runnable v1

.gitignore

@@ -7,3 +7,4 @@ dataset/FB15k-237/masked_*.txt
 dataset/FB15k-237/cached_*.pkl
 **/__pycache__/
 **/.DS_Store
+nohup.out

.vscode/launch.json

@@ -0,0 +1,38 @@
+{
+    // Use IntelliSense to learn about possible attributes.
+    // Hover to view descriptions of existing attributes.
+    // For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
+    "version": "0.2.0",
+    "configurations": [
+        {
+            "name": "Python: Current File",
+            "type": "python",
+            "request": "launch",
+            "program": "${file}",
+            "console": "integratedTerminal",
+            "justMyCode": true,
+            "args": [
+                "--gpus", "1", 
+                "--max_epochs=16",  
+                "--num_workers=32", 
+                "--model_name_or_path",  "bert-base-uncased",
+                "--accumulate_grad_batches", "1", 
+                "--model_class", "BertKGC",
+                "--batch_size", "64",
+                "--checkpoint", "/kg_374/Relphormer/pretrain/output/FB15k-237/epoch=15-step=19299-Eval/hits10=0.96.ckpt",
+                "--pretrain", "0",
+                "--bce", "0",
+                "--check_val_every_n_epoch", "1",
+                "--overwrite_cache",
+                "--data_dir", "dataset/FB15k-237", 
+                "--eval_batch_size", "128",
+                "--max_seq_length", "128",
+                "--lr", "3e-5",
+                "--max_triplet", "64",
+                "--add_attn_bias", "True",
+                "--use_global_node", "True",
+                "--fast_dev_run", "True",
+            ]
+        }
+    ]
+}

data/data_module.py

@@ -10,7 +10,7 @@ from transformers.tokenization_utils_base import (BatchEncoding,
                                                   PreTrainedTokenizerBase)
 
 from .base_data_module import BaseDataModule
-from .processor import KGProcessor, get_dataset
+from .processor import KGProcessor, get_dataset, getNegativeEntityId
 import transformers
 transformers.logging.set_verbosity_error()
 
@@ -79,6 +79,7 @@ class DataCollatorForSeq2Seq:
     label_pad_token_id: int = -100
     return_tensors: str = "pt"
     num_labels: int = 0
+    args: Any = None
 
     def __call__(self, features, return_tensors=None):
 
@@ -105,8 +106,9 @@ class DataCollatorForSeq2Seq:
                 if isinstance(l, int): 
                     new_labels[i][l] = 1
                 else:
-                    for j in l:
+                    if (l[0] != getNegativeEntityId(self.args)):
-                        new_labels[i][j] = 1
+                        for j in l:
+                            new_labels[i][j] = 1
             labels = new_labels
 
         features = self.tokenizer.pad(
@@ -141,6 +143,7 @@ class KGC(BaseDataModule):
             padding="longest",
             max_length=self.args.max_seq_length,
             num_labels = len(entity_list),
+            args=args
         )
         relations_tokens = self.processor.get_relations(args.data_dir)
         self.num_relations = len(relations_tokens)

data/processor.py

@@ -5,7 +5,7 @@ import contextlib
 import sys
 
 from collections import Counter
-from multiprocessing import Pool
+from multiprocessing import Pool, synchronize
 from torch._C import HOIST_CONV_PACKED_PARAMS
 from torch.utils.data import Dataset, Sampler, IterableDataset
 from collections import defaultdict
@@ -110,7 +110,7 @@ def cache_results(_cache_fp, _refresh=False, _verbose=1):
             
             model_name = my_args.model_name_or_path.split("/")[-1]
             is_pretrain = my_args.pretrain
-            cache_filepath = os.path.join(my_args.data_dir, f"cached_{mode}_features{model_name}_pretrain{is_pretrain}_faiss{my_args.faiss_init}_seqlength{my_args.max_seq_length}_{my_args.litmodel_class}.pkl")
+            cache_filepath = os.path.join(my_args.data_dir, f"cached_{func.__name__}_{mode}_features{model_name}_pretrain{is_pretrain}_faiss{my_args.faiss_init}_seqlength{my_args.max_seq_length}_{my_args.litmodel_class}.pkl")
             refresh = my_args.overwrite_cache
 
