ColossalAI/applications/Chat/examples/train_sft.py

205 lines
9.6 KiB
Python
Raw Normal View History

2023-03-28 12:25:36 +00:00
import argparse
import math
2023-03-28 12:25:36 +00:00
import os
import loralib as lora
import torch
import torch.distributed as dist
from coati.dataset import DataCollatorForSupervisedDataset, SFTDataset, SupervisedDataset
from coati.models import convert_to_lora_module
2023-03-28 12:25:36 +00:00
from coati.trainer import SFTTrainer
from coati.trainer.strategies import DDPStrategy, GeminiStrategy, LowLevelZeroStrategy
2023-03-28 12:25:36 +00:00
from datasets import load_dataset
from torch.optim import Adam
from torch.utils.data import DataLoader
from torch.utils.data.distributed import DistributedSampler
from transformers import AutoTokenizer, BloomConfig, BloomForCausalLM, BloomTokenizerFast, LlamaConfig, LlamaForCausalLM
from transformers.models.gpt2.configuration_gpt2 import GPT2Config
from transformers.models.gpt2.modeling_gpt2 import GPT2LMHeadModel
2023-03-28 12:25:36 +00:00
from transformers.models.gpt2.tokenization_gpt2 import GPT2Tokenizer
from transformers.models.opt.configuration_opt import OPTConfig
from transformers.models.opt.modeling_opt import OPTForCausalLM
from transformers.trainer import get_scheduler
2023-03-28 12:25:36 +00:00
from colossalai.logging import get_dist_logger
from colossalai.nn.optimizer import HybridAdam
from colossalai.tensor import ColoParameter
def train(args):
# configure strategy
if args.strategy == 'ddp':
2023-03-28 12:25:36 +00:00
strategy = DDPStrategy()
elif args.strategy == 'colossalai_gemini':
raise NotImplementedError(
'Gemini is not supported .from_pretrained() yet. We will update this after checkpoint io is ready.')
strategy = GeminiStrategy(placement_policy='cuda')
2023-03-28 12:25:36 +00:00
elif args.strategy == 'colossalai_zero2':
strategy = LowLevelZeroStrategy(stage=2, placement_policy='cuda')
elif args.strategy == 'colossalai_zero2_cpu':
strategy = LowLevelZeroStrategy(stage=2, placement_policy='cpu')
2023-03-28 12:25:36 +00:00
else:
raise ValueError(f'Unsupported strategy "{args.strategy}"')
# configure model
with strategy.model_init_context():
if args.model == 'bloom':
model = convert_to_lora_module(BloomForCausalLM.from_pretrained(args.pretrain),
args.lora_rank).half().cuda()
2023-03-28 12:25:36 +00:00
elif args.model == 'opt':
model = convert_to_lora_module(OPTForCausalLM.from_pretrained(args.pretrain), args.lora_rank).half().cuda()
2023-03-28 12:25:36 +00:00
elif args.model == 'gpt2':
model = convert_to_lora_module(GPT2LMHeadModel.from_pretrained(args.pretrain), args.lora_rank).half().cuda()
2023-03-28 12:25:36 +00:00
elif args.model == 'llama':
model = convert_to_lora_module(LlamaForCausalLM.from_pretrained(args.pretrain),
args.lora_rank).half().cuda()
2023-03-28 12:25:36 +00:00
else:
raise ValueError(f'Unsupported model "{args.model}"')
if args.grad_checkpoint:
model.gradient_checkpointing_enable()
2023-03-28 12:25:36 +00:00
# configure tokenizer
if args.model == 'gpt2':
tokenizer = GPT2Tokenizer.from_pretrained('gpt2')
tokenizer.pad_token = tokenizer.eos_token
elif args.model == 'bloom':
tokenizer = BloomTokenizerFast.from_pretrained('bigscience/bloom-560m')
2023-03-28 12:25:36 +00:00
tokenizer.pad_token = tokenizer.eos_token
elif args.model == 'opt':
tokenizer = AutoTokenizer.from_pretrained("facebook/opt-350m")
tokenizer.pad_token = tokenizer.eos_token
2023-03-28 12:25:36 +00:00
elif args.model == 'llama':
tokenizer = AutoTokenizer.from_pretrained(
args.pretrain,
padding_side="right",
use_fast=False,
)
tokenizer.eos_token = '<\s>'
tokenizer.pad_token = tokenizer.unk_token
2023-03-28 12:25:36 +00:00
else:
raise ValueError(f'Unsupported model "{args.model}"')
if args.model == 'llama' and args.strategy == 'colossalai_gemini':
# this is a hack to deal with the resized embedding
# to make sure all parameters are ColoParameter for Colossal-AI Gemini Compatibility
for name, param in model.named_parameters():
if not isinstance(param, ColoParameter):
sub_module_name = '.'.join(name.split('.')[:-1])
weight_name = name.split('.')[-1]
sub_module = model.get_submodule(sub_module_name)
setattr(sub_module, weight_name, ColoParameter(param))
2023-03-28 12:25:36 +00:00
# configure optimizer
if args.strategy.startswith('colossalai'):
