sort_key=lambda x: len(x.text),
sort_within_batch=True,
device=DEVICE)
model = BertForSequenceClassification(
BertConfig(vocab_size=MAX_VOCAB_SIZE,
max_position_embeddings=512,
intermediate_size=1024,
hidden_size=512,
num_attention_heads=8,
num_hidden_layers=6,
type_vocab_size=5,
hidden_dropout_prob=0.1,
attention_probs_dropout_prob=0.1,
num_labels=2))
model.to(DEVICE)
PAD_IDX = TEXT.vocab.stoi[TEXT.pad_token]
UNK_IDX = TEXT.vocab.stoi[TEXT.unk_token]
model.bert.embeddings.word_embeddings.weight.data[PAD_IDX] = torch.zeros(512)
model.bert.embeddings.word_embeddings.weight.data[UNK_IDX] = torch.zeros(512)
print(
f'Parameter: {sum(p.numel() for p in model.parameters() if p.requires_grad):,}'
)
optimizer = RMSprop(model.parameters(), lr=0.000005)
itr = 1
epochs = 20
def main():
parser = argparse.ArgumentParser(description="Training")
## Required parameters
parser.add_argument("--data_dir", default=None, type=str, required=True,
help="The input data dir. Should contain the .tsv files (or other data files) for the task.")
parser.add_argument("--output_dir", type=str, required=True,
help="The output directory (log, checkpoints, parameters, etc.)")
# parser.add_argument("--data_file", type=str, required=True,
# help="The binarized file (tokenized + tokens_to_ids) and grouped by sequence.")
# parser.add_argument("--token_counts", type=str, required=True,
# help="The token counts in the data_file for MLM.")
parser.add_argument("--force", action='store_true',
help="Overwrite output_dir if it already exists.")
parser.add_argument("--task_name", default=None, type=str, required=True,
help="The name of the task to train selected in the list: " + ", ".join(processors.keys()))
parser.add_argument("--do_lower_case", action='store_true',
help="Set this flag if you are using an uncased model.")
# parser.add_argument("--output_dir", default=None, type=str, required=True,
# help="The output directory where the model predictions and checkpoints will be written.")
parser.add_argument("--vocab_size", default=30522, type=int,
help="The vocabulary size.")
parser.add_argument("--max_position_embeddings", default=512, type=int,
help="Maximum sequence length we can model (including [CLS] and [SEP]).")
parser.add_argument("--sinusoidal_pos_embds", action='store_false',
help="If true, the position embeddings are simply fixed with sinusoidal embeddings.")
parser.add_argument("--n_layers", default=6, type=int,
help="Number of Transformer blocks.")
parser.add_argument("--n_heads", default=12, type=int,
help="Number of heads in the self-attention module.")
parser.add_argument("--dim", default=768, type=int,
help="Dimension through the network. Must be divisible by n_heads")
parser.add_argument("--hidden_dim", default=3072, type=int,
help="Intermediate dimension in the FFN.")
parser.add_argument("--dropout", default=0.1, type=float,
help="Dropout.")
parser.add_argument("--attention_dropout", default=0.1, type=float,
help="Dropout in self-attention.")
parser.add_argument("--activation", default='gelu', type=str,
help="Activation to use in self-attention")
parser.add_argument("--tie_weights_", action='store_false',
help="If true, we tie the embeddings matrix with the projection over the vocabulary matrix. Default is true.")
# parser.add_argument("--from_pretrained_weights", default=None, type=str,
# help="Load student initialization checkpoint.")
# parser.add_argument("--from_pretrained_config", default=None, type=str,
# help="Load student initialization architecture config.")
parser.add_argument("--teacher_name", default="bert-base-uncased", type=str,
help="The teacher model.")
parser.add_argument("--temperature", default=2., type=float,
help="Temperature for the softmax temperature.")
parser.add_argument("--alpha_ce", default=1.0, type=float,
help="Linear weight for the distillation loss. Must be >=0.")
# parser.add_argument("--alpha_mlm", default=0.5, type=float,
# help="Linear weight for the MLM loss. Must be >=0.")
parser.add_argument("--alpha_mse", default=0.0, type=float,
help="Linear weight of the MSE loss. Must be >=0.")
parser.add_argument("--alpha_cos", default=0.0, type=float,
help="Linear weight of the cosine embedding loss. Must be >=0.")
