pip install accelerate

# 업그레이드
pip install accelerate --upgrade

from transformers import TrainingArguments

training_args = TrainingArguments(
    output_dir="your-model",
    learning_rate=2e-5,
    per_device_train_batch_size=16,
    per_device_eval_batch_size=16,
    num_train_epochs=2,
    weight_decay=0.01,
    eval_strategy="epoch",
    save_strategy="epoch",
    load_best_model_at_end=True,
    push_to_hub=True,
)

from transformers import Trainer

trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=dataset["train"],
    eval_dataset=dataset["test"],
    tokenizer=tokenizer,
    data_collator=data_collator,
    compute_metrics=compute_metrics,
)

trainer.train()

# 최신 체크포인트에서 재개
trainer.train(resume_from_checkpoint=True)

# 출력 디렉토리에 저장된 특정 체크포인트에서 재개
trainer.train(resume_from_checkpoint="your-model/checkpoint-1000")

from torch import nn
from transformers import Trainer

class CustomTrainer(Trainer):
    def compute_loss(self,

 model, inputs, return_outputs=False):
        labels = inputs.pop("labels")
        # 순방향 전파
        outputs = model(**inputs)
        logits = outputs.get("logits")
        # 서로 다른 가중치로 3개의 레이블에 대한 사용자 정의 손실을 계산
        loss_fct = nn.CrossEntropyLoss(weight=torch.tensor([1.0, 2.0, 3.0], device=model.device))
        loss = loss_fct(logits.view(-1, self.model.config.num_labels), labels.view(-1))
        return (loss, outputs) if return_outputs else loss

from transformers import TrainerCallback

class EarlyStoppingCallback(TrainerCallback):
    def __init__(self, num_steps=10):
        self.num_steps = num_steps
    
    def on_step_end(self, args, state, control, **kwargs):
        if state.global_step >= self.num_steps:
            return {"should_training_stop": True}
        else:
            return {}

from transformers import Trainer

trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=dataset["train"],
    eval_dataset=dataset["test"],
    tokenizer=tokenizer,
    data_collator=data_collator,
    compute_metrics=compute_metrics,
    callbacks=[EarlyStoppingCallback()],
)

logger = logging.getLogger(__name__)

logging.basicConfig(
    format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
    datefmt="%m/%d/%Y %H:%M:%S",
    handlers=[logging.StreamHandler(sys.stdout)],
)

log_level = training_args.get_process_log_level()
logger.setLevel(log_level)
datasets.utils.logging.set_verbosity(log_level)
transformers.utils.logging.set_verbosity(log_level)

trainer = Trainer(...)

from transformers import TrainingArguments, Trainer

training_args = TrainingArguments(..., neftune_noise_alpha=0.1)
trainer = Trainer(..., args=training_args)

pip install galore-torch

import datasets
from trl import SFTConfig, SFTTrainer

train_dataset = datasets.load_dataset('imdb', split='train')
args = SFTConfig(
    output_dir="./test-galore",
    max_steps=100,
    optim="galore_adamw",
    optim_target_modules=[r".*.attn.*", r".*.mlp.*"],
    gradient_checkpointing=True,
)
trainer = SFTTrainer(
    model="google/gemma-2b",
    args=args,
    train_dataset=train_dataset,
)
trainer.train()

import datasets
from trl import SFTConfig, SFTTrainer

train_dataset = datasets.load_dataset('imdb', split='train')
args = SFTConfig(
    output_dir="./test-galore",
    max_steps=100,
    optim="galore_adamw",
    optim_target_modules=[r".*.attn.*", r".*.mlp.*"],
    optim_args="rank=64, update_proj_gap=100, scale=0.10",
    gradient_checkpointing=True,
)
trainer = SFTTrainer(
    model="google/gemma-2b",
    args=args,
    train_dataset=train_dataset,
)
trainer.train()

import datasets
from trl import SFTConfig, SFTTrainer

train_dataset = datasets.load_dataset('imdb', split='train')
args = SFTConfig(
    output_dir="./test-galore",
    max_steps=100,
    optim="galore_adamw_layerwise",
    optim_target_modules=[r".*.attn.*", r".*.mlp.*"],
    gradient_checkpointing=True,
)
trainer = SFTTrainer(
    model="google/gemma-2b",
    args=args,
    train_dataset=train_dataset,
)
trainer.train()

import datasets
from trl import SFTConfig, SFTTrainer

train_dataset = datasets.load_dataset('imdb', split='train')
args = SFTConfig(
    output_dir="./test-lomo",
    max_steps=100,
    optim="adalomo",
    gradient_checkpointing=True,
)
trainer = SFTTrainer(
    model="google/gemma-2b",
    args=args,
    train_dataset=train_dataset,
)
trainer.train()

accelerate launch \
    ./examples/pytorch/text-classification/run_glue.py \
    --model_name_or_path google-bert/bert-base-cased \
    --task_name $TASK_NAME \
    --do_train \
    --do_eval \
    --max_seq_length 128 \
    --per_device_train_batch_size 16 \
    --learning_rate 5e-5 \
    --num_train_epochs 3 \
    --output_dir /tmp/$TASK_NAME/ \
    --overwrite_output_dir

accelerate launch --num_processes=2 \
    --use_fsdp \
    --mixed_precision=bf16 \
    --fsdp_auto_wrap_policy=TRANSFORMER_BASED_WRAP  \
    --fsdp_transformer_layer_cls_to_wrap="BertLayer" \
    --fsdp_sharding_strategy=1 \
    --fsdp_state_dict_type=FULL_STATE_DICT \
    ./examples/pytorch/text-classification/run_glue.py \
    --model_name_or_path google-bert/bert-base-cased \
    --task_name $TASK_NAME \
    --do_train \
    --do_eval \
    --max_seq_length 128 \
    --per_device_train_batch_size 16 \
    --learning_rate 5e-5 \
    --num_train_epochs 3 \
    --output_dir /tmp/$TASK_NAME/ \
    --overwrite_output_dir