-
Notifications
You must be signed in to change notification settings - Fork 87
/
post_training.py
262 lines (225 loc) · 10 KB
/
post_training.py
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
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
'''
Refer to
https://github.com/tloen/alpaca-lora/blob/main/finetune.py
'''
import os
import sys
import argparse
from typing import List
from pathlib import Path
import torch
import transformers
from datasets import load_dataset
from LLMPruner.peft import (
LoraConfig,
get_peft_model,
get_peft_model_state_dict,
prepare_model_for_int8_training,
set_peft_model_state_dict,
)
from LLMPruner.utils.prompter import Prompter, ZeroPrompter
from LLMPruner.datasets.ppl_dataset import get_loaders
device = "cuda" if torch.cuda.is_available() else "cpu"
def main(args):
# Set WanDB
os.environ["WANDB_PROJECT"] = args.wandb_project
# Load Pruned Model
pruned_dict = torch.load(args.prune_model, map_location='cpu')
tokenizer, model = pruned_dict['tokenizer'], pruned_dict['model']
gradient_accumulation_steps = args.batch_size // args.micro_batch_size
if not args.no_instruction:
prompter = Prompter(args.prompt_template_name)
else:
prompter = ZeroPrompter()
world_size = int(os.environ.get("WORLD_SIZE", 1))
ddp = world_size != 1
if ddp:
gradient_accumulation_steps = gradient_accumulation_steps // world_size
if device == 'cuda':
model.half()
tokenizer.pad_token_id = 0
tokenizer.padding_side = "left"
def tokenize(prompt, add_eos_token=True):
result = tokenizer(
prompt,
truncation=True,
max_length=args.cutoff_len,
padding=False,
return_tensors=None,
)
if (
result["input_ids"][-1] != tokenizer.eos_token_id
and len(result["input_ids"]) < args.cutoff_len
and add_eos_token
):
result["input_ids"].append(tokenizer.eos_token_id)
result["attention_mask"].append(1)
result["labels"] = result["input_ids"].copy()
return result
def generate_and_tokenize_prompt(data_point):
if 'lamini' in args.data_path.lower():
full_prompt = prompter.generate_prompt(
data_point["instruction"],
None,
data_point["response"],
)
elif 'alpaca' in args.data_path.lower():
full_prompt = prompter.generate_prompt(
data_point["instruction"],
data_point["input"],
data_point["output"],
)
else:
raise NotImplementedError
tokenized_full_prompt = tokenize(full_prompt)
if not args.train_on_inputs:
user_prompt = prompter.generate_prompt(
data_point["instruction"], data_point["input"] if 'input' in data_point.keys() else None,
)
tokenized_user_prompt = tokenize(
user_prompt, add_eos_token=args.add_eos_token
)
user_prompt_len = len(tokenized_user_prompt["input_ids"])
if args.add_eos_token:
user_prompt_len -= 1
tokenized_full_prompt["labels"] = [
-100
] * user_prompt_len + tokenized_full_prompt["labels"][
user_prompt_len:
] # could be sped up, probably
return tokenized_full_prompt
def split_and_tokenizer(test_data, tokenizer, seq_len, field_name):
test_ids = tokenizer("\n\n".join(test_data[field_name]), return_tensors='pt').input_ids[0]
nsamples = test_ids.numel() // seq_len
test_set = []
for i in range(nsamples):
batch = test_ids[(i * seq_len):((i + 1) * seq_len)]
test_set.append({
'input_ids': batch,
'labels': batch
})
return test_set
# Prepare For LoRA
model = prepare_model_for_int8_training(model)
config = LoraConfig(
r=args.lora_r,
lora_alpha=args.lora_alpha,
target_modules=args.lora_target_modules.split(","),
lora_dropout=args.lora_dropout,
bias="none",
task_type="CAUSAL_LM",
)
model = get_peft_model(model, config)
model.print_trainable_parameters()
# Load Train Dataset
data = load_dataset(args.data_path)
if args.cache_dataset and os.path.exists('datasets/cache/{}.bin'.format(args.data_path)):
preprocess_data = torch.load('datasets/cache/{}.bin'.format(args.data_path))
train_data, val_data = preprocess_data['train'], preprocess_data['val']
else:
train_val = data["train"].train_test_split(
test_size=args.val_set_size, shuffle=True, seed=42
)
train_data = (
train_val["train"].shuffle().map(generate_and_tokenize_prompt)
)
val_data = {
args.data_path: train_val["test"].shuffle().map(generate_and_tokenize_prompt),
}
if args.cache_dataset and args.local_rank == 0:
cache_file = 'datasets/cache/{}.bin'.format(args.data_path)
cache_dir = '/'.join(cache_file.split('/')[:-1])
directory = Path(cache_dir)
directory.mkdir(parents=True, exist_ok=True)
torch.save({
'train': train_data, 'val': val_data
