mirror of
https://github.com/modelscope/modelscope.git
synced 2025-12-24 03:59:23 +01:00
Added full parameter sft to llm (#402)
* Optimized code * update parse_args * fix get_logger bug * update parse_args * Added full parameter fine-tuning * Add support_bf16 warning * Modify the code format and fix bugs
This commit is contained in:
Jintao
@@ -1,20 +1,25 @@
|
||||
# ### Setting up experimental environment.
|
||||
"""
|
||||
pip install modelscope
|
||||
pip install numpy pandas matplotlib scikit-learn
|
||||
pip install transformers datasets
|
||||
conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia
|
||||
pip install tqdm tensorboard torchmetrics sentencepiece charset_normalizer
|
||||
pip install accelerate transformers_stream_generator
|
||||
|
||||
pip install numpy -U # Resolve torchmetrics dependencies and update numpy
|
||||
# Install the latest version of modelscope from source
|
||||
git clone https://github.com/modelscope/modelscope.git
|
||||
cd modelscope
|
||||
pip install .
|
||||
|
||||
# Resolve torchmetrics dependencies and update numpy
|
||||
pip install numpy -U
|
||||
"""
|
||||
|
||||
from _common import *
|
||||
|
||||
|
||||
@dataclass
|
||||
class Arguments:
|
||||
class SftArguments:
|
||||
device: str = '0,1' # e.g. '-1'; '0'; '0,1'
|
||||
seed: int = 42
|
||||
model_type: str = field(
|
||||
@@ -23,29 +28,52 @@ class Arguments:
|
||||
'choices':
|
||||
['baichuan-7b', 'baichuan-13b', 'chatglm2', 'llama2-7b']
|
||||
})
|
||||
# baichuan-7b: 'lora': 16G; 'full': 80G
|
||||
sft_type: str = field(
|
||||
default='lora', metadata={'choices': ['lora', 'full']})
|
||||
data_sample: Optional[int] = None
|
||||
#
|
||||
|
||||
lora_target_modules: Optional[List[str]] = None
|
||||
lora_rank: int = 8
|
||||
lora_alpha: int = 32
|
||||
lora_dropout_p: float = 0.1
|
||||
#
|
||||
|
||||
gradient_checkpoint: bool = True
|
||||
batch_size: int = 1
|
||||
max_epochs: int = 1
|
||||
eval_interval: int = 500
|
||||
learning_rate: float = 1e-4
|
||||
learning_rate: Optional[float] = None
|
||||
weight_decay: float = 0.01
|
||||
n_accumulate_grad: int = 16
|
||||
grad_clip_norm: float = 1.
|
||||
warmup_iters: int = 200
|
||||
|
||||
save_trainer_state: Optional[bool] = None
|
||||
eval_interval: int = 500
|
||||
last_save_interval: Optional[int] = None
|
||||
last_max_checkpoint_num: int = 1
|
||||
best_max_checkpoint_num: int = 1
|
||||
#
|
||||
logging_interval: int = 5
|
||||
tb_interval: int = 5
|
||||
|
||||
def __post_init__(self):
|
||||
if self.sft_type == 'lora':
|
||||
if self.learning_rate is None:
|
||||
self.learning_rate = 1e-4
|
||||
if self.save_trainer_state is None:
|
||||
self.save_trainer_state = True
|
||||
if self.last_save_interval is None:
|
||||
self.last_save_interval = self.eval_interval
|
||||
elif self.sft_type == 'full':
|
||||
if self.learning_rate is None:
|
||||
self.learning_rate = 1e-5
|
||||
if self.save_trainer_state is None:
|
||||
self.save_trainer_state = False # save disk space
|
||||
if self.last_save_interval is None:
|
||||
# Saving the model takes a long time
|
||||
self.last_save_interval = self.eval_interval * 4
|
||||
else:
|
||||
raise ValueError(f'sft_type: {self.sft_type}')
|
||||
|
||||
if self.lora_target_modules is None:
|
||||
if self.model_type in {'baichuan-7b', 'baichuan-13b'}:
|
||||
self.lora_target_modules = ['W_pack']
|
||||
@@ -57,181 +85,179 @@ class Arguments:
|
||||
raise ValueError(f'model_type: {self.model_type}')
|
||||
|
||||
|
||||
def parse_args() -> Arguments:
|
||||
args, = HfArgumentParser([Arguments]).parse_args_into_dataclasses()