             if cache_filepath is not None and refresh is False:
@@ -137,6 +137,116 @@ def cache_results(_cache_fp, _refresh=False, _verbose=1):
 
     return wrapper_
 
+def cache_results_load_once(_cache_fp, _refresh=False, _verbose=1, _global_var=None):
+    r"""
+    === USE IN TRAINING MODE ONLY ===
+    cache_results是fastNLP中用于cache数据的装饰器。通过下面的例子看一下如何使用::
+
+        import time
+        import numpy as np
+        from fastNLP import cache_results
+        
+        @cache_results('cache.pkl')
+        def process_data():
+            # 一些比较耗时的工作，比如读取数据，预处理数据等，这里用time.sleep()代替耗时
+            time.sleep(1)
+            return np.random.randint(10, size=(5,))
+        
+        start_time = time.time()
+        print("res =",process_data())
+        print(time.time() - start_time)
+        
+        start_time = time.time()
+        print("res =",process_data())
+        print(time.time() - start_time)
+        
+        # 输出内容如下，可以看到两次结果相同，且第二次几乎没有花费时间
+        # Save cache to cache.pkl.
+        # res = [5 4 9 1 8]
+        # 1.0042750835418701
+        # Read cache from cache.pkl.
+        # res = [5 4 9 1 8]
+        # 0.0040721893310546875
+
+    可以看到第二次运行的时候，只用了0.0001s左右，是由于第二次运行将直接从cache.pkl这个文件读取数据，而不会经过再次预处理::
+
+        # 还是以上面的例子为例，如果需要重新生成另一个cache，比如另一个数据集的内容，通过如下的方式调用即可
+        process_data(_cache_fp='cache2.pkl')  # 完全不影响之前的‘cache.pkl'
+
+    上面的_cache_fp是cache_results会识别的参数，它将从'cache2.pkl'这里缓存/读取数据，即这里的'cache2.pkl'覆盖默认的
+    'cache.pkl'。如果在你的函数前面加上了@cache_results()则你的函数会增加三个参数[_cache_fp, _refresh, _verbose]。
+    上面的例子即为使用_cache_fp的情况，这三个参数不会传入到你的函数中，当然你写的函数参数名也不可能包含这三个名称::
+
+        process_data(_cache_fp='cache2.pkl', _refresh=True)  # 这里强制重新生成一份对预处理的cache。
+        #  _verbose是用于控制输出信息的，如果为0,则不输出任何内容;如果为1,则会提醒当前步骤是读取的cache还是生成了新的cache
+
+    :param str _cache_fp: 将返回结果缓存到什么位置;或从什么位置读取缓存。如果为None，cache_results没有任何效用，除非在
+        函数调用的时候传入_cache_fp这个参数。
+    :param bool _refresh: 是否重新生成cache。
+    :param int _verbose: 是否打印cache的信息。
+    :return:
+    """
+
+    def wrapper_(func):
+        signature = inspect.signature(func)
+        for key, _ in signature.parameters.items():
+            if key in ('_cache_fp', '_refresh', '_verbose', '_global_var'):
+                raise RuntimeError("The function decorated by cache_results cannot have keyword `{}`.".format(key))
+        v = globals().get(_global_var, None)
+        if (v is not None):
+            return v
+
+        def wrapper(*args, **kwargs):
+            
+            my_args = args[0]
+            mode = "train"
+            if '_cache_fp' in kwargs:
+                cache_filepath = kwargs.pop('_cache_fp')
+                assert isinstance(cache_filepath, str), "_cache_fp can only be str."
+            else:
+                cache_filepath = _cache_fp
+            if '_refresh' in kwargs:
+                refresh = kwargs.pop('_refresh')
+                assert isinstance(refresh, bool), "_refresh can only be bool."
+            else:
+                refresh = _refresh
+            if '_verbose' in kwargs:
+                verbose = kwargs.pop('_verbose')
+                assert isinstance(verbose, int), "_verbose can only be integer."
+            else:
+                verbose = _verbose
+            refresh_flag = True
+            
+            model_name = my_args.model_name_or_path.split("/")[-1]
+            is_pretrain = my_args.pretrain
+            cache_filepath = os.path.join(my_args.data_dir, f"cached_{func.__name__}_{mode}_features{model_name}_pretrain{is_pretrain}.pkl")
+            refresh = my_args.overwrite_cache
+
+            if cache_filepath is not None and refresh is False:
+                # load data
+                if os.path.exists(cache_filepath):
+                    with open(cache_filepath, 'rb') as f:
+                        results = pickle.load(f)
+                        globals()[_global_var] = results
+                    if verbose == 1:
+                        logger.info("Read cache from {}.".format(cache_filepath))
+                    refresh_flag = False
+
+            if refresh_flag:
+                results = func(*args, **kwargs)
+                if cache_filepath is not None:
+                    if results is None:
+                        raise RuntimeError("The return value is None. Delete the decorator.")
+                    with open(cache_filepath, 'wb') as f:
+                        pickle.dump(results, f)
+                    logger.info("Save cache to {}.".format(cache_filepath))
+
+            return results
+
+        return wrapper
+
+    return wrapper_
+
 
