Skip to content

🧠 两小时从 0 训出你的 64M 大模型。Train your own 64M LLM from scratch in 2 hours.

添加了argparse,方便命令行输入参数

Yu Chengzhang51dcf51c5dfe
diff --git a/1-pretrain.py b/1-pretrain.py
index 56c937d..d6bfff6 100644
--- a/1-pretrain.py
+++ b/1-pretrain.py
@@ -1,5 +1,6 @@
import os
import platform
+import argparse
import time
import math
import warnings
@@ -23,66 +24,65 @@ def Logger(content):
def get_lr(it, all):
- warmup_iters = 0
+ warmup_iters = args.warmup_iters
lr_decay_iters = all
- min_lr = learning_rate / 10
+ min_lr = args.learning_rate / 10
if it < warmup_iters:
- return learning_rate * it / warmup_iters
+ return args.learning_rate * it / warmup_iters
if it > lr_decay_iters:
return min_lr
decay_ratio = (it - warmup_iters) / (lr_decay_iters - warmup_iters)
assert 0 <= decay_ratio <= 1
coeff = 0.5 * (1.0 + math.cos(math.pi * decay_ratio))
- return min_lr + coeff * (learning_rate - min_lr)
+ return min_lr + coeff * (args.learning_rate - min_lr)
-def train_epoch(epoch, wandb, accumulation_steps=8):
+def train_epoch(epoch, wandb):
start_time = time.time()
for step, (X, Y) in enumerate(train_loader):
- X = X.to(device)
- Y = Y.to(device)
+ X = X.to(args.device)
+ Y = Y.to(args.device)
- lr = get_lr(epoch * iter_per_epoch + step, epochs * iter_per_epoch)
+ lr = get_lr(epoch * iter_per_epoch + step, args.epochs * iter_per_epoch)
for param_group in optimizer.param_groups:
param_group['lr'] = lr
with ctx:
out = model(X, Y)
- loss = out.last_loss / accumulation_steps
+ loss = out.last_loss / args.accumulation_steps
scaler.scale(loss).backward()
- if (step + 1) % accumulation_steps == 0:
+ if (step + 1) % args.accumulation_steps == 0:
scaler.unscale_(optimizer)
- torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)
+ torch.nn.utils.clip_grad_norm_(model.parameters(), args.grad_clip)
scaler.step(optimizer)
scaler.update()
optimizer.zero_grad(set_to_none=True)
- if step % 100 == 0:
+ if step % args.log_interval == 0:
spend_time = time.time() - start_time
Logger(
'Epoch:[{}/{}]({}/{}) loss:{:.3f} lr:{:.7f} epoch_Time:{}min:'.format(
epoch,
- epochs,
+ args.epochs,
step,
iter_per_epoch,
- loss.item() * accumulation_steps,
+ loss.item() * args.accumulation_steps,
optimizer.param_groups[-1]['lr'],
spend_time / (step + 1) * iter_per_epoch // 60 - spend_time // 60))
- if wandb != None:
- wandb.log({"loss": loss.item() * accumulation_steps,
+ if wandb is not None:
+ wandb.log({"loss": loss.item() * args.accumulation_steps,
"lr": optimizer.param_groups[-1]['lr'],
"epoch_Time": spend_time / (step + 1) * iter_per_epoch // 60 - spend_time // 60})
- if (step + 1) % 1000 == 0 and (not ddp or dist.get_rank() == 0):
+ if (step + 1) % args.save_interval == 0 and (not ddp or dist.get_rank() == 0):
model.eval()
- # torch.save(model.state_dict(), '{}/iter_{}.pth'.format(save_dir, int(step + epoch * iter_per_epoch)))
moe_path = '_moe' if lm_config.use_moe else ''
- ckp = f'{save_dir}/pretrain_{lm_config.dim}{moe_path}.pth'
+ ckp = f'{args.save_dir}/pretrain_{lm_config.dim}{moe_path}.pth'
if isinstance(model, torch.nn.parallel.DistributedDataParallel):
state_dict = model.module.state_dict()
@@ -97,17 +97,8 @@ def init_model():
def count_parameters(model):
return sum(p.numel() for p in model.parameters() if p.requires_grad)
- # model init
- model = Transformer(lm_config).to(device)
+ model = Transformer(lm_config).to(args.device)
moe_path = '_moe' if lm_config.use_moe else ''
- # ckp = f'{save_dir}/pretrain_{lm_config.dim}{moe_path}.pth'
- #
- # state_dict = torch.load(ckp, map_location=device)
- # unwanted_prefix = '_orig_mod.'
