pavan
/
master-thesis
1
1
Fork 0
Code Issues Pull Requests Packages Projects Releases Activity

Added baseline training, decoding & evaluation scripts

Browse Source
main
Pavan Mandava 3 years ago
parent ce79e6254d
commit 37524e4294
Signed by: pavan
GPG Key ID: CCBE5F9F1AF07B4D
9 changed files with 1153 additions and 0 deletions
Show Stats Download Patch File Download Diff File

0
baseline/__init__.py
Unescape View File

36
baseline/decode_baseline.sh
Unescape View File

@ -0,0 +1,36 @@
#!/bin/bash
# temp 0.7 - 1.5
# top_p 0.2 - 0.8
# Check whether the required environment vars are set
if [[ -z "${SAVED_MODELS_BASELINE}" ]]; then
echo "SAVED_MODELS_BASELINE Environment variable not set. Run set_env_var.sh bash script"
return
fi
# Check whether the MODEL_CHECKPOINT env var set
if [[ -z "${MODEL_CHECKPOINT}" ]]; then
echo "MODEL_CHECKPOINT Environment variable not set. Run \"export MODEL_CHECKPOINT=<path>\""
return
fi
if [ ! -d "${SAVED_MODELS_BASELINE}/${MODEL_CHECKPOINT}" ]; then
echo "Directory ${MODEL_CHECKPOINT} doesn't exist! Provide a valid Saved Model Checkpoint dir."
return
fi
NS=5
TEMP=1
TOP_P=0.5
OUTPUT_DIR=${OUTPUTS_DIR_BASELINE}/${MODEL_CHECKPOINT}
mkdir -p "${OUTPUT_DIR}"
python soloist_decode.py \
--model_type=gpt2 \
--model_name_or_path="${SAVED_MODELS_BASELINE}"/"${MODEL_CHECKPOINT}" \
--num_samples $NS \
--input_file=../data/baseline/test/test.soloist.json \
--top_p $TOP_P \
--temperature $TEMP \
--output_file="${OUTPUT_DIR}"/output_test.json \
--max_turn 15

122
baseline/evaluate.py
Unescape View File

@ -0,0 +1,122 @@
import json
BELIEF_PREFIX = 'belief :'
INVALID_SLOT_VALUES = ["", "dontcare", "not mentioned", "don't care", "dont care", "do n't care", "none"]
def create_slot_value_map_from_belief_prediction(belief_prediction):
# remove 'belief:' from the beginning
if belief_prediction.startswith(BELIEF_PREFIX):
belief_prediction = belief_prediction[len(BELIEF_PREFIX):]
belief_state_splits = belief_prediction.split('|')
belief_slot_value_map = {}
for belief_state in belief_state_splits:
if belief_state == '':
continue
if len(belief_state.split()) == 0:
continue
domain = belief_state.split()[0]
if domain == 'none':
continue
# remove domain from belief state
belief_state = ' '.join(belief_state.split()[1:])
# split belief state slot-value pairs
slot_value_pairs = belief_state.split(';')
for slot_value_pair in slot_value_pairs:
if slot_value_pair.strip() == '':
continue
slot_values = slot_value_pair.split(' = ')
if len(slot_values) != 2:
continue
else:
slot, value = slot_values
slot = slot.strip().lower()
value = value.strip().lower()
if value in INVALID_SLOT_VALUES:
continue
belief_slot_value_map[slot] = value
return belief_slot_value_map
class BaselineDSTEvaluator:
def __init__(self, predictions_output_file, evaluation_file):
"""
create an Evaluator object for evaluating Baseline DST predictions
Args:
predictions_output_file: path of the predictions output JSON file
evaluation_file: path of the test/valid data file for extracting ground truth data
"""
# load predictions output json file
self.predictions = json.load(open(predictions_output_file))
# load test/valid data json file
self.eval_data = json.load(open(evaluation_file))
# do some length checks here
if len(self.predictions) == 0 or len(self.eval_data) == 0:
raise ValueError('Invalid Data (no items) in prediction or evaluation data file!')
if len(self.predictions) != len(self.eval_data):
raise ValueError('Length mismatch!')
def parse_prediction_belief_states(self):
belief_state_list = []
for prediction in self.predictions:
last_line = prediction[-1]
last_line = last_line.strip()
# remove system responses from the predictions
belief_prediction = last_line.split('system :')[0]
belief_slot_value_map = create_slot_value_map_from_belief_prediction(belief_prediction)
belief_state_list.append(belief_slot_value_map)
return belief_state_list
def parse_true_belief_states(self):
true_belief_state_list = []
for item in self.eval_data:
belief_prediction = item['belief']
belief_slot_value_map = create_slot_value_map_from_belief_prediction(belief_prediction)
true_belief_state_list.append(belief_slot_value_map)
return true_belief_state_list
def compute_joint_goal_accuracy(self, true_states, prediction_states):
print('Computing Joint Goal Accuracy metric...!')
if len(true_states) != len(prediction_states):
raise ValueError('Length mismatch!')
correctly_predicted, total_turns = 0, 0
for truth, prediction in zip(true_states, prediction_states):
total_turns += 1
if set(truth.keys()) != set(prediction.keys()):
continue
for slot in truth:
if truth[slot] != prediction[slot]:
break
correctly_predicted += 1
print('Evaluation :: Joint Goal Accuracy = ', (correctly_predicted / total_turns) * 100)
evaluator = BaselineDSTEvaluator('../data/outputs/experiment-20220831/125-dpd/checkpoint-90000/output_test.json',
'../data/baseline/test/test.soloist.json')
predicted_belief_states = evaluator.parse_prediction_belief_states()
true_belief_states = evaluator.parse_true_belief_states()
evaluator.compute_joint_goal_accuracy(true_belief_states, predicted_belief_states)

