evaluate_mrr(trainer.model)
    print(results)
    print(mrr)
    encoding_map=encode_documents(documents,trainer.model,trainer.collator)
    for e in range(epochs):
        num_batch = 0

        # step=0
        iter_ = tqdm(train_dataloader, desc="Training")
        for step, batch in enumerate(iter_):
            #batch=data_processing.create_batch_ent(batch[0],list(entities[batch[0]]),random.sample(list(documents),1000),doc_to_ent)
            #batch=data_processing.create_batch_index(batch[0],entities,list(entities[batch[0]]),encoding_map,index,doc_to_ent)
            batch=data_processing.create_batch_label_noise(batch[0],entities,list(entities[batch[0]]),encoding_map,index,doc_to_ent,num_noise_labels=0)
            #batch = data_processing.create_batch_index_document(batch[0], entities,  encoding_map,
            #                                           index, doc_to_ent)
            logits, loss = trainer.make_forward_pass(batch,step)

            if trainer.grad_acc_steps > 1:
                loss = loss / trainer.grad_acc_steps
            loss.backward()
            if (step + 1) % trainer.grad_acc_steps == 0:
                torch.nn.utils.clip_grad_norm_(
                    trainer.model.parameters(), trainer.params["max_grad_norm"]
                )
                noise_function=trainer.params["noise_approach"]
                if noise_function=="anticorrelated_noise_prev_term":
                    noise.add_anticorrelated_noise_prev_term(optimizer, device)
                if noise_function == "gausian_noise":
                    noise.add_gausian_noise(optimizer, device)
                if noise_function == "anticorrelated_noise_gradient":
                    noise.add_anticorrelated_noise_gradient(optimizer, device)
                optimizer.step()
                scheduler.step()
                optimizer.zero_grad()

            num_batch += 1
            # print("Train Epoch " + str(e), "batch " + str(num_batch) + " Loss: " + str(loss.item()))
            if num_batch % trainer.evaluate_after == 0:
                print("Start evaluation in epoch:" + str(e) + " batch: " + str(num_batch))
                trainer.model.eval()
                #print(evaluator.evaluate(trainer.model))
                #print(evaluator.evaluate_mrr(trainer.model))
                #index,results = evaluator.evaluate(trainer.model)
                _, results, mrr = evaluator.evaluate(trainer.model)
                #index, mrr = evaluator.evaluate_mrr(trainer.model)
                print(results)
                print(mrr)
                encoding_map = encode_documents(documents, trainer.model, trainer.collator)
                #epoch_output_folder_path = os.path.join(
                #    "ranker_gr", "epoch_{}_{}".format(e, num_batch))
                #save_model(model, model.tokenizer, epoch_output_folder_path)
                trainer.model.train()

        print("Start evaluation after epoch: " + str(e))
        trainer.model.eval()
        #index,results = evaluator.evaluate(trainer.model)
        #index, mrr = evaluator.evaluate_mrr(trainer.model)
        _, results, mrr = evaluator.evaluate(trainer.model)
        print("---------------------------Results in Epoch------------------------:" + str(e))
        print(results)
        #Recall writing in a file
        f = open(trainer.params["training_result_update_file"], 'a+')
        f.write("Results in Epoch: " + str(e)+ str(results)+ '\n')
        f.close()
        #writing mrrs
        f1 = open('Results_Mrr_gaussian.txt', 'a+')
        f1.write("Results in Epoch: " + str(e) + str(mrr) + '\n')
        f1.close()
        encoding_map = encode_documents(documents, trainer.model, trainer.collator)
        epoch_output_folder_path = os.path.join(
            trainer.params["model_dump_folder"], "epoch_{}".format(e)
        )

        #save_model(trainer.model,trainer.tokenizer,  epoch_output_folder_path)
        trainer.model.train()
train(1)