- # for k, v in list(state_dict.items()):
- # if k.startswith(unwanted_prefix):
- # state_dict[k[len(unwanted_prefix):]] = state_dict.pop(k)
- # model.load_state_dict(state_dict, strict=False)
Logger(f'LLM总参数量:{count_parameters(model) / 1e6:.3f} 百万')
return model
@@ -125,81 +116,78 @@ def init_distributed_mode():
torch.cuda.set_device(DEVICE)
# torchrun --nproc_per_node 2 1-pretrain.py
-# I/O
if __name__ == "__main__":
- # -----------------------------------------------------------------------------
+ parser = argparse.ArgumentParser(description="MiniMind Pretraining")
+ parser.add_argument("--out_dir", type=str, default="out", help="Output directory")
+ parser.add_argument("--epochs", type=int, default=20, help="Number of epochs")
+ parser.add_argument("--batch_size", type=int, default=64, help="Batch size")
+ parser.add_argument("--learning_rate", type=float, default=2e-4, help="Learning rate")
+ parser.add_argument("--device", type=str, default="cuda:0" if torch.cuda.is_available() else "cpu", help="Device to use")
+ parser.add_argument("--dtype", type=str, default="bfloat16", help="Data type")
+ parser.add_argument("--use_wandb", action="store_true", help="Use Weights & Biases")
+ parser.add_argument("--wandb_project", type=str, default="MiniMind-Pretrain", help="Weights & Biases project name")
+ parser.add_argument("--num_workers", type=int, default=8, help="Number of workers for data loading")
+ parser.add_argument("--data_path", type=str, default="./dataset/pretrain_data.bin", help="Path to training data")
+ parser.add_argument("--ddp", action="store_true", help="Use DistributedDataParallel")
+ parser.add_argument("--accumulation_steps", type=int, default=8, help="Gradient accumulation steps")
+ parser.add_argument("--grad_clip", type=float, default=1.0, help="Gradient clipping threshold")
+ parser.add_argument("--warmup_iters", type=int, default=0, help="Number of warmup iterations")
+ parser.add_argument("--log_interval", type=int, default=100, help="Logging interval")
+ parser.add_argument("--save_interval", type=int, default=1000, help="Model saving interval")
+
+ args = parser.parse_args()
+
lm_config = LMConfig()
max_seq_len = lm_config.max_seq_len
- out_dir = 'out'
- epochs = 20
- batch_size = 64
- learning_rate = 2e-4
- device = 'cuda:0' if torch.cuda.is_available() else 'cpu'
- dtype = 'bfloat16'
- save_dir = os.path.join(out_dir)
- os.makedirs(save_dir, exist_ok=True)
- os.makedirs(out_dir, exist_ok=True)
- tokens_per_iter = batch_size * max_seq_len
+ args.save_dir = os.path.join(args.out_dir)
+ os.makedirs(args.save_dir, exist_ok=True)
+ os.makedirs(args.out_dir, exist_ok=True)
+ tokens_per_iter = args.batch_size * max_seq_len
torch.manual_seed(1337)
- device_type = device if "cuda" in device else "cpu"
-
- use_wandb = True #是否使用wandb
- wandb_project = "MiniMind-Pretrain"
- wandb_run_name = f"MiniMind-Pretrain-Epoch-{epochs}-BatchSize-{batch_size}-LearningRate-{learning_rate}"
+ device_type = "cuda" if "cuda" in args.device else "cpu"
+ args.wandb_run_name = f"MiniMind-Pretrain-Epoch-{args.epochs}-BatchSize-{args.batch_size}-LearningRate-{args.learning_rate}"
- ctx = (
- nullcontext()
- if device_type == "cpu"
- else torch.cuda.amp.autocast()
- )
+ ctx = nullcontext() if device_type == "cpu" else torch.cuda.amp.autocast()
ddp = int(os.environ.get("RANK", -1)) != -1 # is this a ddp run?