# parser.add_argument("--mlm_mask_prop", default=0.15, type=float,
# help="Proportion of tokens for which we need to make a prediction.")
# parser.add_argument("--word_mask", default=0.8, type=float,
# help="Proportion of tokens to mask out.")
# parser.add_argument("--word_keep", default=0.1, type=float,
# help="Proportion of tokens to keep.")
# parser.add_argument("--word_rand", default=0.1, type=float,
# help="Proportion of tokens to randomly replace.")
# parser.add_argument("--mlm_smoothing", default=0.7, type=float,
# help="Smoothing parameter to emphasize more rare tokens (see XLM, similar to word2vec).")
# parser.add_argument("--restrict_ce_to_mask", action='store_true',
# help="If true, compute the distilation loss only the [MLM] prediction distribution.")
parser.add_argument("--n_epoch", type=int, default=3,
help="Number of pass on the whole dataset.")
parser.add_argument("--batch_size", type=int, default=5,
help="Batch size (for each process).")
parser.add_argument("--per_gpu_eval_batch_size", default=8, type=int,
help="Batch size per GPU/CPU for evaluation.")
# parser.add_argument("--tokens_per_batch", type=int, default=-1,
# help="If specified, modify the batches so that they have approximately this number of tokens.")
# parser.add_argument("--shuffle", action='store_false',
# help="If true, shuffle the sequence order. Default is true.")
parser.add_argument("--group_by_size", action='store_false',
help="If true, group sequences that have similar length into the same batch. Default is true.")
parser.add_argument("--gradient_accumulation_steps", type=int, default=50,
help="Gradient accumulation for larger training batches.")
parser.add_argument("--max_steps", default=-1, type=int,
help="If > 0: set total number of training steps to perform. Override num_train_epochs.")
parser.add_argument("--warmup_prop", default=0.05, type=float,
help="Linear warmup proportion.")
parser.add_argument("--weight_decay", default=0.0, type=float,
help="Weight deay if we apply some.")
parser.add_argument("--learning_rate", default=5e-4, type=float,
help="The initial learning rate for Adam.")
parser.add_argument("--adam_epsilon", default=1e-6, type=float,
help="Epsilon for Adam optimizer.")
parser.add_argument("--max_grad_norm", default=5.0, type=float,
help="Max gradient norm.")
parser.add_argument("--initializer_range", default=0.02, type=float,
help="Random initialization range.")
# parser.add_argument('--fp16', action='store_true',
# help="Whether to use 16-bit (mixed) precision (through NVIDIA apex) instead of 32-bit")
# parser.add_argument('--fp16_opt_level', type=str, default='O1',
# help="For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
# "See details at https://nvidia.github.io/apex/amp.html")
parser.add_argument("--n_gpu", type=int, default=1,
help="Number of GPUs in the node.")
parser.add_argument("--local_rank", type=int, default=-1,
help="Distributed training - Local rank")
parser.add_argument("--seed", type=int, default=56,
help="Random seed")
parser.add_argument("--log_interval", type=int, default=10,
help="Tensorboard logging interval.")
parser.add_argument('--log_examples', action='store_false',
help="Show input examples on the command line during evaluation. Enabled by default.")
parser.add_argument("--evaluate_during_training", action='store_true',
help="Rul evaluation during training at each logging step.")
parser.add_argument("--checkpoint_interval", type=int, default=100,
help="Checkpoint interval.")
parser.add_argument("--no_cuda", type=bool, default=False,
help="Avoid using CUDA when available")
parser.add_argument("--toy_mode", action='store_true', help="Toy mode for development.")
parser.add_argument("--rich_eval", action='store_true', help="Rich evaluation (more metrics + mistake reporting).")
args = parser.parse_args()
args.no_cuda = False
print("NO CUDA", args.no_cuda)
## ARGS ##
init_gpu_params(args)
set_seed(args)
if args.is_master:
if os.path.exists(args.output_dir):
if not args.force:
raise ValueError(f'Serialization dir {args.output_dir} already exists, but you have not precised wheter to overwrite it'
'Use `--force` if you want to overwrite it')
else:
shutil.rmtree(args.output_dir)
if not os.path.exists(args.output_dir):
os.makedirs(args.output_dir)
logger.info(f'Experiment will be dumped and logged in {args.output_dir}')
### SAVE PARAMS ###
logger.info(f'Param: {args}')
with open(os.path.join(args.output_dir, 'parameters.json'), 'w') as f:
json.dump(vars(args), f, indent=4)
# git_log(args.output_dir)
print(args.local_rank)
print(args.no_cuda)
# Setup CUDA, GPU & distributed training
if args.local_rank == -1 or args.no_cuda:
device = torch.device("cuda" if torch.cuda.is_available() and not args.no_cuda else "cpu")
args.n_gpu = torch.cuda.device_count()
args.local_rank = -1
else: # Initializes the distributed backend which will take care of sychronizing nodes/GPUs
device = torch.device("cuda", args.local_rank)
args.device = device
logger.info("DEVICE: {}".format(args.device))
logger.info("N_GPU: {}".format(args.n_gpu))
# exit(0)
print(args.local_rank)
### TOKENIZER ###
# if args.teacher_type == 'bert':
tokenizer = BertTokenizer.from_pretrained(args.teacher_name, do_lower_case=args.do_lower_case)
# elif args.teacher_type == 'roberta':
# tokenizer = RobertaTokenizer.from_pretrained(args.teacher_name)
special_tok_ids = {}
for tok_name, tok_symbol in tokenizer.special_tokens_map.items():
idx = tokenizer.all_special_tokens.index(tok_symbol)
special_tok_ids[tok_name] = tokenizer.all_special_ids[idx]
logger.info(f'Special tokens {special_tok_ids}')
args.special_tok_ids = special_tok_ids
# Prepare GLUE task
args.task_name = args.task_name.lower()
if args.task_name not in processors:
raise ValueError("Task not found: %s" % (args.task_name))
args.model_name_or_path = args.teacher_name
args.max_seq_length = args.max_position_embeddings
args.model_type = "bert"
args.output_mode = output_modes[args.task_name]
train_dataset = load_and_cache_examples(args, args.task_name, tokenizer, evaluate=False)
## DATA LOADER ##
# logger.info(f'Loading data from {args.data_file}')
# with open(args.data_file, 'rb') as fp:
# data = pickle.load(fp)
# assert os.path.isfile(args.token_counts)
# logger.info(f'Loading token counts from {args.token_counts} (already pre-computed)')
# with open(args.token_counts, 'rb') as fp:
# counts = pickle.load(fp)
# assert len(counts) == args.vocab_size
# token_probs = np.maximum(counts, 1) ** -args.mlm_smoothing
# for idx in special_tok_ids.values():
# token_probs[idx] = 0. # do not predict special tokens
# token_probs = torch.from_numpy(token_probs)
# train_dataloader = Dataset(params=args, data=data)
train_sampler = RandomSampler(train_dataset)
train_dataloader = DataLoader(train_dataset, sampler=train_sampler, batch_size=args.batch_size)
logger.info(f'Data loader created.')
## STUDENT ##
# if args.from_pretrained_weights is not None:
# assert os.path.isfile(args.from_pretrained_weights)
# assert os.path.isfile(args.from_pretrained_config)
# logger.info(f'Loading pretrained weights from {args.from_pretrained_weights}')
# logger.info(f'Loading pretrained config from {args.from_pretrained_config}')
# stu_architecture_config = DistilBertConfig.from_json_file(args.from_pretrained_config)
# stu_architecture_config.output_hidden_states = True
# student = DistilBertForMaskedLM.from_pretrained(args.from_pretrained_weights,
# config=stu_architecture_config)
# else:
# args.vocab_size_or_config_json_file = args.vocab_size
# stu_architecture_config = DistilBertConfig(**vars(args), output_hidden_states=True)
# student = DistilBertForMaskedLM(stu_architecture_config)
student_config = BertConfig(
vocab_size_or_config_json_file=args.vocab_size,
hidden_size=args.dim,
num_hidden_layers=args.n_layers,
num_attention_heads=args.n_heads,
intermediate_size=args.hidden_dim,
hidden_dropout_prob=args.dropout,
attention_probs_dropout_prob=args.attention_dropout,
max_position_embeddings=args.max_position_embeddings,
hidden_act=args.activation,
initializer_range=0.02)
student = BertForSequenceClassification(student_config)
if args.n_gpu > 0:
student.to(f'cuda:{args.local_rank}')
logger.info(f'Student loaded.')
## TEACHER ##
teacher = BertForSequenceClassification.from_pretrained(args.teacher_name) # take outputs[1] for the logits
if args.n_gpu > 0:
teacher.to(f'cuda:{args.local_rank}')
logger.info(f'Teacher loaded from {args.teacher_name}.')
## DISTILLER ##
torch.cuda.empty_cache()
distiller = Distiller(params=args,
dataloader=train_dataloader,
# token_probs=token_probs,
student=student,
teacher=teacher,
tokenizer=tokenizer)
distiller.train()
logger.info("Let's go get some drinks.")
model = BertForSequenceClassification(
BertConfig(
vocab_size=MAX_VOCAB_SIZE,
max_position_embeddings=512,
intermediate_size=1024,
hidden_size=512,
num_attention_heads=8,
num_hidden_layers=6,
type_vocab_size=5,
hidden_dropout_prob=0.1,
attention_probs_dropout_prob=0.1,
num_labels=2
)
)
model.to(device)
print(f'모델의 파라미터 수는 {sum(p.numel() for p in model.parameters() if p.requires_grad):,} 개 입니다.')
optimizer = RMSprop(model.parameters(), lr=1e-6)
itr = 1
epochs = 30
total_loss = 0
total_len = 0
total_correct = 0
model.train()
for epoch in range(epochs):
for batch in train_iterator:
optimizer.zero_grad()