201
baseline/soloist_decode.py
Unescape View File

@ -0,0 +1,201 @@
from __future__ import absolute_import, division, print_function, unicode_literals
import argparse
import logging
from tqdm import trange
import json
import torch
import torch.nn.functional as F
import numpy as np
import sys
sys.path.append('.')
sys.path.append('./transformers')
sys.path.append('./transformers/')
from transformers import GPT2Config
from transformers import GPT2LMHeadModel, GPT2Tokenizer
logging.basicConfig(format = '%(asctime)s - %(levelname)s - %(name)s - %(message)s',
datefmt = '%m/%d/%Y %H:%M:%S',
level = logging.INFO)
logger = logging.getLogger(__name__)
MAX_LENGTH = int(150) # Hardcoded max length to avoid infinite loop
MODEL_CLASSES = {
'gpt2': (GPT2LMHeadModel, GPT2Tokenizer)
}
def set_seed(args):
np.random.seed(args.seed)
torch.manual_seed(args.seed)
if args.n_gpu > 0:
torch.cuda.manual_seed_all(args.seed)
def top_k_top_p_filtering(logits, top_k=0, top_p=0.0, filter_value=-float('Inf')):
""" Filter a distribution of logits using top-k and/or nucleus (top-p) filtering
Args:
logits: logits distribution shape (batch size x vocabulary size)
top_k > 0: keep only top k tokens with highest probability (top-k filtering).
top_p > 0.0: keep the top tokens with cumulative probability >= top_p (nucleus filtering).
Nucleus filtering is described in Holtzman et al. (http://arxiv.org/abs/1904.09751)
From: https://gist.github.com/thomwolf/1a5a29f6962089e871b94cbd09daf317
"""
top_k = min(top_k, logits.size(-1)) # Safety check
if top_k > 0:
# Remove all tokens with a probability less than the last token of the top-k
indices_to_remove = logits < torch.topk(logits, top_k)[0][..., -1, None]
logits[indices_to_remove] = filter_value
if top_p > 0.0:
sorted_logits, sorted_indices = torch.sort(logits, descending=True)
cumulative_probs = torch.cumsum(F.softmax(sorted_logits, dim=-1), dim=-1)
# Remove tokens with cumulative probability above the threshold
sorted_indices_to_remove = cumulative_probs > top_p
# Shift the indices to the right to keep also the first token above the threshold
sorted_indices_to_remove[..., 1:] = sorted_indices_to_remove[..., :-1].clone()
sorted_indices_to_remove[..., 0] = 0
# scatter sorted tensors to original indexing
indices_to_remove = sorted_indices_to_remove.scatter(dim=1, index=sorted_indices, src=sorted_indices_to_remove)
logits[indices_to_remove] = filter_value
return logits
def sample_sequence(model, length, context, token_type_ids, system_token_id, num_samples=1, temperature=1, top_k=0, top_p=0.0, repetition_penalty=1.0, device='cpu'):
context = torch.tensor(context, dtype=torch.long, device=device)
context = context.unsqueeze(0).repeat(num_samples, 1)
token_type_ids = torch.tensor(token_type_ids, dtype=torch.long, device=device)
token_type_ids = token_type_ids.unsqueeze(0).repeat(num_samples, 1)
system_token_id = torch.tensor(system_token_id, dtype=torch.long, device=device)
system_token_id = system_token_id.unsqueeze(0).repeat(num_samples, 1)
generated = context
with torch.no_grad():
for _ in range(length):
inputs = {'input_ids': generated, 'token_type_ids':token_type_ids}
outputs = model(**inputs) # Note: we could also use 'past' with GPT-2/Transfo-XL/XLNet/CTRL (cached hidden-states)
next_token_logits = outputs[0][:, -1, :] / (temperature if temperature > 0 else 1.)
# repetition penalty from CTRL (https://arxiv.org/abs/1909.05858)
for i in range(num_samples):
for _ in set(generated[i].tolist()):
next_token_logits[i, _] /= repetition_penalty
filtered_logits = top_k_top_p_filtering(next_token_logits, top_k=top_k, top_p=top_p)
if temperature == 0: # greedy sampling:
next_token = torch.argmax(filtered_logits, dim=-1).unsqueeze(-1)
else:
next_token = torch.multinomial(F.softmax(filtered_logits, dim=-1), num_samples=1)
generated = torch.cat((generated, next_token), dim=1)
token_type_ids = torch.cat((token_type_ids, system_token_id), dim=1)
return generated
def main():
parser = argparse.ArgumentParser()
parser.add_argument("--model_type", default='gpt2', type=str, help="Model type selected in the list: " + ", ".join(MODEL_CLASSES.keys()))
parser.add_argument("--model_name_or_path", default=None, type=str, required=True, help="Path to pre-trained model or shortcut name selected in the list")
parser.add_argument("--padding_text", type=str, default="")
parser.add_argument("--xlm_lang", type=str, default="", help="Optional language when used with the XLM model.")
parser.add_argument("--length", type=int, default=110)
parser.add_argument("--num_samples", type=int, default=1)
parser.add_argument("--temperature", type=float, default=1.0, help="temperature of 0 implies greedy sampling")
parser.add_argument("--repetition_penalty", type=float, default=1.0, help="primarily useful for CTRL model; in that case, use 1.2")
parser.add_argument("--top_k", type=int, default=0)
parser.add_argument("--top_p", type=float, default=0.9)
parser.add_argument("--no_cuda", action='store_true', help="Avoid using CUDA when available")
parser.add_argument('--seed', type=int, default=42, help="random seed for initialization")
parser.add_argument('--stop_token', type=str, default='<|endoftext|>', help="Token at which text generation is stopped")
parser.add_argument('--input_file', type=str, default=None, help="input json file to decoding")
parser.add_argument('--output_file', type=str, default=None, help="save path")
parser.add_argument('--max_turn', type=int, default=15, help="number of turns used as context")
args = parser.parse_args()
# setup CUDA device
args.device = torch.device("cuda" if torch.cuda.is_available() and not args.no_cuda else "cpu")
args.n_gpu = torch.cuda.device_count()
set_seed(args)
# setup HuggingFace Model
args.model_type = args.model_type.lower()
model_class, tokenizer_class = MODEL_CLASSES[args.model_type]
tokenizer = tokenizer_class.from_pretrained(args.model_name_or_path)
model = model_class.from_pretrained(args.model_name_or_path)
model.to(args.device)
model.eval()
if args.length < 0 and model.config.max_position_embeddings > 0:
args.length = model.config.max_position_embeddings
elif 0 < model.config.max_position_embeddings < args.length:
args.length = model.config.max_position_embeddings # No generation bigger than model size
elif args.length < 0:
args.length = MAX_LENGTH # avoid infinite loop
logger.info(args)
inputs = json.load(open(args.input_file))
output_tests = []
system_token_id = tokenizer.convert_tokens_to_ids(['system'])
user_token_id = tokenizer.convert_tokens_to_ids(['user'])
for idx in range(len(inputs)):
logger.info(f"PROGRESS: {int(idx/len(inputs)*100)}%")
example = inputs[idx]
history = example['history']
context = history[-args.max_turn:]
context_ids = []
token_ids_for_context = []
for cxt in context:
ids = tokenizer.convert_tokens_to_ids(tokenizer.tokenize(cxt))
context_ids += ids
if 'user :' in cxt:
token_ids_for_context += user_token_id * len(ids)
else:
token_ids_for_context += system_token_id * len(ids)
response = '=>'
response_id = tokenizer.convert_tokens_to_ids(tokenizer.tokenize(response))
context_tokens = context_ids + response_id
token_type_ids = token_ids_for_context + system_token_id
assert( len(context_tokens) == len(token_type_ids))
out = sample_sequence(
model=model,
context=context_tokens,
token_type_ids=token_type_ids,
system_token_id=system_token_id,
num_samples=args.num_samples,
length=args.length,
temperature=args.temperature,
top_k=args.top_k,
top_p=args.top_p,
repetition_penalty=args.repetition_penalty,
device=args.device,
)
out = out[:, len(context_tokens):].tolist()
examples = []
for o in out:
text = tokenizer.decode(o, clean_up_tokenization_spaces=True)
text = text[: text.find(args.stop_token) if args.stop_token else None]
examples.append(text)
output_tests.append(examples)
print(output_tests)
json.dump(output_tests, open(args.output_file,'w'), indent=2)
return text
if __name__ == '__main__':
main()