ddp_local_rank, DEVICE = 0, "cuda:0"
if ddp:
init_distributed_mode()
- device = torch.device(DEVICE)
+ args.device = torch.device(DEVICE)
- if use_wandb and (not ddp or ddp_local_rank == 0):
+ if args.use_wandb and (not ddp or ddp_local_rank == 0):
import wandb
- wandb.init(project=wandb_project, name=wandb_run_name)
+ wandb.init(project=args.wandb_project, name=args.wandb_run_name)
else:
wandb = None
- # -----------------------------------------------------------------------------
- # -----init dataloader------
- data_path_list = ['./dataset/pretrain_data.bin']
+ data_path_list = [args.data_path]
train_ds = PretrainDataset(data_path_list, max_length=max_seq_len, memmap=True)
train_sampler = DistributedSampler(train_ds) if ddp else None
- num_workers = 8 # 可以根据系统的 CPU 核心数来调整
train_loader = DataLoader(
train_ds,
- batch_size=batch_size,
+ batch_size=args.batch_size,
pin_memory=True,
drop_last=False,
shuffle=False,
- num_workers=num_workers,
+ num_workers=args.num_workers,
sampler=train_sampler
)
- # init model
model = init_model()
- scaler = torch.cuda.amp.GradScaler(enabled=(dtype == dtype))
- # optimizer
- optimizer = optim.Adam(model.parameters(), lr=learning_rate)
- # compile the model
+ scaler = torch.cuda.amp.GradScaler(enabled=(args.dtype == args.dtype))
+ optimizer = optim.Adam(model.parameters(), lr=args.learning_rate)
+
if False and platform.system() != 'Windows' and float(torch.__version__.split('.')[0]) >= 2:
Logger("compiling the model... (takes a ~minute)")
unoptimized_model = model
model = torch.compile(model)
if ddp:
- # Ignore the freqs_cis buffer so that DDP does not broadcast it at
- # construction time since NCCL does not support ComplexFloat
model._ddp_params_and_buffers_to_ignore = {"pos_cis"}
model = DistributedDataParallel(model, device_ids=[ddp_local_rank])
- # training loop
iter_per_epoch = len(train_loader)
- for epoch in range(epochs):
+ for epoch in range(args.epochs):
train_epoch(epoch, wandb)
diff --git a/3-full_sft.py b/3-full_sft.py
index c50dedf..fc58bb3 100644
--- a/3-full_sft.py
+++ b/3-full_sft.py
@@ -1,5 +1,6 @@
import os
import platform
+import argparse
import time
import math
import warnings
@@ -12,7 +13,6 @@ from contextlib import nullcontext
from torch import optim
from torch.nn.parallel import DistributedDataParallel
-from torch.optim.lr_scheduler import CosineAnnealingLR
from torch.utils.data import DataLoader, DistributedSampler
from transformers import AutoTokenizer, AutoModel
from model.model import Transformer
@@ -28,28 +28,27 @@ def Logger(content):
def get_lr(it, all):
- warmup_iters = 0
+ warmup_iters = args.warmup_iters
lr_decay_iters = all
- min_lr = learning_rate / epochs
+ min_lr = args.learning_rate / 10
if it < warmup_iters:
- return learning_rate * it / warmup_iters
+ return args.learning_rate * it / warmup_iters
if it > lr_decay_iters:
return min_lr
decay_ratio = (it - warmup_iters) / (lr_decay_iters - warmup_iters)
assert 0 <= decay_ratio <= 1
coeff = 0.5 * (1.0 + math.cos(math.pi * decay_ratio))
- return min_lr + coeff * (learning_rate - min_lr)
+ return min_lr + coeff * (args.learning_rate - min_lr)
