-
Notifications
You must be signed in to change notification settings - Fork 18
/
finetune.py
234 lines (209 loc) · 7.59 KB
/
finetune.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
import os
import sys
import argparse
import torch
import torch.nn as nn
import bitsandbytes as bnb
from datasets import load_dataset
import transformers
from peft import PeftModel
from transformers import LlamaForCausalLM, LlamaTokenizer
from peft import (
prepare_model_for_int8_training,
LoraConfig,
get_peft_model,
get_peft_model_state_dict,
)
def generate_prompt(data_point):
# sorry about the formatting disaster gotta move fast
if data_point["input"]:
return f"""下方是一個關於任務的指令,以及一個提供與任務相關之資訊的輸入。請撰寫一個能適當地完成該任務指令需求的回覆。
### 指令:
{data_point["instruction"]}
### 輸入:
{data_point["input"]}
### 回覆:
{data_point["output"]}"""
else:
return f"""下方是一個關於任務的指令。請撰寫一個能適當地完成該任務指令需求的回覆。
### 輸入:
{data_point["instruction"]}
### 回覆:
{data_point["output"]}"""
def tokenize(prompt):
# there's probably a way to do this with the tokenizer settings
# but again, gotta move fast
result = tokenizer(
prompt,
truncation=True,
max_length=CUTOFF_LEN + 1,
padding="max_length",
)
return {
"input_ids": result["input_ids"][:-1],
"attention_mask": result["attention_mask"][:-1],
}
def generate_and_tokenize_prompt(data_point):
# This function masks out the labels for the input,
# so that our loss is computed only on the response.
user_prompt = (
(
f"""下方是一個關於任務的指令,以及一個提供與任務相關之資訊的輸入。請撰寫一個能適當地完成該任務需求的回覆。
### 指令:
{data_point["instruction"]}
### 輸入:
{data_point["input"]}
### 回覆:
"""
)
if data_point["input"]
else (
f"""下方是一個關於任務的指令。請撰寫一個能適當地完成該任務需求的回覆。
### 輸入:
{data_point["instruction"]}
### 回覆:
"""
)
)
len_user_prompt_tokens = (
len(
tokenizer(
user_prompt,
truncation=True,
max_length=CUTOFF_LEN + 1,
padding="max_length",
)["input_ids"]
)
- 1
) # no eos token
full_tokens = tokenizer(
user_prompt + data_point["output"],
truncation=True,
max_length=CUTOFF_LEN + 1,
padding="max_length",
)["input_ids"][:-1]
return {
"input_ids": full_tokens,
"labels": [-100] * len_user_prompt_tokens
+ full_tokens[len_user_prompt_tokens:],
"attention_mask": [1] * (len(full_tokens)),
}
def parse_config():
parser = argparse.ArgumentParser()
parser.add_argument("--model_name", type=str, default="decapoda-research/llama-7b-hf")
parser.add_argument("--from_ckpt", action="store_true")
parser.add_argument("--ckpt_name", type=str)
parser.add_argument("--dataset_dir", type=str, required=True)
parser.add_argument("--output_dir", type=str, required=True)
parser.add_argument("--cache_dir", type=str, default="../cache")
parser.add_argument("--num_epoch", type=int, default=3)
parser.add_argument("--logging_steps", type=int, default=20)
parser.add_argument("--save_steps", type=int, default=600)
parser.add_argument("--save_total_limit", type=int, default=3)
parser.add_argument("--report_to", type=str, default="wandb")
parser.add_argument("--wandb_run_name", type=str, required=True)
return parser.parse_args()
if __name__ == '__main__':
args = parse_config()
if args.report_to == 'wandb':
import wandb
wandb.login()
# Probably no need to change the following settings
# -------------------------------------------------------------------------
# optimized for RTX 4090. for larger GPUs, increase some of these?
MICRO_BATCH_SIZE = 4 # this could actually be 5 but i like powers of 2
BATCH_SIZE = 128
GRADIENT_ACCUMULATION_STEPS = BATCH_SIZE // MICRO_BATCH_SIZE
# EPOCHS = 6 # we don't always need 3 tbh
LEARNING_RATE = 3e-4 # the Karpathy constant
CUTOFF_LEN = 256 # 256 accounts for about 96% of the data
LORA_R = 8
LORA_ALPHA = 16
LORA_DROPOUT = 0.05
VAL_SET_SIZE = 0
TARGET_MODULES = [
"q_proj",
"v_proj",
]
device_map = "auto"
world_size = int(os.environ.get("WORLD_SIZE", 1))
ddp = world_size != 1
if ddp:
device_map = {"": int(os.environ.get("LOCAL_RANK") or 0)}
GRADIENT_ACCUMULATION_STEPS = GRADIENT_ACCUMULATION_STEPS // world_size
# -------------------------------------------------------------------------
# main logic starts here
model = LlamaForCausalLM.from_pretrained(
args.model_name,
load_in_8bit=True,
device_map=device_map,
cache_dir=args.cache_dir
)
tokenizer = LlamaTokenizer.from_pretrained(
args.model_name,
add_eos_token=True,
cache_dir=args.cache_dir
)
if args.from_ckpt:
# Resume from checkpoint
model = PeftModel.from_pretrained(model, args.ckpt_name)
model = prepare_model_for_int8_training(model)
config = LoraConfig(
r=LORA_R,
lora_alpha=LORA_ALPHA,
target_modules=TARGET_MODULES,
lora_dropout=LORA_DROPOUT,
bias="none",
task_type="CAUSAL_LM",
)
model = get_peft_model(model, config)
tokenizer.pad_token_id = 0 # unk. we want this to be different from the eos token
data = load_dataset(
"json",
data_files=args.dataset_dir,
cache_dir=args.cache_dir
)
if VAL_SET_SIZE > 0:
train_val = data["train"].train_test_split(
test_size=VAL_SET_SIZE, shuffle=True, seed=42
)
train_data = train_val["train"].shuffle().map(generate_and_tokenize_prompt)
val_data = train_val["test"].shuffle().map(generate_and_tokenize_prompt)
else:
train_data = data['train'].shuffle().map(generate_and_tokenize_prompt)
val_data = None
trainer = transformers.Trainer(
model=model,
train_dataset=train_data,
eval_dataset=val_data,
args=transformers.TrainingArguments(
per_device_train_batch_size=MICRO_BATCH_SIZE,
gradient_accumulation_steps=GRADIENT_ACCUMULATION_STEPS,
warmup_steps=100,
num_train_epochs=args.num_epoch,
learning_rate=LEARNING_RATE,
fp16=True,
logging_steps=args.logging_steps,
# evaluation_strategy="steps" if VAL_SET_SIZE > 0 else "no",
save_strategy="steps",
# eval_steps=200 if VAL_SET_SIZE > 0 else None,
save_steps=args.save_steps,
output_dir=args.output_dir,
save_total_limit=args.save_total_limit,
# load_best_model_at_end=True if VAL_SET_SIZE > 0 else False,
ddp_find_unused_parameters=False if ddp else None,
report_to=args.report_to,
run_name=args.wandb_run_name if args.report_to == 'wandb' else None,
),
data_collator=transformers.DataCollatorForLanguageModeling(tokenizer, mlm=False),
)
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))
if torch.__version__ >= "2" and sys.platform != 'win32':
model = torch.compile(model)
trainer.train()
model.save_pretrained(args.output_dir)
print("\n If there's a warning about missing keys above, please disregard :)")