739
baseline/soloist_train.py
Unescape View File

@ -0,0 +1,739 @@
from __future__ import absolute_import, division, print_function
import argparse
import glob
import logging
import os
import pickle
import random
import re
import shutil
import time
import json
import sys
sys.path.append('.')
sys.path.append('./transformers')
sys.path.append('./transformers/')
import numpy as np
import torch
from torch.utils.data import DataLoader, Dataset, SequentialSampler, RandomSampler
from torch.utils.data.distributed import DistributedSampler
try:
from torch.utils.tensorboard import SummaryWriter
except:
from tensorboardX import SummaryWriter
from tqdm import tqdm, trange
from transformers import (WEIGHTS_NAME, AdamW, get_linear_schedule_with_warmup,GPT2Config, GPT2DoubleHeadsModel, GPT2Tokenizer)
logging.basicConfig(format = '%(asctime)s - %(levelname)s - %(name)s - %(message)s',
datefmt = '%m/%d/%Y %H:%M:%S',
level = logging.INFO)
logger = logging.getLogger(__name__)
MODEL_CLASSES = {
'gpt2': (GPT2Config, GPT2DoubleHeadsModel, GPT2Tokenizer)
}
s = 0
s_time = time.time()
def log_every_n_interval(interval, msg):
global s, s_time
s += 1
if s % interval == 0:
if s_time is None:
s_time = time.time()
iter_p_s = 0
else:
e_time = time.time()
elapse = e_time - s_time
iter_p_s = interval / elapse
s_time = e_time
logger.info(f'MSG: {msg}; ITER: {iter_p_s:.2f}/s')
class JsonDataset(Dataset):
def __init__(self, tokenizer, args, file_path='train', max_seq=80, max_turn=1, seperator=' & '):
assert os.path.isfile(file_path)
directory, filename = os.path.split(file_path)
cached_features_file = os.path.join(directory, args.output_dir + '_cached_lm' + '_seqlen_' + str(max_seq) + '_' + filename)
if os.path.exists(cached_features_file) and not args.overwrite_cache:
logger.info("Loading features from cached file %s", cached_features_file)
with open(cached_features_file, 'rb') as handle:
self.examples = pickle.load(handle)
else:
logger.info(f"Creating features from dataset file at {directory}")
self.examples = []
self.labels = []
self.token_ids = []
self.attention_masks = []
self.mc_token_ids = []
self.mc_labels = []
system_token_id = tokenizer.convert_tokens_to_ids(['system'])
user_token_id = tokenizer.convert_tokens_to_ids(['user'])
induction_token_id = tokenizer.convert_tokens_to_ids(['=>'])
examples = json.load(open(file_path))
response_pool = []
belief_pool = []
idxs = list(range(len(examples)))
for i in examples:
response = i['reply'] + ' '+tokenizer.eos_token
response_id = tokenizer.convert_tokens_to_ids(tokenizer.tokenize(response))
response_pool.append(response_id)
belief_id = tokenizer.convert_tokens_to_ids(tokenizer.tokenize(i['belief']))
belief_pool.append(belief_id)
for example in examples:
history = example['history']
context = history[-max_turn:]
context_ids = []
token_ids_for_context = []
for cxt in context:
ids = tokenizer.convert_tokens_to_ids(tokenizer.tokenize(cxt))
context_ids += ids
if 'user :' in cxt:
token_ids_for_context += user_token_id * len(ids)
else:
token_ids_for_context += system_token_id * len(ids)
history = ' '.join(history[-max_turn:])
kb = ' '#example['kb']
belief = example['belief']
belief_id = tokenizer.convert_tokens_to_ids(tokenizer.tokenize(belief))
response = example['reply'] + ' '+tokenizer.eos_token
response_id = tokenizer.convert_tokens_to_ids(tokenizer.tokenize(response))
token_id = token_ids_for_context + system_token_id + system_token_id * len(belief_id) + system_token_id * len(response_id)
source = context_ids + induction_token_id + belief_id + response_id
if args.with_LM:
target = source
else:
target = [-1] * len(context_ids) + [-1] * len(induction_token_id) + belief_id + response_id
if len(source) < max_seq:
attention_mask = [0] * max_seq
attention_mask[:len(source)] = [1] * len(source)
self.mc_token_ids.append(len(source) - 1)
source += [0] * (max_seq - len(source))
target += [-1] * (max_seq - len(target))
token_id += [0] * (max_seq - len(token_id))
else:
attention_mask = [1] * max_seq
self.mc_token_ids.append(max_seq - 1)
source = source[-max_seq:]
target = target[-max_seq:]
token_id = token_id[-max_seq:]
self.mc_labels.append(0)
if not len(source) == len(target) == len(token_id) == len(attention_mask):
import pdb
pdb.set_trace()
self.examples.append(source)
self.labels.append(target)
self.token_ids.append(token_id)
self.attention_masks.append(attention_mask)
if args.add_same_belief_response_prediction:
for _ in range(args.num_candidates):
random_idx = random.choice(idxs)
new_response_id = response_pool[random_idx]
new_belief_id = belief_pool[random_idx]
source = context_ids + induction_token_id + new_belief_id + new_response_id
token_id = token_ids_for_context + system_token_id + system_token_id * len(new_belief_id) + system_token_id * len(new_response_id)
if len(source) < max_seq:
attention_mask = [0] * max_seq
attention_mask[:len(source)] = [1] * len(source)
self.mc_token_ids.append(len(source)-1)
source += [0] * (max_seq - len(source))
token_id += [0] * (max_seq - len(token_id))
else:
attention_mask = [1] * max_seq
self.mc_token_ids.append(max_seq - 1)
source = source[-max_seq:]
target = target[-max_seq:]
token_id = token_id[-max_seq:]
self.examples.append(source)
self.labels.append([-1] * len(source))
self.token_ids.append(token_id)
self.mc_labels.append(1)
self.attention_masks.append(attention_mask)
if args.add_response_prediction:
for _ in range(args.num_candidates):
random_idx = random.choice(idxs)
new_response_id = response_pool[random_idx]
new_belief_id = belief_pool[random_idx]
source = context_ids + induction_token_id + belief_id + new_response_id
token_id = token_ids_for_context + system_token_id + system_token_id * len(belief_id) + system_token_id * len(new_response_id)
if len(source) < max_seq:
attention_mask = [0] * max_seq
attention_mask[:len(source)] = [1] * len(source)
self.mc_token_ids.append(len(source)-1)
source += [0] * (max_seq - len(source))