-# ------------------------------------------------------------------------------
def train_epoch(epoch, wandb):
start_time = time.time()
for step, (X, Y, loss_mask) in enumerate(train_loader):
- X = X.to(device)
- Y = Y.to(device)
- loss_mask = loss_mask.to(device)
- lr = get_lr(epoch * iter_per_epoch + step, epochs * iter_per_epoch)
+ X = X.to(args.device)
+ Y = Y.to(args.device)
+ loss_mask = loss_mask.to(args.device)
+ lr = get_lr(epoch * iter_per_epoch + step, args.epochs * iter_per_epoch)
for param_group in optimizer.param_groups:
param_group['lr'] = lr
@@ -59,41 +58,38 @@ def train_epoch(epoch, wandb):
loss_mask = loss_mask.view(-1)
loss = torch.sum(loss * loss_mask) / loss_mask.sum()
- # Backward pass
scaler.scale(loss).backward()
- # Unscale gradients and clip them
- scaler.unscale_(optimizer)
- torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)
+ if (step + 1) % args.accumulation_steps == 0:
+ scaler.unscale_(optimizer)
+ torch.nn.utils.clip_grad_norm_(model.parameters(), args.grad_clip)
- # Update parameters
- scaler.step(optimizer)
- scaler.update()
+ scaler.step(optimizer)
+ scaler.update()
- # Zero the gradients
- optimizer.zero_grad(set_to_none=True)
+ optimizer.zero_grad(set_to_none=True)
- # 打印日志
- if step % 100 == 0:
+ if step % args.log_interval == 0:
spend_time = time.time() - start_time
Logger(
- 'Epoch:[{}/{}]({}/{}) loss:{:.3f} lr:{:.8f} epoch_Time:{}min:'.format(
+ 'Epoch:[{}/{}]({}/{}) loss:{:.3f} lr:{:.7f} epoch_Time:{}min:'.format(
epoch,
- epochs,
+ args.epochs,
step,
iter_per_epoch,
- loss,
+ loss.item(),
optimizer.param_groups[-1]['lr'],
spend_time / (step + 1) * iter_per_epoch // 60 - spend_time // 60))
- if use_wandb != None:
- wandb.log({"loss": loss, "lr": optimizer.param_groups[-1]['lr'],
- "epoch_Time": spend_time / (step + 1) * iter_per_epoch // 60 - spend_time // 60})
+ if wandb is not None:
+ wandb.log({"loss": loss.item(),
+ "lr": optimizer.param_groups[-1]['lr'],
+ "epoch_Time": spend_time / (step + 1) * iter_per_epoch // 60 - spend_time // 60})
- if (step + 1) % 1000 == 0 and (not ddp or dist.get_rank() == 0):
+ if (step + 1) % args.save_interval == 0 and (not ddp or dist.get_rank() == 0):
model.eval()
- # torch.save(model.state_dict(), '{}/sft_iter_{}.pth'.format(save_dir, int(step + epoch * iter_per_epoch)))
moe_path = '_moe' if lm_config.use_moe else ''
- ckp = f'{save_dir}/full_sft_{lm_config.dim}{moe_path}.pth'
+ ckp = f'{args.save_dir}/full_sft_{lm_config.dim}{moe_path}.pth'
+
if isinstance(model, torch.nn.parallel.DistributedDataParallel):
state_dict = model.module.state_dict()
else:
@@ -103,7 +99,7 @@ def train_epoch(epoch, wandb):
model.train()
-def init_model(lm_config):
+def init_model():
tokenizer = AutoTokenizer.from_pretrained('./model/minimind_tokenizer')
model_from = 1 # 1从权重,2用transformers
@@ -114,7 +110,7 @@ def init_model(lm_config):
model = Transformer(lm_config)
moe_path = '_moe' if lm_config.use_moe else ''
ckp = f'./out/pretrain_{lm_config.dim}{moe_path}.pth'
- state_dict = torch.load(ckp, map_location=device)
+ state_dict = torch.load(ckp, map_location=args.device)
unwanted_prefix = '_orig_mod.'
