-
Notifications
You must be signed in to change notification settings - Fork 10
/
finetune.py
324 lines (286 loc) · 12.4 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
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
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
# this code is modified from lora_alpaca https://github.com/tloen/alpaca-lora under Apache-2.0 license
import os
from typing import List
import fire
import torch
import transformers
from datasets import load_dataset
from transformers import BertTokenizerFast
"""
Unused imports:
import torch.nn as nn
import bitsandbytes as bnb
"""
from peft import (
LoraConfig,
get_peft_model,
get_peft_model_state_dict,
prepare_model_for_int8_training,
set_peft_model_state_dict,
)
from transformers import LlamaForCausalLM, LlamaTokenizer
from utils.prompter import Prompter
def train(
# model/data params
base_model: str = "", # the only required argument
data_path: str = "",
output_dir: str = "",
# training hyperparams
batch_size: int = 128,
micro_batch_size: int = 4,
num_epochs: int = 3,
learning_rate: float = 3e-4,
cutoff_len: int = 256,
val_set_size: int = 2000,
# lora hyperparams
lora_r: int = 8,
lora_alpha: int = 16,
lora_dropout: float = 0.05,
lora_target_modules: List[str] = [
"q_proj",
"v_proj",
],
# llm hyperparams
train_on_inputs: bool = True, # if False, masks out inputs in loss
add_eos_token: bool = False,
group_by_length: bool = False, # faster, but produces an odd training loss curve
# wandb params
wandb_project: str = "gama",
wandb_run_name: str = "",
wandb_watch: str = "false", # options: false | gradients | all
wandb_log_model: str = "false", # options: false | true
resume_from_checkpoint: str = None, # either training checkpoint or final adapter
prompt_template_name: str = "alpaca_short", # The prompt template to use, will default to alpaca.
save_steps: int = 100,
trainable_params = 'all'
):
if int(os.environ.get("LOCAL_RANK", 0)) == 0:
print(
f"Training Alpaca-LoRA model with params:\n"
f"base_model: {base_model}\n"
f"data_path: {data_path}\n"
f"output_dir: {output_dir}\n"
f"batch_size: {batch_size}\n"
f"micro_batch_size: {micro_batch_size}\n"
f"num_epochs: {num_epochs}\n"
f"learning_rate: {learning_rate}\n"
f"cutoff_len: {cutoff_len}\n"
f"val_set_size: {val_set_size}\n"
f"lora_r: {lora_r}\n"
f"lora_alpha: {lora_alpha}\n"
f"lora_dropout: {lora_dropout}\n"
f"lora_target_modules: {lora_target_modules}\n"
f"train_on_inputs: {train_on_inputs}\n"
f"add_eos_token: {add_eos_token}\n"
f"group_by_length: {group_by_length}\n"
f"wandb_project: {wandb_project}\n"
f"wandb_run_name: {wandb_run_name}\n"
f"wandb_watch: {wandb_watch}\n"
f"wandb_log_model: {wandb_log_model}\n"
f"resume_from_checkpoint: {resume_from_checkpoint or False}\n"
f"prompt template: {prompt_template_name}\n"
)
assert (
base_model
), "Please specify a --base_model, e.g. --base_model='huggyllama/llama-7b'"
# trick to load checkpoints correctly from HF
if '/fs/nexus-projects/brain_project/acl_sk_24/GAMA/src/Llama-2-7b-chat-hf-qformer' not in base_model:
# start from a different model with original vicuna
# temporally first load the original vicuna, then load the actual checkpoint
start_model = base_model # need to point to a specific bin file that contains state dict.
# TODO: change to your vicuna_tltr path
base_model = '/fs/nexus-projects/brain_project/acl_sk_24/GAMA/src/Llama-2-7b-chat-hf-qformer'
print('Will load from {:s} later, for implementation purpose, first load from {:s}'.format(start_model, base_model))
else:
start_model = None
gradient_accumulation_steps = batch_size // micro_batch_size
prompter = Prompter(prompt_template_name)
device_map = "auto"
world_size = int(torch.cuda.device_count())
ddp = world_size != 1
if ddp:
device_map = {"": int(os.environ.get("LOCAL_RANK") or 0)}
gradient_accumulation_steps = gradient_accumulation_steps // world_size
use_wandb = len(wandb_project) > 0 or (
"WANDB_PROJECT" in os.environ and len(os.environ["WANDB_PROJECT"]) > 0
)
# Only overwrite environ if wandb param passed
if len(wandb_project) > 0:
os.environ["WANDB_PROJECT"] = wandb_project
if len(wandb_watch) > 0:
os.environ["WANDB_WATCH"] = wandb_watch
if len(wandb_log_model) > 0:
os.environ["WANDB_LOG_MODEL"] = wandb_log_model
# base_model = '/fs/nexus-projects/brain_project/acl_sk_24/GAMA/src/Llama-2-7b-chat-hf-qformer'
model = LlamaForCausalLM.from_pretrained(
base_model,
load_in_8bit=False,
# torch_dtype=torch.float16,
device_map=device_map,
)
tokenizer = LlamaTokenizer.from_pretrained(base_model)
tokenizer.pad_token = tokenizer.eos_token
tokenizer.padding_side = "left" # Allow batched inference
bert_tokenizer = BertTokenizerFast.from_pretrained("google-bert/bert-base-uncased")
def tokenize(prompt, add_eos_token=True):
result = tokenizer(
prompt,
truncation=True,
max_length=cutoff_len,
padding=False,
return_tensors=None,
)
if (
result["input_ids"][-1] != tokenizer.eos_token_id
and len(result["input_ids"]) < 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 flatten_c(example):
if 'tokenized_full_prompt' in example:
example.update(example['tokenized_full_prompt']) # Merge 'c' into the root
del example['tokenized_full_prompt'] # Remove 'c' from the example
return example
def generate_and_tokenize_prompt(data_point):
full_prompt = prompter.generate_prompt(
data_point["instruction"],
data_point["input"],
data_point["output"]
)
tokenized_full_prompt = tokenize(full_prompt)
if not train_on_inputs:
user_prompt = prompter.generate_prompt(
data_point["instruction"], data_point["input"]
)
tokenized_user_prompt = tokenize(
user_prompt, add_eos_token=add_eos_token
)
user_prompt_len = len(tokenized_user_prompt["input_ids"])
if 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
tokenizer_input_bert = []
# print(tokenized_full_prompt)
return tokenized_full_prompt
# return {'tokenized_full_prompt': tokenized_full_prompt, 'tokenizer_input_bert':tokenizer_input_bert}
config = LoraConfig(
r=lora_r,
lora_alpha=lora_alpha,
target_modules=lora_target_modules,
lora_dropout=lora_dropout,
bias="none",
task_type="CAUSAL_LM",
)
model = get_peft_model(model, config)
# print(model)
# for audio params, lora always trainable, llama always frozen
for name, param in model.named_parameters():
if trainable_params == 'all':
if "audio" in name:
param.requires_grad = True
if trainable_params == 'proj':
if "audio_proj" in name:
param.requires_grad = True
if trainable_params == 'qformer':
if "audio_aggregator_layer_1" in name or "audio_aggregator_layer_2" in name or "audio_proj_qformer" in name or "audio_proj_audioenc" in name or "audio_proj_norm_qformer" in name or "audio_proj_norm_audioenc" in name:
param.requires_grad = True
if trainable_params == 'qformer_all':
if "audio_aggregator_layer_1" in name or "audio_aggregator_layer_2" in name or "audio_proj_qformer" in name or "audio_proj_audioenc" in name or "audio_proj_norm_qformer" in name or "audio_proj_norm_audioenc" in name or 'audio_encoder' in name or 'Qformer' in name or 'query_tokens' in name or 'qformer_proj_norm' in name:
param.requires_grad = True
if data_path.endswith(".json") or data_path.endswith(".jsonl"):
data = load_dataset("json", data_files=data_path)
else:
data = load_dataset(data_path)
if resume_from_checkpoint:
# Check the available weights and load them
checkpoint_name = os.path.join(
resume_from_checkpoint, "pytorch_model.bin"
) # Full checkpoint
if not os.path.exists(checkpoint_name):
checkpoint_name = os.path.join(
resume_from_checkpoint, "adapter_model.bin"
) # only LoRA model - LoRA config above has to fit
resume_from_checkpoint = (
False # So the trainer won't try loading its state
)
# The two files above have a different name depending on how they were saved, but are actually the same.
if os.path.exists(checkpoint_name):
state_dict = torch.load(checkpoint_name, map_location='cpu')
msg = model.load_state_dict(state_dict, strict=False)
else:
print(f"Checkpoint {checkpoint_name} not found")
# # now load from real checkpoint
if start_model != None and (resume_from_checkpoint == None or resume_from_checkpoint == False):
state_dict = torch.load(start_model, map_location='cpu')
msg = model.load_state_dict(state_dict, strict=False)
# print('load checkpoint', msg)
model.print_trainable_parameters() # Be more transparent about the % of trainable params.
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
# train_data = train_data.map(flatten_c)
if not ddp and torch.cuda.device_count() > 1:
# keeps Trainer from trying its own DataParallelism when more than 1 gpu is available
model.is_parallelizable = True
model.model_parallel = True
from transformers import TrainerCallback
class PrecisionLoggingCallback(TrainerCallback):
def on_log(self, args, state, control, logs=None, **kwargs):
# Modify this method to log the loss with more decimal points
if logs is not None and 'loss' in logs:
# Assuming 'logs' is a dictionary that contains the loss
high_precision_loss = format(logs['loss'], '.10f') # Adjust the '.4f' for more or fewer decimals
# print(f"High Precision Loss: {high_precision_loss}")
trainer = transformers.Trainer(
model=model,
train_dataset=train_data,
eval_dataset=val_data,
callbacks=[PrecisionLoggingCallback],
args=transformers.TrainingArguments(
per_device_train_batch_size=micro_batch_size,
gradient_accumulation_steps=gradient_accumulation_steps,
warmup_steps=100,
num_train_epochs=num_epochs,
learning_rate=learning_rate,
bf16=True,
logging_steps=10,
optim="adamw_torch",
evaluation_strategy="no",
save_strategy="steps",
eval_steps=None,
save_steps=save_steps,
dataloader_num_workers=8,
output_dir=output_dir,
save_total_limit=50,
load_best_model_at_end=False,
ddp_find_unused_parameters=True,
group_by_length=group_by_length,
report_to="wandb" if use_wandb else None,
run_name=wandb_run_name if use_wandb else None,
remove_unused_columns=False ),
data_collator=transformers.DataCollatorForSeq2Seq(
tokenizer, pad_to_multiple_of=8, return_tensors="pt", padding=True
),
)
model.config.use_cache = False
trainer.train(resume_from_checkpoint=resume_from_checkpoint)
model.save_pretrained(output_dir)
if __name__ == "__main__":
fire.Fire(train)