token_id += [0] * (max_seq - len(token_id))
else:
attention_mask = [1] * max_seq
self.mc_token_ids.append(max_seq - 1)
source = source[-max_seq:]
target = target[-max_seq:]
token_id = token_id[-max_seq:]
self.examples.append(source)
self.labels.append([-1] * len(source))
self.token_ids.append(token_id)
self.mc_labels.append(1)
self.attention_masks.append(attention_mask)
if args.add_belief_prediction:
for _ in range(args.num_candidates):
random_idx = random.choice(idxs)
new_response_id = response_pool[random_idx]
new_belief_id = belief_pool[random_idx]
source = context_ids + induction_token_id + new_belief_id + response_id
token_id = token_ids_for_context + system_token_id + system_token_id * len(new_belief_id) + system_token_id * len(response_id)
if len(source) < max_seq:
attention_mask = [0] * max_seq
attention_mask[:len(source)] = [1] * len(source)
self.mc_token_ids.append(len(source)-1)
source += [0] * (max_seq - len(source))
token_id += [0] * (max_seq - len(token_id))
else:
attention_mask = [1] * max_seq
self.mc_token_ids.append(max_seq - 1)
source = source[-max_seq:]
target = target[-max_seq:]
token_id = token_id[-max_seq:]
self.examples.append(source)
self.labels.append([-1] * len(source))
self.token_ids.append(token_id)
self.mc_labels.append(1)
self.attention_masks.append(attention_mask)
def __len__(self):
return len(self.examples)
def __getitem__(self, item):
return torch.tensor(self.examples[item]), torch.tensor(self.token_ids[item]), torch.tensor(self.labels[item]), torch.tensor(self.attention_masks[item]), torch.tensor(self.mc_labels[item]), torch.tensor(self.mc_token_ids[item]),
def load_and_cache_examples(args, tokenizer, evaluate=False):
dataset = JsonDataset(tokenizer, args, file_path=args.eval_data_file if evaluate else args.train_data_file, max_seq=args.max_seq, max_turn=args.max_turn)
return dataset
def set_seed(args):
random.seed(args.seed)
np.random.seed(args.seed)
torch.manual_seed(args.seed)
if args.n_gpu > 0:
torch.cuda.manual_seed_all(args.seed)
def _rotate_checkpoints(args, checkpoint_prefix, use_mtime=False):
if not args.save_total_limit:
return
if args.save_total_limit <= 0:
return
# Check if we should delete older checkpoint(s)
glob_checkpoints = glob.glob(os.path.join(args.output_dir, '{}-*'.format(checkpoint_prefix)))
if len(glob_checkpoints) <= args.save_total_limit:
return
ordering_and_checkpoint_path = []
for path in glob_checkpoints:
if use_mtime:
ordering_and_checkpoint_path.append((os.path.getmtime(path), path))
else:
regex_match = re.match('.*{}-([0-9]+)'.format(checkpoint_prefix), path)
if regex_match and regex_match.groups():
ordering_and_checkpoint_path.append((int(regex_match.groups()[0]), path))
checkpoints_sorted = sorted(ordering_and_checkpoint_path)
checkpoints_sorted = [checkpoint[1] for checkpoint in checkpoints_sorted]
number_of_checkpoints_to_delete = max(0, len(checkpoints_sorted) - args.save_total_limit)
checkpoints_to_be_deleted = checkpoints_sorted[:number_of_checkpoints_to_delete]
for checkpoint in checkpoints_to_be_deleted:
logger.info("Deleting older checkpoint [{}] due to args.save_total_limit".format(checkpoint))
shutil.rmtree(checkpoint)
def mask_tokens(inputs, tokenizer, args):
""" Prepare masked tokens inputs/labels for masked language modeling: 80% MASK, 10% random, 10% original. """
labels = inputs.clone()
# We sample a few tokens in each sequence for masked-LM training (with probability args.mlm_probability defaults to 0.15 in Bert/RoBERTa)
probability_matrix = torch.full(labels.shape, args.mlm_probability)
special_tokens_mask = [tokenizer.get_special_tokens_mask(val, already_has_special_tokens=True) for val in labels.tolist()]
probability_matrix.masked_fill_(torch.tensor(special_tokens_mask, dtype=torch.bool), value=0.0)
masked_indices = torch.bernoulli(probability_matrix).bool()
labels[~masked_indices] = -1 # We only compute loss on masked tokens
# 80% of the time, we replace masked input tokens with tokenizer.mask_token ([MASK])
indices_replaced = torch.bernoulli(torch.full(labels.shape, 0.8)).bool() & masked_indices
inputs[indices_replaced] = tokenizer.convert_tokens_to_ids(tokenizer.mask_token)
# 10% of the time, we replace masked input tokens with random word
indices_random = torch.bernoulli(torch.full(labels.shape, 0.5)).bool() & masked_indices & ~indices_replaced
random_words = torch.randint(len(tokenizer), labels.shape, dtype=torch.long)
inputs[indices_random] = random_words[indices_random]
# The rest of the time (10% of the time) we keep the masked input tokens unchanged
return inputs, labels
def train(args, train_dataset, model, tokenizer):
""" Train the model """
if args.local_rank in [-1, 0]:
tb_writer = SummaryWriter()
args.train_batch_size = args.per_gpu_train_batch_size * max(1, args.n_gpu)
train_sampler = RandomSampler(train_dataset) if args.local_rank == -1 else DistributedSampler(train_dataset)
train_dataloader = DataLoader(train_dataset, sampler=train_sampler, batch_size=args.train_batch_size)
if args.max_steps > 0:
t_total = args.max_steps
args.num_train_epochs = args.max_steps // (len(train_dataloader) // args.gradient_accumulation_steps) + 1
else:
t_total = len(train_dataloader) // args.gradient_accumulation_steps * args.num_train_epochs
# Prepare optimizer and schedule (linear warmup and decay)
no_decay = ['bias', 'LayerNorm.weight']
optimizer_grouped_parameters = [
{'params': [p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay)], 'weight_decay': args.weight_decay},
{'params': [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)], 'weight_decay': 0.0}
]
optimizer = AdamW(optimizer_grouped_parameters, lr=args.learning_rate, eps=args.adam_epsilon)
scheduler = get_linear_schedule_with_warmup(optimizer, num_warmup_steps=args.warmup_steps, num_training_steps=t_total)
if args.fp16:
try:
from apex import amp
except ImportError:
raise ImportError("Please install apex from https://www.github.com/nvidia/apex to use fp16 training.")