for k, v in list(state_dict.items()):
if k.startswith(unwanted_prefix):
@@ -124,7 +120,7 @@ def init_model(lm_config):
model = AutoModel.from_pretrained('./minimind', trust_remote_code=True)
Logger(f'LLM总参数量:{count_parameters(model) / 1e6:.3f} 百万')
- model = model.to(device)
+ model = model.to(args.device)
return model, tokenizer
@@ -141,83 +137,78 @@ def init_distributed_mode():
torch.cuda.set_device(DEVICE)
-# I/O
if __name__ == "__main__":
- # -----------------------------------------------------------------------------
+ parser = argparse.ArgumentParser(description="MiniMind Full SFT")
+ parser.add_argument("--out_dir", type=str, default="out", help="Output directory")
+ parser.add_argument("--epochs", type=int, default=19, help="Number of epochs")
+ parser.add_argument("--batch_size", type=int, default=40, help="Batch size")
+ parser.add_argument("--learning_rate", type=float, default=1e-4, help="Learning rate")
+ parser.add_argument("--device", type=str, default="cuda:0" if torch.cuda.is_available() else "cpu", help="Device to use")
+ parser.add_argument("--dtype", type=str, default="bfloat16", help="Data type")
+ parser.add_argument("--use_wandb", action="store_true", help="Use Weights & Biases")
+ parser.add_argument("--wandb_project", type=str, default="MiniMind-Full-SFT", help="Weights & Biases project name")
+ parser.add_argument("--num_workers", type=int, default=8, help="Number of workers for data loading")
+ parser.add_argument("--ddp", action="store_true", help="Use DistributedDataParallel")
+ parser.add_argument("--accumulation_steps", type=int, default=1, help="Gradient accumulation steps")
+ parser.add_argument("--grad_clip", type=float, default=1.0, help="Gradient clipping threshold")
+ parser.add_argument("--warmup_iters", type=int, default=0, help="Number of warmup iterations")
+ parser.add_argument("--log_interval", type=int, default=100, help="Logging interval")
+ parser.add_argument("--save_interval", type=int, default=1000, help="Model saving interval")
+
+ args = parser.parse_args()
+
lm_config = LMConfig()
max_seq_len = lm_config.max_seq_len
- out_dir = 'out'
- epochs = 19
- gradient_accumulation_steps = 1
- batch_size = 40
- learning_rate = 1e-4
- device = 'cuda:0' if torch.cuda.is_available() else 'cpu'
- dtype = 'bfloat16'
- # dtype = 'float16'
- save_dir = os.path.join(out_dir)
- os.makedirs(save_dir, exist_ok=True)
- tokens_per_iter = gradient_accumulation_steps * batch_size * max_seq_len
- os.makedirs(out_dir, exist_ok=True)
+ args.save_dir = os.path.join(args.out_dir)
+ os.makedirs(args.save_dir, exist_ok=True)
+ os.makedirs(args.out_dir, exist_ok=True)
+ tokens_per_iter = args.batch_size * max_seq_len
torch.manual_seed(1337)
- device_type = device if "cuda" in device else "cpu"
+ device_type = "cuda" if "cuda" in args.device else "cpu"
- use_wandb = True #是否使用wandb
- wandb_project = "MiniMind-Full-SFT"
- wandb_run_name = f"MiniMind-Full-SFT-Epoch-{epochs}-BatchSize-{batch_size}-LearningRate-{learning_rate}"
- if use_wandb:
- import wandb
- wandb.init(project=wandb_project, name=wandb_run_name)
- else:
- wandb = None
-
- ctx = (
- nullcontext()
- if device_type == "cpu"
- else torch.cuda.amp.autocast()
- )
+ args.wandb_run_name = f"MiniMind-Full-SFT-Epoch-{args.epochs}-BatchSize-{args.batch_size}-LearningRate-{args.learning_rate}"
- ### ddp config
+ ctx = nullcontext() if device_type == "cpu" else torch.cuda.amp.autocast()
ddp = int(os.environ.get("RANK", -1)) != -1 # is this a ddp run?