|
||||
def parse_args() -> SftArguments:
|
||||
# return_remaining_strings=True for notebook compatibility
|
||||
args, remaining_args = HfArgumentParser([
|
||||
SftArguments
|
||||
]).parse_args_into_dataclasses(return_remaining_strings=True)
|
||||
logger.info(f'args: {args}')
|
||||
if len(remaining_args) > 0:
|
||||
logger.warning(f'remaining_args: {remaining_args}')
|
||||
return args
|
||||
|
||||
|
||||
args = parse_args()
|
||||
logger.info(args)
|
||||
select_device(args.device)
|
||||
seed_everything(args.seed)
|
||||
def llm_sft(args: SftArguments) -> None:
|
||||
select_device(args.device)
|
||||
seed_everything(args.seed)
|
||||
|
||||
# ### Loading Model and Tokenizer
|
||||
if args.model_type == 'baichuan-7b':
|
||||
model_dir = snapshot_download('baichuan-inc/baichuan-7B', 'v1.0.5')
|
||||
model, tokenizer = get_baichuan_model_tokenizer(model_dir)
|
||||
elif args.model_type == 'baichuan-13b':
|
||||
model_dir = snapshot_download('baichuan-inc/Baichuan-13B-Base', 'v1.0.2')
|
||||
model, tokenizer = get_baichuan_model_tokenizer(model_dir)
|
||||
elif args.model_type == 'chatglm2':
|
||||
model_dir = snapshot_download('ZhipuAI/chatglm2-6b', 'v1.0.6')
|
||||
model, tokenizer = get_chatglm2_model_tokenizer(model_dir)
|
||||
elif args.model_type == 'llama2-7b':
|
||||
model_dir = snapshot_download('modelscope/Llama-2-7b-ms', 'v1.0.0')
|
||||
model, tokenizer = get_llama2_model_tokenizer(model_dir)
|
||||
else:
|
||||
raise ValueError(f'model_type: {args.model_type}')
|
||||
# ### Loading Model and Tokenizer
|
||||
support_bf16 = torch.cuda.is_bf16_supported()
|
||||
if not support_bf16:
|
||||
logger.warning(f'support_bf16: {support_bf16}')
|
||||
model, tokenizer, model_dir = get_model_tokenizer(
|
||||
args.model_type, torch_dtype=torch.bfloat16)
|
||||
|
||||
#
|
||||
if args.gradient_checkpoint:
|
||||
# baichuan13B does not implement the `get_input_embeddings` function
|
||||
if args.model_type == 'baichuan-13b':
|
||||
if args.gradient_checkpoint:
|
||||
# baichuan-13b does not implement the `get_input_embeddings` function
|
||||
if args.model_type == 'baichuan-13b':
|
||||
model.__class__.get_input_embeddings = MethodType(
|
||||
lambda self: self.model.embed_tokens, model)
|
||||
model.gradient_checkpointing_enable()
|
||||
model.enable_input_require_grads()
|
||||
|
||||
def get_input_embeddings(self):
|
||||
return self.model.embed_tokens
|
||||
# ### Preparing lora
|
||||
if args.sft_type == 'lora':
|
||||
lora_config = LoRAConfig(
|
||||
replace_modules=args.lora_target_modules,
|
||||
rank=args.lora_rank,
|
||||
lora_alpha=args.lora_alpha,
|
||||
lora_dropout=args.lora_dropout_p)
|
||||
logger.info(f'lora_config: {lora_config}')
|
||||
Swift.prepare_model(model, lora_config)
|
||||
|
||||
model.__class__.get_input_embeddings = get_input_embeddings.__get__(
|
||||
model)
|
||||
model.gradient_checkpointing_enable()
|
||||
model.enable_input_require_grads()
|
||||
show_freeze_layers(model)
|
||||
print_model_info(model)
|
||||
# check the device and dtype of the model
|
||||
_p: Parameter = list(model.parameters())[-1]
|
||||
logger.info(f'device: {_p.device}, dtype: {_p.dtype}')
|
||||
|
||||
# ### Preparing lora
|
||||
lora_config = LoRAConfig(
|
||||
replace_modules=args.lora_target_modules,
|
||||
rank=args.lora_rank,
|
||||
lora_alpha=args.lora_alpha,
|
||||
lora_dropout=args.lora_dropout_p)
|
||||
logger.info(f'lora_config: {lora_config}')
|
||||
Swift.prepare_model(model, lora_config)
|
||||
#
|
||||
show_freeze_layers(model)
|
||||
print_model_info(model)
|