model, optimizer = amp.initialize(model, optimizer, opt_level=args.fp16_opt_level)
model.resize_token_embeddings(len(tokenizer))
# multi-gpu training (should be after apex fp16 initialization)
if args.n_gpu > 1:
model = torch.nn.DataParallel(model)
# Distributed training (should be after apex fp16 initialization)
if args.local_rank != -1:
model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.local_rank],
output_device=args.local_rank,
find_unused_parameters=True)
# Train!
logger.info("***** Running training *****")
logger.info(" Num examples = %d", len(train_dataset))
logger.info(" Num Epochs = %d", args.num_train_epochs)
logger.info(" Instantaneous batch size per GPU = %d", args.per_gpu_train_batch_size)
logger.info(" Total train batch size (w. parallel, distributed & accumulation) = %d",
args.train_batch_size * args.gradient_accumulation_steps * (torch.distributed.get_world_size() if args.local_rank != -1 else 1))
logger.info(" Gradient Accumulation steps = %d", args.gradient_accumulation_steps)
logger.info(" Total optimization steps = %d", t_total)
global_step = 0
tr_loss, logging_loss = 0.0, 0.0
model.zero_grad()
train_iterator = trange(int(args.num_train_epochs), desc="Epoch", disable=args.local_rank not in [-1, 0])
set_seed(args) # Added here for reproducibility (even between python 2 and 3)
for e in train_iterator:
# epoch_iterator = tqdm(train_dataloader, desc="Iteration", disable=args.local_rank not in [-1, 0])
for step, batch in enumerate(train_dataloader):
# inputs, labels = mask_tokens(batch, tokenizer, args) if args.mlm else (batch, batch)
# logger.info(f" PROGRESS: {float(global_step)/t_total*100}%")
log_every_n_interval(500, f" PROGRESS: {int(float(global_step)/t_total*100)}%")
if step % 500 == 0:
logger.info(f" PROGRESS: {int(float(global_step)/t_total*100)}%")
inputs, tokens, labels, masks,mc_labels, mc_token_ids = batch
inputs = inputs.to(args.device)
tokens = tokens.to(args.device)
labels = labels.to(args.device)
masks = masks.to(args.device)
mc_labels = mc_labels.to(args.device)
mc_token_ids = mc_token_ids.to(args.device)
model.train()
outputs = model(inputs, lm_labels=labels, mc_labels=mc_labels, mc_token_ids=mc_token_ids, token_type_ids=tokens, attention_mask=masks)
lm_loss = outputs[0] # model outputs are always tuple in transformers (see doc)
mc_loss = outputs[1]
loss = lm_loss + args.mc_loss_efficient * mc_loss
if args.n_gpu > 1:
loss = loss.mean() # mean() to average on multi-gpu parallel training
if args.gradient_accumulation_steps > 1:
loss = loss / args.gradient_accumulation_steps
if args.fp16:
with amp.scale_loss(loss, optimizer) as scaled_loss:
scaled_loss.backward()
else:
loss.backward()
tr_loss += loss.item()
if (step + 1) % args.gradient_accumulation_steps == 0:
if args.fp16:
torch.nn.utils.clip_grad_norm_(amp.master_params(optimizer), args.max_grad_norm)
else:
torch.nn.utils.clip_grad_norm_(model.parameters(), args.max_grad_norm)
optimizer.step()
scheduler.step() # Update learning rate schedule
model.zero_grad()
global_step += 1
if args.local_rank in [-1, 0] and args.logging_steps > 0 and global_step % args.logging_steps == 0:
# Log metrics
if args.local_rank == -1 and args.evaluate_during_training: # Only evaluate when single GPU otherwise metrics may not average well
results = evaluate(args, model, tokenizer)
for key, value in results.items():
tb_writer.add_scalar('eval_{}'.format(key), value, global_step)
tb_writer.add_scalar('lr', scheduler.get_lr()[0], global_step)
tb_writer.add_scalar('loss', (tr_loss - logging_loss)/args.logging_steps, global_step)
logger.info(f" EVALERR: {(tr_loss - logging_loss)/float(args.logging_steps)}")
logging_loss = tr_loss
if args.local_rank in [-1, 0] and args.save_steps > 0 and global_step % args.save_steps == 0:
checkpoint_prefix = 'checkpoint'
# Save model checkpoint
output_dir = os.path.join(args.output_dir, '{}-{}'.format(checkpoint_prefix, global_step))
if not os.path.exists(output_dir):
os.makedirs(output_dir)
model_to_save = model.module if hasattr(model, 'module') else model # Take care of distributed/parallel training
model_to_save.save_pretrained(output_dir)
tokenizer.save_pretrained(output_dir)
torch.save(args, os.path.join(output_dir, 'training_args.bin'))
logger.info("Saving model checkpoint to %s", output_dir)
_rotate_checkpoints(args, checkpoint_prefix)
# if args.max_steps > 0 and global_step > args.max_steps:
# epoch_iterator.close()
# break
if args.max_steps > 0 and global_step > args.max_steps:
train_iterator.close()
break
if args.local_rank in [-1, 0]:
tb_writer.close()
return global_step, tr_loss / global_step
def evaluate(args, model, tokenizer, prefix=""):
# Loop to handle MNLI double evaluation (matched, mis-matched)
eval_output_dir = args.output_dir
eval_dataset = load_and_cache_examples(args, tokenizer, evaluate=True)
if not os.path.exists(eval_output_dir) and args.local_rank in [-1, 0]:
os.makedirs(eval_output_dir)
args.eval_batch_size = args.per_gpu_eval_batch_size * max(1, args.n_gpu)
# Note that DistributedSampler samples randomly
eval_sampler = SequentialSampler(eval_dataset) if args.local_rank == -1 else DistributedSampler(eval_dataset)
eval_dataloader = DataLoader(eval_dataset, sampler=eval_sampler, batch_size=args.eval_batch_size)
# multi-gpu evaluate
if args.n_gpu > 1:
model = torch.nn.DataParallel(model)
# Eval!
logger.info("***** Running evaluation {} *****".format(prefix))
logger.info(" Num examples = %d", len(eval_dataset))
logger.info(" Batch size = %d", args.eval_batch_size)
eval_loss = 0.0
nb_eval_steps = 0
model.eval()
for batch in tqdm(eval_dataloader, desc="Evaluating"):
# inputs, labels = mask_tokens(batch, tokenizer, args) if args.mlm else (batch, batch)
inputs, tokens, labels, masks = batch
# import pdb
# pdb.set_trace()
inputs = inputs.to(args.device)
tokens = tokens.to(args.device)
labels = labels.to(args.device)