ddp_local_rank, DEVICE = 0, "cuda:0"
if ddp:
init_distributed_mode()
- device = torch.device(DEVICE)
- # -----------------------------------------------------------------------------
+ args.device = torch.device(DEVICE)
+
+ if args.use_wandb and (not ddp or ddp_local_rank == 0):
+ import wandb
+ wandb.init(project=args.wandb_project, name=args.wandb_run_name)
+ else:
+ wandb = None
+
+ model, tokenizer = init_model()
- model, tokenizer = init_model(lm_config)
- # -----init dataloader------
df = pd.read_csv('./dataset/sft_data_single.csv')
df = df.sample(frac=1.0)
train_ds = SFTDataset(df, tokenizer, max_length=max_seq_len)
train_sampler = DistributedSampler(train_ds) if ddp else None
train_loader = DataLoader(
train_ds,
- batch_size=batch_size,
- pin_memory=False,
+ batch_size=args.batch_size,
+ pin_memory=True,
drop_last=False,
shuffle=False,
- num_workers=8,
+ num_workers=args.num_workers,
sampler=train_sampler
)
- scaler = torch.cuda.amp.GradScaler(enabled=(dtype == dtype))
- # optimizer
- optimizer = optim.Adam(model.parameters(), lr=learning_rate)
+ scaler = torch.cuda.amp.GradScaler(enabled=(args.dtype == args.dtype))
+ optimizer = optim.Adam(model.parameters(), lr=args.learning_rate)
- iter_per_epoch = len(train_loader)
- # compile the model
- if False and not lm_config.use_moe and platform.system() != 'Windows' and float(
- torch.__version__.split('.')[0]) >= 2:
+ if False and not lm_config.use_moe and platform.system() != 'Windows' and float(torch.__version__.split('.')[0]) >= 2:
Logger("compiling the model... (takes a ~minute)")
unoptimized_model = model
- model = torch.compile(model) # requires PyTorch 2.0
+ model = torch.compile(model)
if ddp:
- # Ignore the pos_cis buffer so that DDP does not broadcast it at
- # construction time since NCCL does not support ComplexFloat
model._ddp_params_and_buffers_to_ignore = {"pos_cis"}
model = DistributedDataParallel(model, device_ids=[ddp_local_rank])
- # training loop
- for epoch in range(epochs,wandb):
- train_epoch(epoch)
+ iter_per_epoch = len(train_loader)
+ for epoch in range(args.epochs):
+ train_epoch(epoch, wandb)
diff --git a/4-lora_sft.py b/4-lora_sft.py
index 128041a..44d69ba 100644
--- a/4-lora_sft.py
+++ b/4-lora_sft.py
@@ -1,5 +1,6 @@
import os
import platform
+import argparse
import time
import math
import warnings
@@ -16,32 +17,36 @@ from torch.utils.data import DataLoader
from model.LMConfig import LMConfig
from model.dataset import SFTDataset
-warnings.filterwarnings('ignore', category=UserWarning)
+warnings.filterwarnings('ignore')
-def get_lr(it):
- warmup_iters = 1000
- lr_decay_iters = 80000
- min_lr = 1e-5
+def Logger(content):
+ print(content)
+
+
+def get_lr(it, all):
+ warmup_iters = args.warmup_iters
+ lr_decay_iters = all
+ min_lr = args.learning_rate / 10
if it < warmup_iters:
- return learning_rate * it / warmup_iters
+ return args.learning_rate * it / warmup_iters
if it > lr_decay_iters:
return min_lr
decay_ratio = (it - warmup_iters) / (lr_decay_iters - warmup_iters)
assert 0 <= decay_ratio <= 1
coeff = 0.5 * (1.0 + math.cos(math.pi * decay_ratio))
- return min_lr + coeff * (learning_rate - min_lr)
+ return min_lr + coeff * (args.learning_rate - min_lr)
-# ------------------------------------------------------------------------------
def train_epoch(epoch, wandb):
start_time = time.time()
for step, (X, Y, loss_mask) in enumerate(train_loader):
- X = X.to(device)
- Y = Y.to(device)