||||
_p: Parameter = list(model.parameters())[100]
|
||||
logger.info(f'device: {_p.device}, dtype: {_p.dtype}')
|
||||
model.bfloat16()
|
||||
# ### Loading Dataset
|
||||
tokenize_func = partial(tokenize_function, tokenizer=tokenizer)
|
||||
train_dataset, val_dataset = get_alpaca_en_zh_dataset(
|
||||
tokenize_func, split_seed=42, data_sample=args.data_sample)
|
||||
# Data analysis
|
||||
stat_dataset(train_dataset)
|
||||
stat_dataset(val_dataset)
|
||||
data_collator = partial(data_collate_fn, tokenizer=tokenizer)
|
||||
print_example(train_dataset[0], tokenizer)
|
||||
|
||||
# ### Loading Dataset
|
||||
tokenize_function = partial(tokenize_function, tokenizer=tokenizer)
|
||||
train_dataset, val_dataset = get_alpaca_en_zh_dataset(
|
||||
tokenize_function, split_seed=42, data_sample=args.data_sample)
|
||||
# Data analysis
|
||||
stat_dataset(train_dataset)
|
||||
stat_dataset(val_dataset)
|
||||
data_collate_fn = partial(data_collate_fn, tokenizer=tokenizer)
|
||||
print_example(train_dataset[0], tokenizer)
|
||||
# ### Setting Config
|
||||
cfg_file = os.path.join(model_dir, 'configuration.json')
|
||||
|
||||
# ### Setting Config
|
||||
cfg_file = os.path.join(model_dir, 'configuration.json')
|
||||
#
|
||||
T_max = get_T_max(len(train_dataset), args.batch_size, args.max_epochs, True)
|
||||
work_dir = get_work_dir(f'runs/{args.model_type}')
|
||||
config = Config({
|
||||
'train': {
|
||||
'dataloader': {
|
||||
'batch_size_per_gpu': args.batch_size,
|
||||
'workers_per_gpu': 1,
|
||||
'shuffle': True,
|
||||
'drop_last': True,
|
||||
'pin_memory': True
|
||||
},
|
||||
'max_epochs':
|
||||
args.max_epochs,
|
||||
'work_dir':
|
||||
work_dir,
|
||||
'optimizer': {
|
||||
'type': 'AdamW',
|
||||
'lr': args.learning_rate,
|
||||
'weight_decay': args.weight_decay,
|
||||
'options': {
|
||||
'cumulative_iters': args.n_accumulate_grad,
|
||||
'grad_clip': {
|
||||
'norm_type': 2,
|
||||
'max_norm': args.grad_clip_norm
|
||||
T_max = get_T_max(
|
||||
len(train_dataset), args.batch_size, args.max_epochs, True)
|
||||
work_dir = get_work_dir(f'runs/{args.model_type}')
|
||||
config = Config({
|
||||
'train': {
|
||||
'dataloader': {
|
||||
'batch_size_per_gpu': args.batch_size,
|
||||
'workers_per_gpu': 1,
|
||||
'shuffle': True,
|
||||
'drop_last': True,
|
||||
'pin_memory': True
|
||||
},
|
||||
'max_epochs':
|
||||
args.max_epochs,
|
||||
'work_dir':
|
||||
work_dir,
|
||||
'optimizer': {
|
||||
'type': 'AdamW',
|
||||
'lr': args.learning_rate,
|
||||
'weight_decay': args.weight_decay,
|
||||
'options': {
|
||||
'cumulative_iters': args.n_accumulate_grad,
|
||||
'grad_clip': {
|
||||
'norm_type': 2,
|
||||
'max_norm': args.grad_clip_norm
|
||||
}
|
||||
}
|
||||
}
|
||||
},
|
||||
'lr_scheduler': {
|
||||
'type': 'CosineAnnealingLR',
|
||||
'T_max': T_max,
|
||||
'eta_min': 0,
|
||||
'options': {
|
||||
'by_epoch': False,
|
||||
'warmup': {
|
||||
'type': 'LinearWarmup',
|
||||
'warmup_ratio': 0.1,
|
||||
'warmup_iters': args.warmup_iters
|
||||
},
|
||||
'lr_scheduler': {
|
||||
'type': 'CosineAnnealingLR',
|
||||
'T_max': T_max,
|
||||
'eta_min': args.learning_rate * 0.1,
|
||||
'options': {
|
||||
'by_epoch': False,
|
||||
'warmup': {
|
||||
'type': 'LinearWarmup',
|
||||
'warmup_ratio': 0.1,
|
||||
'warmup_iters': args.warmup_iters
|
||||
}
|
||||
}
|
||||
}
|
||||
},
|
||||
'hooks': [
|
||||
{
|
||||
'type': 'CheckpointHook',
|
||||
'by_epoch': False,
|
||||
'interval': args.last_save_interval,
|
||||
'max_checkpoint_num': args.last_max_checkpoint_num,
|
||||
'save_trainer_state': args.save_trainer_state