 import argparse
 import csv
@@ -235,6 +345,31 @@ class DataProcessor(object):
 
 import copy
 
+from collections import deque
+import threading
+
+
+class _LiveState(type):
+    _instances = {}
+    _lock = threading.Lock()
+
+    def __call__(cls, *args, **kwargs):
+        if cls not in cls._instances:
+            with cls._lock:
+                if cls not in cls._instances:
+                    cls._instances[cls] = super(_LiveState, cls).__call__(*args, **kwargs)
+        
+        return cls._instances[cls]
+
+class LiveState(metaclass=_LiveState):
+
+    def __init__(self):
+        self._pool_size = 4
+        self._deq = deque(maxlen=self._pool_size)
+    def put(self, item):
+        self._deq.append(item)
+    def get(self):
+        return list(self._deq)
 
 def solve_get_knowledge_store(line, set_type="train", pretrain=1):
     """
@@ -364,6 +499,58 @@ def solve(line,  set_type="train", pretrain=1, max_triplet=32):
             InputExample(guid=guid, text_a="[MASK]", text_b="[PAD]", text_c = "[PAD]", text_d = list(masked_head_seq), label=lmap(lambda x: ent2id[x], b), real_label=ent2id[line[0]], en=[line[1], line[2]], en_id = [rel2id[line[1]], ent2id[line[2]]], rel=rel2id[line[1]], text_d_id = list(masked_head_seq_id), graph_inf = masked_head_graph_list))
         examples.append(
             InputExample(guid=guid, text_a="[PAD]", text_b="[PAD]", text_c = "[MASK]", text_d = list(masked_tail_seq), label=lmap(lambda x: ent2id[x], a), real_label=ent2id[line[2]], en=[line[0], line[1]], en_id = [ent2id[line[0]], rel2id[line[1]]], rel=rel2id[line[1]], text_d_id = list(masked_tail_seq_id), graph_inf = masked_tail_graph_list))
+        
+        liveState = LiveState()
+        _prev = liveState.get()
+
+        if (set_type == "train" and len(_prev) > 0):
+
+            for prev_ent in _prev:
+
+                if (prev_ent == line[0] or prev_ent == line[2]):
+                    continue
+
+                z = head_filter_entities["\t".join([prev_ent,line[1]])]
+                if (len(z) == 0):
+                    z.append('[NEG]')
+                    z.append(line[2])
+                    z.append(line[0])
+                masked_neg_seq = set()
+                masked_neg_seq_id = set()
+
+                masked_neg_graph_list = masked_tail_neighbor.get("\t".join([prev_ent, line[1]]), []) if len(masked_tail_neighbor.get("\t".join([prev_ent, line[1]]), [])) < max_triplet else \
+                    random.sample(masked_tail_neighbor["\t".join([prev_ent, line[1]])], max_triplet)
+                if (len(masked_head_graph_list) == 0):
+                    masked_head_graph_list.append(['[NEG]', line[1], '[NEG]'])
+
+                for item in masked_neg_graph_list:
+                    masked_neg_seq.add(item[0])
+                    masked_neg_seq.add(item[1])
+                    masked_neg_seq.add(item[2])
+                    masked_neg_seq_id.add(ent2id[item[0]])
+                    masked_neg_seq_id.add(rel2id[item[1]])
+                    masked_neg_seq_id.add(ent2id[item[2]])
+                    
+                masked_neg_seq = masked_neg_seq.difference({line[0]})
+                masked_neg_seq = masked_neg_seq.difference({line[2]})