masks = masks.to(args.device)
# inputs = inputs.to(args.device)
# labels = labels.to(args.device)
with torch.no_grad():
outputs = model(inputs, masked_lm_labels=labels, token_type_ids=tokens) if args.mlm else model(inputs, labels=labels)
lm_loss = outputs[0]
eval_loss += lm_loss.mean().item()
nb_eval_steps += 1
eval_loss = eval_loss / nb_eval_steps
perplexity = torch.exp(torch.tensor(eval_loss))
result = {
"perplexity": perplexity
}
output_eval_file = os.path.join(eval_output_dir, prefix, "eval_results.txt")
with open(output_eval_file, "w") as writer:
logger.info("***** Eval results {} *****".format(prefix))
for key in sorted(result.keys()):
logger.info(" %s = %s", key, str(result[key]))
writer.write("%s = %s\n" % (key, str(result[key])))
return result
def main():
parser = argparse.ArgumentParser()
## Required parameters
parser.add_argument("--train_data_file", default=None, type=str, required=True,
help="The input training data file (a text file).")
parser.add_argument("--output_dir", default=None, type=str, required=True,
help="The output directory where the model predictions and checkpoints will be written.")
## Other parameters
parser.add_argument("--eval_data_file", default=None, type=str,
help="An optional input evaluation data file to evaluate the perplexity on (a text file).")
parser.add_argument("--model_type", default="bert", type=str,
help="The model architecture to be fine-tuned.")
parser.add_argument("--model_name_or_path", default="bert-base-cased", type=str,
help="The model checkpoint for weights initialization.")
parser.add_argument("--mlm", action='store_true',
help="Train with masked-language modeling loss instead of language modeling.")
parser.add_argument("--mlm_probability", type=float, default=0.15,
help="Ratio of tokens to mask for masked language modeling loss")
parser.add_argument("--config_name", default="", type=str,
help="Optional pretrained config name or path if not the same as model_name_or_path")
parser.add_argument("--tokenizer_name", default="", type=str,
help="Optional pretrained tokenizer name or path if not the same as model_name_or_path")
parser.add_argument("--cache_dir", default="", type=str,
help="Optional directory to store the pre-trained models downloaded from s3 (instread of the default one)")
parser.add_argument("--block_size", default=80, type=int,
help="Optional input sequence length after tokenization."
"The training dataset will be truncated in block of this size for training."
"Default to the model max input length for single sentence inputs (take into account special tokens).")
parser.add_argument("--do_train", action='store_true',
help="Whether to run training.")
parser.add_argument("--do_eval", action='store_true',
help="Whether to run eval on the dev set.")
parser.add_argument("--evaluate_during_training", action='store_true',
help="Run evaluation during training at each logging step.")
parser.add_argument("--do_lower_case", action='store_true',
help="Set this flag if you are using an uncased model.")
parser.add_argument("--per_gpu_train_batch_size", default=1, type=int,
help="Batch size per GPU/CPU for training.")
parser.add_argument("--per_gpu_eval_batch_size", default=1, type=int,
help="Batch size per GPU/CPU for evaluation.")
parser.add_argument('--gradient_accumulation_steps', type=int, default=1,
help="Number of updates steps to accumulate before performing a backward/update pass.")
parser.add_argument("--learning_rate", default=5e-5, type=float,
help="The initial learning rate for Adam.")
parser.add_argument("--weight_decay", default=0.0, type=float,
help="Weight deay if we apply some.")
parser.add_argument("--adam_epsilon", default=1e-8, type=float,
help="Epsilon for Adam optimizer.")
parser.add_argument("--max_grad_norm", default=1.0, type=float,
help="Max gradient norm.")
parser.add_argument("--num_train_epochs", default=1.0, type=float,
help="Total number of training epochs to perform.")
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_steps", default=0, type=int,
help="Linear warmup over warmup_steps.")
parser.add_argument('--logging_steps', type=int, default=10,
help="Log every X updates steps.")
parser.add_argument('--save_steps', type=int, default=5000,
help="Save checkpoint every X updates steps.")
parser.add_argument('--save_total_limit', type=int, default=None,
help='Limit the total amount of checkpoints, delete the older checkpoints in the output_dir, does not delete by default')
parser.add_argument("--eval_all_checkpoints", action='store_true',
help="Evaluate all checkpoints starting with the same prefix as model_name_or_path ending and ending with step number")
parser.add_argument("--no_cuda", action='store_true',
help="Avoid using CUDA when available")
parser.add_argument('--overwrite_output_dir', action='store_true',
help="Overwrite the content of the output directory")
parser.add_argument('--overwrite_cache', action='store_true',
help="Overwrite the cached training and evaluation sets")
parser.add_argument('--seed', type=int, default=42,
help="random seed for initialization")
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("--local_rank", type=int, default=-1,
help="For distributed training: local_rank")
parser.add_argument('--server_ip', type=str, default='', help="For distant debugging.")
parser.add_argument('--server_port', type=str, default='', help="For distant debugging.")
parser.add_argument('--text_chunk', action='store_true', help="")
parser.add_argument('--with_LM', type=bool, default=True, help="")
parser.add_argument("--max_seq", default=200, type=int, help="")
parser.add_argument("--max_turn", default=1, type=int, help="")
parser.add_argument("--mc_loss_efficient", default=1, type=float, help="")
parser.add_argument("--num_candidates", default=1, type=int, help="")
parser.add_argument("--add_special_action_tokens", default='', type=str)
parser.add_argument("--add_same_belief_response_prediction", action='store_true')