|
||||
},
|
||||
{
|
||||
'type': 'EvaluationHook',
|
||||
'by_epoch': False,
|
||||
'interval': args.eval_interval
|
||||
},
|
||||
{
|
||||
'type': 'BestCkptSaverHook',
|
||||
'metric_key': 'loss',
|
||||
'save_best': True,
|
||||
'rule': 'min',
|
||||
'max_checkpoint_num': args.best_max_checkpoint_num,
|
||||
'save_trainer_state': args.save_trainer_state
|
||||
},
|
||||
{
|
||||
'type': 'TextLoggerHook',
|
||||
'by_epoch': True, # Whether EpochBasedTrainer is used
|
||||
'interval': args.logging_interval
|
||||
},
|
||||
{
|
||||
'type': 'TensorboardHook',
|
||||
'by_epoch': False,
|
||||
'interval': args.tb_interval
|
||||
}
|
||||
]
|
||||
},
|
||||
'hooks': [
|
||||
{
|
||||
'type': 'CheckpointHook',
|
||||
'by_epoch': False,
|
||||
'interval': args.eval_interval,
|
||||
'max_checkpoint_num': args.last_max_checkpoint_num
|
||||
'evaluation': {
|
||||
'dataloader': {
|
||||
'batch_size_per_gpu': args.batch_size,
|
||||
'workers_per_gpu': 1,
|
||||
'shuffle': False,
|
||||
'drop_last': False,
|
||||
'pin_memory': True
|
||||
},
|
||||
{
|
||||
'type': 'EvaluationHook',
|
||||
'by_epoch': False,
|
||||
'interval': args.eval_interval
|
||||
},
|
||||
{
|
||||
'type': 'BestCkptSaverHook',
|
||||
'metric_key': 'loss',
|
||||
'save_best': True,
|
||||
'rule': 'min',
|
||||
'max_checkpoint_num': args.best_max_checkpoint_num
|
||||
},
|
||||
{
|
||||
'type': 'TextLoggerHook',
|
||||
'by_epoch': True, # Whether EpochBasedTrainer is used
|
||||
'interval': args.logging_interval
|
||||
},
|
||||
{
|
||||
'type': 'TensorboardHook',
|
||||
'by_epoch': False,
|
||||
'interval': args.tb_interval
|
||||
}
|
||||
]
|
||||
},
|
||||
'evaluation': {
|
||||
'dataloader': {
|
||||
'batch_size_per_gpu': args.batch_size,
|
||||
'workers_per_gpu': 1,
|
||||
'shuffle': False,
|
||||
'drop_last': False,
|
||||
'pin_memory': True
|
||||
},
|
||||
'metrics': [{
|
||||
'type': 'my_metric',
|
||||
'vocab_size': tokenizer.vocab_size
|
||||
}]
|
||||
}
|
||||
})
|
||||
'metrics': [{
|
||||
'type': 'my_metric',
|
||||
'vocab_size': tokenizer.vocab_size
|
||||
}]
|
||||
}
|
||||
})
|
||||
|
||||
# ### Finetuning
|
||||
# ### Finetuning
|
||||
|
||||
def cfg_modify_fn(cfg: Config) -> Config:
|
||||
cfg.update(config)
|
||||
return cfg
|
||||
|
||||
trainer = EpochBasedTrainer(
|
||||
model=model,
|
||||
cfg_file=cfg_file,
|
||||
data_collator=data_collator,
|
||||
train_dataset=train_dataset,
|
||||
eval_dataset=val_dataset,
|
||||
remove_unused_data=True,
|
||||
seed=42,
|
||||
device='cpu', # No placement for model, leave the model to `device_map`
|
||||
cfg_modify_fn=cfg_modify_fn,
|
||||
)
|
||||
|
||||
trainer.train()
|
||||
|
||||
# ### Visualization
|
||||
tb_dir = os.path.join(work_dir, 'tensorboard_output')
|
||||
plot_image(tb_dir, ['loss'], 0.9)
|
||||
|
||||
|
||||
def cfg_modify_fn(cfg: Config) -> Config:
|
||||
cfg.update(config)
|
||||
return cfg
|
||||
|
||||
|
||||
trainer = EpochBasedTrainer(
|
||||
model=model,
|
||||
cfg_file=cfg_file,
|
||||
data_collator=data_collate_fn,
|
||||
train_dataset=train_dataset,
|
||||
eval_dataset=val_dataset,
|
||||
remove_unused_data=True,
|
||||
seed=42,
|
||||
device='cpu', # No placement for model, leave the model to `device_map`
|
||||
cfg_modify_fn=cfg_modify_fn,
|
||||
)
|
||||
|
||||
trainer.train()
|
||||
|
||||
# ### Visualization
|
||||
tb_dir = os.path.join(work_dir, 'tensorboard_output')
|
||||
plot_image(tb_dir, ['loss'], 0.9)
|
||||
if __name__ == '__main__':
|
||||
args = parse_args()
|
||||
llm_sft(args)
|
||||
|
||||
Reference in New Issue
Block a user