+                masked_neg_seq = masked_neg_seq.difference({line[1]})
+                masked_neg_seq = masked_neg_seq.difference({prev_ent})
+                masked_neg_seq_id = masked_neg_seq_id.difference({ent2id[line[0]]})
+                masked_neg_seq_id = masked_neg_seq_id.difference({rel2id[line[1]]})
+                masked_neg_seq_id = masked_neg_seq_id.difference({ent2id[line[2]]})
+                masked_neg_seq_id = masked_neg_seq_id.difference({ent2id[prev_ent]})
+                # examples.append(
+                #     InputExample(guid=guid, text_a="[MASK]", text_b=' '.join(text_b.split(' ')[:16]) + " [PAD]", text_c = "[PAD]" + " " + ' '.join(text_c.split(' ')[:16]), text_d = masked_head_seq, label=lmap(lambda x: ent2id[x], b), real_label=ent2id[line[0]], en=[rel2id[line[1]], ent2id[line[2]]], rel=rel2id[line[1]]))
+                # examples.append(
+                #     InputExample(guid=guid, text_a="[PAD] ", text_b=' '.join(text_b.split(' ')[:16]) + " [PAD]", text_c = "[MASK]" +" " + ' '.join(text_a.split(' ')[:16]), text_d = masked_tail_seq, label=lmap(lambda x: ent2id[x], a), real_label=ent2id[line[2]], en=[ent2id[line[0]], rel2id[line[1]]], rel=rel2id[line[1]]))
+                examples.append(
+                    InputExample(guid=guid, text_a="[MASK]", text_b="[PAD]", text_c = "[PAD]", text_d = list(masked_neg_seq), label=lmap(lambda x: ent2id[x], z), real_label=ent2id[line[0]], en=[line[1], prev_ent], en_id = [rel2id[line[1]], ent2id[prev_ent]], rel=rel2id[line[1]], text_d_id = list(masked_neg_seq_id), graph_inf = masked_neg_graph_list))
+                examples.append(    
+                    InputExample(guid=guid, text_a="[MASK]", text_b="[PAD]", text_c = "[PAD]", text_d = list(masked_neg_seq), label=lmap(lambda x: ent2id[x], z), real_label=ent2id[line[2]], en=[line[1], prev_ent], en_id = [rel2id[line[1]], ent2id[prev_ent]], rel=rel2id[line[1]], text_d_id = list(masked_neg_seq_id), graph_inf = masked_neg_graph_list))
+        
+
+        liveState.put(line[0])
+        liveState.put(line[2])
     return examples
 
 def filter_init(head, tail, t1,t2, ent2id_, ent2token_, rel2id_, masked_head_neighbor_, masked_tail_neighbor_, rel2token_):
@@ -377,6 +564,7 @@ def filter_init(head, tail, t1,t2, ent2id_, ent2token_, rel2id_, masked_head_nei
     global masked_head_neighbor
     global masked_tail_neighbor
     global rel2token
+    # global negativeEntity
 
     head_filter_entities = head
     tail_filter_entities = tail
@@ -388,11 +576,23 @@ def filter_init(head, tail, t1,t2, ent2id_, ent2token_, rel2id_, masked_head_nei
     masked_head_neighbor = masked_head_neighbor_
     masked_tail_neighbor = masked_tail_neighbor_
     rel2token = rel2token_
+    # negativeEntity = ent2id['[NEG]']
+    print("Initialized negative entity ID")
 
 def delete_init(ent2text_):
     global ent2text
     ent2text = ent2text_
 
+def getEntityIdByName(name):
+    global ent2id
+    return ent2id[name]
+
+@cache_results_load_once(_cache_fp="./dataset", _global_var='negativeEntity')
+def getNegativeEntityId(args):
+    global negativeEntity
+    negativeEntity = ent2id['[NEG]']
+    return negativeEntity
+
 
