Trainer¶

pip install accelerate

# 升级
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"],
    processing_class=tokenizer,
    data_collator=data_collator,
    compute_metrics=compute_metrics,
)

trainer.train()

# 从最新的检查点恢复
trainer.train(resume_from_checkpoint=True)

# 从输出目录中保存的特定检查点恢复
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")
        # 前向传播
        outputs = model(**inputs)
        logits = outputs.get("logits")
        # 计算自定义损失，为3个标签设置不同的权重
        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"],
    processing_class=tokenizer,
    data_collator=data_collator,
    compute_metrics=compute_metrics,
    callback=[EarlyStoppingCallback()],
)

import logging
import sys

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(...)

my_app.py ... --log_level warning --log_level_replica error

my_app.py ... --log_level warning --log_level_replica error --log_on_each_node 0

# set to only report errors
my_app.py ... --log_level error --log_level_replica error --log_on_each_node 0

from transformers import TrainingArguments, Trainer

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

pip install galore-torch

import torch
import datasets
import trl

from transformers import TrainingArguments, AutoConfig, AutoTokenizer, AutoModelForCausalLM

train_dataset = datasets.load_dataset('imdb', split='train')

args = TrainingArguments(
    output_dir="./test-galore",
    max_steps=100,
    per_device_train_batch_size=2,
    optim="galore_adamw",
    optim_target_modules=[r".*.attn.*", r".*.mlp.*"]
)

model_id = "google/gemma-2b"

config = AutoConfig.from_pretrained(model_id)

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_config(config).to(0)

trainer = trl.SFTTrainer(
    model=model,
    args=args,
    train_dataset=train_dataset,
    dataset_text_field='text',
    max_seq_length=512,
)

trainer.train()

import torch
import datasets
import trl

from transformers import TrainingArguments, AutoConfig, AutoTokenizer, AutoModelForCausalLM

train_dataset = datasets.load_dataset('imdb', split='train')

args = TrainingArguments(
    output_dir="./test-galore",
    max_steps=100,
    per_device_train_batch_size=2,
    optim="galore_adamw",
    optim_target_modules=[r".*.attn.*", r".*.mlp.*"],
    optim_args="rank=64, update_proj_gap=100, scale=0.10",
)

model_id = "google/gemma-2b"

config = AutoConfig.from_pretrained(model_id)

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_config(config).to(0)

trainer = trl.SFTTrainer(
    model=model,
    args=args,
    train_dataset=train_dataset,
    dataset_text_field='text',
    max_seq_length=512,
)

trainer.train()

import torch
import datasets
import trl

from transformers import TrainingArguments, AutoConfig, AutoTokenizer, AutoModelForCausalLM

train_dataset = datasets.load_dataset('imdb', split='train')

args = TrainingArguments(
    output_dir="./test-galore",
    max_steps=100,
    per_device_train_batch_size=2,
    optim="galore_adamw_layerwise",
    optim_target_modules=[r".*.attn.*", r".*.mlp.*"]
)

model_id = "google/gemma-2b"

config = AutoConfig.from_pretrained(model_id)

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_config(config).to(0)

trainer = trl.SFTTrainer(
    model=model,
    args=args,
    train_dataset=train_dataset,
    dataset_text_field='text',
    max_seq_length=512,
)

trainer.train()

pip install liger-kernel

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,
    use_liger_kernel=True
)

import torch
import datasets
from transformers import TrainingArguments, AutoTokenizer, AutoModelForCausalLM
import trl

train_dataset = datasets.load_dataset('imdb', split='train')

args = TrainingArguments(
    output_dir="./test-lomo",
    max_steps=1000,
    per_device_train_batch_size=4,
    optim="adalomo",
    gradient_checkpointing=True,
    logging_strategy="steps",
    logging_steps=1,
    learning_rate=2e-6,
    save_strategy="no",
    run_name="lomo-imdb",
)

model_id = "google/gemma-2b"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, low_cpu_mem_usage=True).to(0)

trainer = trl.SFTTrainer(
    model=model,
    args=args,
    train_dataset=train_dataset,
    dataset_text_field='text',
    max_seq_length=1024,
)

trainer.train()

import torch
import datasets
from transformers import TrainingArguments, AutoTokenizer, AutoModelForCausalLM, Trainer

# 加载 IMDB 数据集
train_dataset = datasets.load_dataset('imdb', split='train')

# 定义训练参数
args = TrainingArguments(
    output_dir="./test-grokadamw",
    max_steps=1000,
    per_device_train_batch_size=4,
    optim="grokadamw",
    logging_strategy="steps",
    logging_steps=1,
    learning_rate=2e-5,
    save_strategy="no",
    run_name="grokadamw-imdb",
)

# 加载模型和分词器
model_id = "google/gemma-2b"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, low_cpu_mem_usage=True).to(0)

# 初始化 Trainer
trainer = Trainer(
    model=model,
    args=args,
    train_dataset=train_dataset,
)

# 训练模型
trainer.train()

import torch
import datasets
from transformers import TrainingArguments, AutoTokenizer, AutoModelForCausalLM
import trl

train_dataset = datasets.load_dataset('imdb', split='train')

args = TrainingArguments(
    output_dir="./test-schedulefree",
    max_steps=1000,
    per_device_train_batch_size=4,
    optim="schedule_free_adamw",
    gradient_checkpointing=True,
    logging_strategy="steps",
    logging_steps=1,
    learning_rate=2e-6,
    save_strategy="no",
    run_name="sfo-imdb",
)

model_id = "google/gemma-2b"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, low_cpu_mem_usage=True).to(0)

trainer = trl.SFTTrainer(
    model=model, 
    args=args,
    train_dataset=train_dataset,
    dataset_text_field='text',
    max_seq_length=1024,
)

trainer.train()

compute_environment: LOCAL_MACHINE
distributed_type: MULTI_GPU
downcast_bf16: 'no'
gpu_ids: all
machine_rank: 0 #根据节点更改 rank
main_process_ip: 192.168.20.1
main_process_port: 9898
main_training_function: main
mixed_precision: fp16
num_machines: 2
num_processes: 8
rdzv_backend: static
same_network: true
tpu_env: []
tpu_use_cluster: false
tpu_use_sudo: false
use_cpu: false

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