parser.add_argument("--add_response_prediction", action='store_true')
parser.add_argument("--add_belief_prediction", action='store_true')
args = parser.parse_args()
if args.model_type in ["bert", "roberta", "distilbert"] and not args.mlm:
raise ValueError("BERT and RoBERTa do not have LM heads but masked LM heads. They must be run using the --mlm "
"flag (masked language modeling).")
if args.eval_data_file is None and args.do_eval:
raise ValueError("Cannot do evaluation without an evaluation data file. Either supply a file to --eval_data_file "
"or remove the --do_eval argument.")
if os.path.exists(args.output_dir) and os.listdir(args.output_dir) and args.do_train and not args.overwrite_output_dir:
raise ValueError("Output directory ({}) already exists and is not empty. Use --overwrite_output_dir to overcome.".format(args.output_dir))
# Setup distant debugging if needed
if args.server_ip and args.server_port:
# Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
import ptvsd
print("Waiting for debugger attach")
ptvsd.enable_attach(address=(args.server_ip, args.server_port), redirect_output=True)
ptvsd.wait_for_attach()
# 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()
else: # Initializes the distributed backend which will take care of sychronizing nodes/GPUs
torch.cuda.set_device(args.local_rank)
device = torch.device("cuda", args.local_rank)
torch.distributed.init_process_group(backend='nccl')
args.n_gpu = 1
args.device = device
# Setup logging
logging.basicConfig(format = '%(asctime)s - %(levelname)s - %(name)s - %(message)s',
datefmt = '%m/%d/%Y %H:%M:%S',
level = logging.INFO if args.local_rank in [-1, 0] else logging.WARN)
logger.warning("Process rank: %s, device: %s, n_gpu: %s, distributed training: %s, 16-bits training: %s",
args.local_rank, device, args.n_gpu, bool(args.local_rank != -1), args.fp16)
# Set seed
set_seed(args)
# Load pretrained model and tokenizer
if args.local_rank not in [-1, 0]:
torch.distributed.barrier() # Barrier to make sure only the first process in distributed training download model & vocab
config_class, model_class, tokenizer_class = MODEL_CLASSES[args.model_type]
config = config_class.from_pretrained(args.config_name if args.config_name else args.model_name_or_path,
cache_dir=args.cache_dir if args.cache_dir else None)
config.num_labels = 2
tokenizer = tokenizer_class.from_pretrained(args.tokenizer_name if args.tokenizer_name else args.model_name_or_path,
do_lower_case=args.do_lower_case,
cache_dir=args.cache_dir if args.cache_dir else None)
if args.block_size <= 0:
args.block_size = tokenizer.max_len_single_sentence # Our input block size will be the max possible for the model
args.block_size = min(args.block_size, tokenizer.max_len_single_sentence)
model = model_class.from_pretrained(args.model_name_or_path,
from_tf=bool('.ckpt' in args.model_name_or_path),
config=config,
cache_dir=args.cache_dir if args.cache_dir else None)
if args.add_special_action_tokens:
special_tokens = []
for line in open(args.add_special_action_tokens):
special_tokens.append(line.strip())
tokenizer.add_tokens(special_tokens)
model.resize_token_embeddings(len(tokenizer))
model.to(args.device)
if args.local_rank == 0:
torch.distributed.barrier() # End of barrier to make sure only the first process in distributed training download model & vocab
logger.info("Training/evaluation parameters %s", args)
# Training
if args.do_train:
if args.local_rank not in [-1, 0]:
torch.distributed.barrier() # Barrier to make sure only the first process in distributed training process the dataset, and the others will use the cache
train_dataset = load_and_cache_examples(args, tokenizer, evaluate=False)
if args.local_rank == 0:
torch.distributed.barrier()
global_step, tr_loss = train(args, train_dataset, model, tokenizer)
logger.info(" global_step = %s, average loss = %s", global_step, tr_loss)
# Saving best-practices: if you use save_pretrained for the model and tokenizer, you can reload them using from_pretrained()
if args.do_train and (args.local_rank == -1 or torch.distributed.get_rank() == 0):
# Create output directory if needed
if not os.path.exists(args.output_dir) and args.local_rank in [-1, 0]:
os.makedirs(args.output_dir)
logger.info("Saving model checkpoint to %s", args.output_dir)
# Save a trained model, configuration and tokenizer using `save_pretrained()`.
# They can then be reloaded using `from_pretrained()`
model_to_save = model.module if hasattr(model, 'module') else model # Take care of distributed/parallel training
model_to_save.save_pretrained(args.output_dir)
tokenizer.save_pretrained(args.output_dir)
# Good practice: save your training arguments together with the trained model
torch.save(args, os.path.join(args.output_dir, 'training_args.bin'))
# Load a trained model and vocabulary that you have fine-tuned
model = model_class.from_pretrained(args.output_dir)
tokenizer = tokenizer_class.from_pretrained(args.output_dir, do_lower_case=args.do_lower_case)
model.to(args.device)
# Evaluation
results = {}
if args.do_eval and args.local_rank in [-1, 0]:
checkpoints = [args.output_dir]
if args.eval_all_checkpoints:
checkpoints = list(os.path.dirname(c) for c in sorted(glob.glob(args.output_dir + '/**/' + WEIGHTS_NAME, recursive=True)))
logging.getLogger("transformers.modeling_utils").setLevel(logging.WARN) # Reduce logging
logger.info("Evaluate the following checkpoints: %s", checkpoints)
for checkpoint in checkpoints:
global_step = checkpoint.split('-')[-1] if len(checkpoints) > 1 else ""
prefix = checkpoint.split('/')[-1] if checkpoint.find('checkpoint') != -1 else ""
model = model_class.from_pretrained(checkpoint)
model.to(args.device)
result = evaluate(args, model, tokenizer, prefix=prefix)
result = dict((k + '_{}'.format(global_step), v) for k, v in result.items())
results.update(result)
return results
if __name__ == "__main__":
main()