}, cache_file)
# Load Extra Validation Dataset
if args.extra_val_dataset:
from LLMPruner.datasets.ppl_dataset import get_wikitext2, get_ptb
seq_len = 128
for extra_dataset in args.extra_val_dataset.split(','):
if 'wikitext2' in extra_dataset:
_, test_data = get_wikitext2(seq_len, None)
test_data = split_and_tokenizer(test_data, tokenizer, seq_len, field_name='text')
if 'ptb' in extra_dataset:
_, test_data = get_ptb(seq_len, None)
test_data = split_and_tokenizer(test_data, tokenizer, seq_len, field_name='sentence')
val_data[extra_dataset] = test_data
trainer = transformers.Trainer(
model=model,
train_dataset=train_data,
eval_dataset=val_data,
args=transformers.TrainingArguments(
per_device_train_batch_size=args.micro_batch_size,
gradient_accumulation_steps=gradient_accumulation_steps,
warmup_steps=100,
num_train_epochs=args.num_epochs,
learning_rate=args.learning_rate,
fp16=True,
logging_steps=10,
logging_first_step=True,
optim="adamw_torch",
evaluation_strategy="steps",
save_strategy="steps",
eval_steps=100,
save_steps=200,
output_dir=args.output_dir,
save_total_limit=20,
load_best_model_at_end=True,
ddp_find_unused_parameters=None,
group_by_length=args.group_by_length,
report_to="wandb",
run_name=args.output_dir.split('/')[-1],
metric_for_best_model="{}_loss".format(args.data_path),
),
data_collator=transformers.DataCollatorForSeq2Seq(
tokenizer, pad_to_multiple_of=8, return_tensors="pt", padding=True
),
)
model.config.use_cache = False
old_state_dict = model.state_dict
model.state_dict = (
lambda self, *_, **__: get_peft_model_state_dict(
self, old_state_dict()
)
).__get__(model, type(model))
trainer.train(resume_from_checkpoint=args.resume_from_checkpoint)
model.state_dict = old_state_dict
model.save_pretrained(args.output_dir)
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='Tuning Pruned LLM')
# Model Type&Path
parser.add_argument('--base_model', type=str, default="decapoda-research/llama-7b-hf", help='base model name')
parser.add_argument('--prune_model', type=str, help='prune model name')
parser.add_argument('--data_path', type=str, default="yahma/alpaca-cleaned", help='data path')
parser.add_argument('--cache_dataset', action="store_true", default=False)
parser.add_argument('--extra_val_dataset', type=str, default=None, help='validation datasets. Split with ","')
parser.add_argument('--output_dir', type=str, default="./lora-alpaca", help='output directory')
# Training Hyperparameters
parser.add_argument('--batch_size', type=int, default=128, help='batch size')
parser.add_argument('--micro_batch_size', type=int, default=4, help='micro batch size')
parser.add_argument('--num_epochs', type=int, default=5, help='number of epochs')
parser.add_argument('--learning_rate', type=float, default=3e-4, help='learning rate')
parser.add_argument('--cutoff_len', type=int, default=256, help='cutoff length')
parser.add_argument('--val_set_size', type=int, default=2000, help='validation set size')
parser.add_argument('--prompt_template_name', type=str, default="alpaca", help="The prompt template to use, will default to alpaca.")
parser.add_argument('--no_instruction', action='store_true', default=False, help="Whether to use the instruction template or not.")
# Lora Configuration
parser.add_argument('--lora_r', type=int, default=8, help='lora r')
parser.add_argument('--lora_alpha', type=int, default=16, help='lora alpha')
parser.add_argument('--lora_dropout', type=float, default=0.05, help='lora dropout')
parser.add_argument('--lora_target_modules', type=str, default="q_proj,k_proj,v_proj,o_proj,gate_proj,down_proj,up_proj", help='lora target modules')
# llm hyperparameters
parser.add_argument('--train_on_inputs', default=False, action="store_true", help='Train on inputs. If False, masks out inputs in loss')
parser.add_argument('--add_eos_token', default=False, action="store_true")
parser.add_argument('--group_by_length', default=False, action="store_true", help="faster, but produces an odd training loss curve")
# wandb params
parser.add_argument('--wandb_project', type=str, default="")
parser.add_argument('--resume_from_checkpoint', type=str, help="either training checkpoint or final adapter")
#ddp
parser.add_argument('--local_rank', type=int, default=-1)
args = parser.parse_args()
torch_version = int(torch.__version__.split('.')[1])
args.torch_version = torch_version
main(args)