optim = HybridAdam(model.parameters(), lr=args.lr, clipping_norm=1.0)
else:
optim = Adam(model.parameters(), lr=args.lr)
logger = get_dist_logger()
# configure dataset
if args.dataset == 'yizhongw/self_instruct':
train_data = load_dataset(args.dataset, 'super_natural_instructions', split='train')
eval_data = load_dataset(args.dataset, 'super_natural_instructions', split='test')
train_dataset = SFTDataset(train_data, tokenizer, args.max_len)
eval_dataset = SFTDataset(eval_data, tokenizer, args.max_len)
2023-03-28 12:25:36 +00:00
else:
train_dataset = SupervisedDataset(tokenizer=tokenizer,
data_path=args.dataset,
max_datasets_size=args.max_datasets_size,
max_length=args.max_len)
2023-03-28 12:25:36 +00:00
eval_dataset = None
2023-04-17 08:47:44 +00:00
data_collator = DataCollatorForSupervisedDataset(tokenizer=tokenizer)
2023-03-28 12:25:36 +00:00
if dist.is_initialized() and dist.get_world_size() > 1:
train_sampler = DistributedSampler(train_dataset,
shuffle=True,
seed=42,
drop_last=True,
rank=dist.get_rank(),
num_replicas=dist.get_world_size())
if eval_dataset is not None:
eval_sampler = DistributedSampler(eval_dataset,
shuffle=False,
seed=42,
drop_last=False,
rank=dist.get_rank(),
num_replicas=dist.get_world_size())
else:
train_sampler = None
eval_sampler = None
train_dataloader = DataLoader(train_dataset,
shuffle=(train_sampler is None),
sampler=train_sampler,
batch_size=args.batch_size,
collate_fn=data_collator,
pin_memory=True)
if eval_dataset is not None:
eval_dataloader = DataLoader(eval_dataset,
shuffle=(eval_sampler is None),
sampler=eval_sampler,
batch_size=args.batch_size,
collate_fn=data_collator,
pin_memory=True)
else:
eval_dataloader = None
num_update_steps_per_epoch = len(train_dataloader) // args.accumulation_steps
max_steps = math.ceil(args.max_epochs * num_update_steps_per_epoch)
lr_scheduler = get_scheduler("cosine",
optim,
num_warmup_steps=math.ceil(max_steps * 0.03),
num_training_steps=max_steps)
strategy_dict = strategy.prepare(
dict(model=model, optimizer=optim, lr_scheduler=lr_scheduler)
)
model = strategy_dict['model']
optim = strategy_dict['optimizer']
lr_scheduler = strategy_dict['lr_scheduler']
2023-03-28 12:25:36 +00:00
trainer = SFTTrainer(model=model,
strategy=strategy,
optim=optim,
lr_scheduler=lr_scheduler,
2023-03-28 12:25:36 +00:00
max_epochs=args.max_epochs,
accumulation_steps=args.accumulation_steps)
2023-03-28 12:25:36 +00:00
trainer.fit(train_dataloader=train_dataloader,
eval_dataloader=eval_dataloader,
logger=logger,
use_wandb=args.use_wandb)
2023-03-28 12:25:36 +00:00
# save model checkpoint after fitting on only rank0
strategy.save_pretrained(model, path=args.save_path, only_rank0=True, tokenizer=tokenizer)
2023-03-28 12:25:36 +00:00
# save optimizer checkpoint on all ranks
if args.need_optim_ckpt:
strategy.save_optimizer(trainer.optimizer,
'rm_optim_checkpoint_%d.pt' % (torch.cuda.current_device()),
only_rank0=False)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--strategy',
choices=['ddp', 'colossalai_gemini', 'colossalai_zero2', 'colossalai_zero2_cpu'],
default='colossalai_zero2')
2023-03-28 12:25:36 +00:00
parser.add_argument('--model', choices=['gpt2', 'bloom', 'opt', 'llama'], default='bloom')
parser.add_argument('--pretrain', type=str, default=None)
parser.add_argument('--dataset', type=str, default=None)
parser.add_argument('--max_datasets_size', type=int, default=None)
parser.add_argument('--save_path', type=str, default='output')
parser.add_argument('--need_optim_ckpt', type=bool, default=False)
parser.add_argument('--max_epochs', type=int, default=3)
parser.add_argument('--batch_size', type=int, default=4)
parser.add_argument('--max_len', type=int, default=512)
2023-03-28 12:25:36 +00:00
parser.add_argument('--lora_rank', type=int, default=0, help="low-rank adaptation matrices rank")
parser.add_argument('--log_interval', type=int, default=100, help="how many steps to log")
parser.add_argument('--lr', type=float, default=5e-6)
parser.add_argument('--accumulation_steps', type=int, default=8)
parser.add_argument('--use_wandb', default=False, action='store_true')
parser.add_argument('--grad_checkpoint', default=False, action='store_true')
2023-03-28 12:25:36 +00:00
args = parser.parse_args()
train(args)