37
baseline/train_baseline.sh
Unescape View File

@ -0,0 +1,37 @@
#!/bin/bash
# lr 1e-5 to 5e-5
# mc_loss_efficient 0.1 to 1
while getopts d: flag
do
case "${flag}" in
d) data_split=${OPTARG};;
esac
done
# Check whether the required environment vars are set
if [[ -z "${SAVED_MODELS_BASELINE}" ]] || [[ -z "${PRE_TRAINED_SOLOIST}" ]]; then
echo "Required Environment variables not set. First run set_env_var.sh"
fi
datetime_now=$(date +"%Y%m%d")
experiment_folder=experiment-${datetime_now}/"${data_split}"
python soloist_train.py \
--output_dir="${SAVED_MODELS_BASELINE}"/"${experiment_folder}" \
--model_type=gpt2 \
--model_name_or_path="${PRE_TRAINED_SOLOIST}" \
--do_train \
--train_data_file=../data/baseline/"${data_split}"/train.soloist.json \
--eval_data_file=../data/baseline/valid/valid.soloist.json \
--add_special_action_tokens=../data/resource/special_tokens.txt \
--per_gpu_train_batch_size 1 \
--num_train_epochs 25 \
--learning_rate 5e-5 \
--overwrite_cache \
--save_steps 5000 \
--max_seq 100 \
--overwrite_output_dir \
--max_turn 15 \
--num_candidates 1 \
--mc_loss_efficient 0.33 \
--add_belief_prediction

0
common/__init__.py
Unescape View File

13
set_env.sh
Unescape View File

@ -0,0 +1,13 @@
#!/bin/bash
# Path for storing baseline saved models
export SAVED_MODELS_BASELINE=/mount/studenten/projects/mandavsi/baseline/trained_models
# path for storing test outputs
export OUTPUTS_DIR_BASELINE=/mount/studenten/projects/mandavsi/baseline/outputs
# path of pretrained SOLOIST model
export PRE_TRAINED_SOLOIST=/mount/studenten/projects/mandavsi/soloist-pretrained/gtg_pretrained
# create dirs if not exist
mkdir -p ${SAVED_MODELS_BASELINE}
mkdir -p ${OUTPUTS_DIR_BASELINE}
mkdir -p ${PRE_TRAINED_SOLOIST}

5
unset_env.sh
Unescape View File

@ -0,0 +1,5 @@
#!/bin/bash
# run this script to unset env variables
unset SAVED_MODELS_BASELINE
unset PRE_TRAINED_SOLOIST
unset OUTPUTS_DIR_BASELINE
Write Preview
Loading…
Cancel
Save