- loss_mask = loss_mask.to(device)
- lr = get_lr(epoch * iter_per_epoch + step)
+ X = X.to(args.device)
+ Y = Y.to(args.device)
+ loss_mask = loss_mask.to(args.device)
+
+ lr = get_lr(epoch * iter_per_epoch + step, args.epochs * iter_per_epoch)
for param_group in optimizer.param_groups:
param_group['lr'] = lr
@@ -50,31 +55,37 @@ def train_epoch(epoch, wandb):
loss = F.cross_entropy(logits.view(-1, logits.size(-1)), Y.view(-1), ignore_index=0, reduction='none')
loss_mask = loss_mask.view(-1)
loss = torch.sum(loss * loss_mask) / loss_mask.sum()
+ loss = loss / args.accumulation_steps
scaler.scale(loss).backward()
- scaler.unscale_(optimizer)
- torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)
+ if (step + 1) % args.accumulation_steps == 0:
+ scaler.unscale_(optimizer)
+ torch.nn.utils.clip_grad_norm_(model.parameters(), args.grad_clip)
- scaler.step(optimizer)
- scaler.update()
+ scaler.step(optimizer)
+ scaler.update()
- optimizer.zero_grad(set_to_none=True)
+ optimizer.zero_grad(set_to_none=True)
- if step % 100 == 0:
+ if step % args.log_interval == 0:
spend_time = time.time() - start_time
- print(
+ Logger(
'Epoch:[{}/{}]({}/{}) loss:{:.3f} lr:{:.7f} epoch_Time:{}min:'.format(
epoch,
- epochs,
+ args.epochs,
step,
iter_per_epoch,
- loss.item(),
+ loss.item() * args.accumulation_steps,
optimizer.param_groups[-1]['lr'],
spend_time / (step + 1) * iter_per_epoch // 60 - spend_time // 60))
- if use_wandb != None:
- wandb.log({"loss": loss.item(), "lr": optimizer.param_groups[-1]['lr'],
- "epoch_Time": spend_time / (step + 1) * iter_per_epoch // 60 - spend_time // 60})
+ if wandb is not None:
+ wandb.log({"loss": loss.item() * args.accumulation_steps,
+ "lr": optimizer.param_groups[-1]['lr'],
+ "epoch_Time": spend_time / (step + 1) * iter_per_epoch // 60 - spend_time // 60})
+
+ if (step + 1) % args.save_interval == 0:
+ model.save_pretrained(args.save_dir)
def find_all_linear_names(model):
@@ -94,7 +105,7 @@ def init_model():
model_name_or_path = "./minimind"
tokenizer_name_or_path = "./minimind"
tokenizer = AutoTokenizer.from_pretrained(tokenizer_name_or_path, trust_remote_code=True, use_fast=False)
- model = AutoModelForCausalLM.from_pretrained(model_name_or_path, trust_remote_code=True).to(device)
+ model = AutoModelForCausalLM.from_pretrained(model_name_or_path, trust_remote_code=True).to(args.device)
target_modules = find_all_linear_names(model)
peft_config = LoraConfig(
@@ -107,73 +118,70 @@ def init_model():
)
model = get_peft_model(model, peft_config)
model.print_trainable_parameters()
- model = model.to(device)
+ model = model.to(args.device)
return model, tokenizer
-# I/O
if __name__ == "__main__":
- # -----------------------------------------------------------------------------
+ parser = argparse.ArgumentParser(description="MiniMind LoRA Fine-tuning")
+ parser.add_argument("--out_dir", type=str, default="out", help="Output directory")
+ parser.add_argument("--epochs", type=int, default=20, help="Number of epochs")
+ parser.add_argument("--batch_size", type=int, default=16, help="Batch size")
+ parser.add_argument("--learning_rate", type=float, default=1e-4, help="Learning rate")
+ parser.add_argument("--device", type=str, default="cuda:0" if torch.cuda.is_available() else "cpu", help="Device to use")
+ parser.add_argument("--dtype", type=str, default="bfloat16", help="Data type")
+ parser.add_argument("--use_wandb", action="store_true", help="Use Weights & Biases")