 class KGProcessor(DataProcessor):
     """Processor for knowledge graph data set."""
@@ -443,6 +643,7 @@ class KGProcessor(DataProcessor):
         """Gets all entities in the knowledge graph."""
         with open(self.entity_path, 'r') as f:
             lines = f.readlines()
+            lines.append('[NEG]\t')
             entities = []
             for line in lines:
                 entities.append(line.strip().split("\t")[0])
@@ -469,6 +670,7 @@ class KGProcessor(DataProcessor):
         ent2text_with_type = {}
         with open(self.entity_path, 'r') as f:
             ent_lines = f.readlines()
+            ent_lines.append('[NEG]\t')
             for line in ent_lines:
                 temp = line.strip().split('\t')
                 try:
@@ -570,6 +772,8 @@ class KGProcessor(DataProcessor):
         threads = min(1, cpu_count())
         filter_init(head_filter_entities, tail_filter_entities,ent2text, rel2text, ent2id, ent2token, rel2id, masked_head_neighbor, masked_tail_neighbor, rel2token
             )
+        # cache this
+        getNegativeEntityId(args)
         
         if hasattr(args, "faiss_init") and args.faiss_init:
             annotate_ = partial(
@@ -579,6 +783,7 @@ class KGProcessor(DataProcessor):
         else:
             annotate_ = partial(
                 solve,
+                set_type=set_type,
                 pretrain=self.args.pretrain,
                 max_triplet=self.args.max_triplet
             )

lit_models/transformer.py

@@ -81,10 +81,13 @@ class TransformerLitModel(BaseLitModel):
         pos = batch.pop("pos")
         try:
             en = batch.pop("en")
+            # self.print("__DEBUG__: en", en)
             rel = batch.pop("rel")
+            # self.print("__DEBUG__: rel", rel)
         except KeyError:
             pass
         input_ids = batch['input_ids']
+        # self.print("__DEBUG__: input_ids", input_ids)
 
         distance_attention = torch.stack([pad_distance(len(input_ids[i]) - len(distance) - 1, distance) for i, distance in enumerate(batch['distance_attention'])])
         distance = batch.pop("distance_attention")
@@ -382,13 +385,17 @@ class CombineEntityEmbeddingLitModel(UseEntityEmbeddingLitModel):
         self.id2entity = {}
         with open("./dataset/FB15k-237/entity2textlong.txt", 'r') as file:
             cnt = 0
-            for line in file.readlines():
+            lines = file.readlines()
+            lines.append('[NEG]\t')
+            for line in lines:
                 e, d = line.strip().split("\t")
                 self.id2entity[cnt] = e
                 cnt += 1
         self.id2entity_t = {}
         with open("./dataset/FB15k-237/entity2text.txt", 'r') as file:
-            for line in file.readlines():
+            lines = file.readlines()
+            lines.append('[NEG]\t')
+            for line in lines:
                 e, d = line.strip().split("\t")
                 self.id2entity_t[e] = d
         for k, v in self.id2entity.items():

main.py

@@ -98,6 +98,7 @@ def main():
     tokenizer = data.tokenizer
 
     lit_model = litmodel_class(args=args, model=model, tokenizer=tokenizer, data_config=data.get_config())
+    print("__DEBUG__: Initialized")
     if args.checkpoint:
         lit_model.load_state_dict(torch.load(args.checkpoint, map_location="cpu")["state_dict"], strict=False)
 
@@ -120,9 +121,12 @@ def main():
     callbacks = [early_callback, model_checkpoint]
 
     # args.weights_summary = "full"  # Print full summary of the model
-    trainer = pl.Trainer.from_argparse_args(args, callbacks=callbacks, logger=logger, default_root_dir="training/logs", accelerator="ddp")
+    trainer = pl.Trainer.from_argparse_args(args, callbacks=callbacks, logger=logger, default_root_dir="training/logs")
     
+    print('__DEBUG__: Init trainer')
+
     if "EntityEmbedding" not in lit_model.__class__.__name__:
+        print('__DEBUG__: Fit trainer')
         trainer.fit(lit_model, datamodule=data)
         path = model_checkpoint.best_model_path
         lit_model.load_state_dict(torch.load(path)["state_dict"], strict=False)

pretrain/data/data_module.py

@@ -10,7 +10,7 @@ from transformers.tokenization_utils_base import (BatchEncoding,
                                                   PreTrainedTokenizerBase)
 
 from .base_data_module import BaseDataModule
-from .processor import KGProcessor, get_dataset
+from .processor import KGProcessor, get_dataset, getNegativeEntityId
 import transformers
 transformers.logging.set_verbosity_error()
 
@@ -106,8 +106,9 @@ class DataCollatorForSeq2Seq:
                 if isinstance(l, int): 
                     new_labels[i][l] = 1
                 else:
-                    for j in l:
+                    if (l[0] != getNegativeEntityId()):
-                        new_labels[i][j] = 1
+                        for j in l:
+                            new_labels[i][j] = 1
             labels = new_labels
 
         features = self.tokenizer.pad(

pretrain/data/processor.py

@@ -314,6 +314,7 @@ def filter_init(head, tail, t1,t2, ent2id_, ent2token_, rel2id_):
     global ent2id
     global ent2token
     global rel2id
+    global negativeEntity
 
     head_filter_entities = head
     tail_filter_entities = tail
@@ -322,11 +323,19 @@ def filter_init(head, tail, t1,t2, ent2id_, ent2token_, rel2id_):
     ent2id = ent2id_
     ent2token = ent2token_
     rel2id = rel2id_
+    negativeEntity = ent2id['[NEG]']
 
 def delete_init(ent2text_):
     global ent2text
     ent2text = ent2text_
 
+def getEntityIdByName(name):
+    global ent2id
+    return ent2id[name]
+
+def getNegativeEntityId():
+    global negativeEntity
+    return negativeEntity
 
 class KGProcessor(DataProcessor):
     """Processor for knowledge graph data set."""
@@ -377,6 +386,7 @@ class KGProcessor(DataProcessor):
         """Gets all entities in the knowledge graph."""
         with open(self.entity_path, 'r') as f:
             lines = f.readlines()
+            lines.append('[NEG]\t')
             entities = []
             for line in lines:
                 entities.append(line.strip().split("\t")[0])
@@ -403,6 +413,7 @@ class KGProcessor(DataProcessor):
         ent2text_with_type = {}
         with open(self.entity_path, 'r') as f:
             ent_lines = f.readlines()
+            ent_lines.append('[NEG]\t')
             for line in ent_lines:
                 temp = line.strip().split('\t')
                 try:

pretrain/lit_models/transformer.py

@@ -364,13 +364,17 @@ class CombineEntityEmbeddingLitModel(UseEntityEmbeddingLitModel):
         self.id2entity = {}
         with open("./dataset/FB15k-237/entity2textlong.txt", 'r') as file:
             cnt = 0
-            for line in file.readlines():
+            lines = file.readlines()
+            lines.append('[NEG]\t')
+            for line in lines:
                 e, d = line.strip().split("\t")
                 self.id2entity[cnt] = e
                 cnt += 1
         self.id2entity_t = {}
         with open("./dataset/FB15k-237/entity2text.txt", 'r') as file:
-            for line in file.readlines():
+            lines = file.readlines()
+            lines.append('[NEG]\t')
+            for line in lines:
                 e, d = line.strip().split("\t")
                 self.id2entity_t[e] = d
         for k, v in self.id2entity.items():

scripts/fb15k-237/fb15k-237.sh

@@ -1,13 +1,12 @@
-nohup python -u main.py --gpus "1" --max_epochs=16  --num_workers=32 \
+nohup python -u main.py --gpus "2," --max_epochs=16  --num_workers=32 \
    --model_name_or_path  bert-base-uncased \
    --accumulate_grad_batches 1 \
    --model_class BertKGC \
    --batch_size 64 \
-   --checkpoint /kg_374/Relphormer/pretrain/output/FB15k-237/epoch=15-step=19299-Eval/hits10=0.96.ckpt \
+   --checkpoint /kg_374/Relphormer/output/FB15k-237/epoch=1-Eval/hits10=Eval/hits1=0.47-Eval/hits1=0.22.ckpt \
    --pretrain 0 \
    --bce 0 \
    --check_val_every_n_epoch 1 \
-   --overwrite_cache \
    --data_dir dataset/FB15k-237 \
    --eval_batch_size 128 \
    --max_seq_length 128 \