+ parser.add_argument("--wandb_project", type=str, default="MiniMind-LoRA", help="Weights & Biases project name")
+ parser.add_argument("--num_workers", type=int, default=0, help="Number of workers for data loading")
+ parser.add_argument("--accumulation_steps", type=int, default=1, help="Gradient accumulation steps")
+ parser.add_argument("--grad_clip", type=float, default=1.0, help="Gradient clipping threshold")
+ parser.add_argument("--warmup_iters", type=int, default=1000, help="Number of warmup iterations")
+ parser.add_argument("--log_interval", type=int, default=100, help="Logging interval")
+ parser.add_argument("--save_interval", type=int, default=1000, help="Model saving interval")
+
+ args = parser.parse_args()
+
lm_config = LMConfig()
max_seq_len = lm_config.max_seq_len
- out_dir = 'out'
- epochs = 20
- gradient_accumulation_steps = 1
- batch_size = 16
- learning_rate = 1e-4
- weight_decay = 1e-1
- device = 'cuda:0'
- dtype = 'bfloat16'
- save_dir = os.path.join(out_dir)
- os.makedirs(save_dir, exist_ok=True)
- tokens_per_iter = gradient_accumulation_steps * batch_size * max_seq_len
- os.makedirs(out_dir, exist_ok=True)
+ args.save_dir = os.path.join(args.out_dir)
+ os.makedirs(args.save_dir, exist_ok=True)
+ os.makedirs(args.out_dir, exist_ok=True)
+ tokens_per_iter = args.batch_size * max_seq_len
torch.manual_seed(1337)
- device_type = device if "cuda" in device else "cpu"
+ device_type = "cuda" if "cuda" in args.device else "cpu"
+
+ args.wandb_run_name = f"MiniMind-LoRA-Epoch-{args.epochs}-BatchSize-{args.batch_size}-LearningRate-{args.learning_rate}"
- use_wandb = True #是否使用wandb
- wandb_project = "MiniMind-LoRA"
- wandb_run_name = f"MiniMind-LoRA-Epoch-{epochs}-BatchSize-{batch_size}-LearningRate-{learning_rate}"
- if use_wandb:
+ ctx = nullcontext() if device_type == "cpu" else torch.cuda.amp.autocast()
+
+ if args.use_wandb:
import wandb
- wandb.init(project=wandb_project, name=wandb_run_name)
+ wandb.init(project=args.wandb_project, name=args.wandb_run_name)
else:
wandb = None
- ctx = (
- nullcontext()
- if device_type == "cpu"
- else torch.cuda.amp.autocast()
- )
- # -----------------------------------------------------------------------------
-
model, tokenizer = init_model()
- # -----init dataloader------
df = pd.read_csv('./dataset/sft_data.csv')
df = df.sample(frac=1.0)
train_ds = SFTDataset(df, tokenizer, max_length=max_seq_len)
train_loader = DataLoader(
train_ds,
- batch_size=batch_size,
- pin_memory=False,
+ batch_size=args.batch_size,
+ pin_memory=True,
drop_last=False,
shuffle=False,
- num_workers=0,
+ num_workers=args.num_workers,
)
- scaler = torch.cuda.amp.GradScaler(enabled=(dtype == 'float16'))
- # optimizer
- optimizer = optim.Adam(model.parameters(), lr=learning_rate)
- iter_per_epoch = len(train_loader)
- # compile the model
+ scaler = torch.cuda.amp.GradScaler(enabled=(args.dtype == 'float16'))
+ optimizer = optim.Adam(model.parameters(), lr=args.learning_rate)
+
if False and platform.system() != 'Windows' and float(torch.__version__.split('.')[0]) >= 2:
- print("compiling the model... (takes a ~minute)")
+ Logger("compiling the model... (takes a ~minute)")
unoptimized_model = model
model = torch.compile(model)
- raw_model = model
- # training loop
- for epoch in range(epochs):
+ iter_per_epoch = len(train_loader)
+ for epoch in range(args.epochs):
train_epoch(epoch, wandb)
- model.save_pretrained('minimind')