lollms/train/train.py

233 lines
8.8 KiB
Python
Raw Normal View History

Create README Create README.md upgraded readme upgraded upgraded Added an example of a console chat Upgraded code upgraded submodule upgraded example upgraded console application Added some logo and readme upgraded upgraded updated updated changed logo upgraded upgrade upgraded upgraded upgraded version Added console app upgraded code and service information changed documentation title upgraded code updated zoo Upgraded logo upgradded Update server_endpoints.md Update README.md Update server_endpoints.md Enhanced code enhanced work + added training fixed error in README upgraded readme Fixed console problem enhanced code Added reference to models upgraded version Update README.md upgraded binding Update README.md enhanced server upgraded console and server upgraded tool upgraded upgraded Upgraded to new Version enhanced updated personalities zoo personalities_zoo upgraded readme Possibility to send files to personalities Possibility to send files to personalities upgraded code bugfix updated upgraded upgraded console updated readme version upgrade Update README.md Added menu build at startup change upgraded code now you select a personality of not selected upgraded upgraded documentation upgraded documentation updated Upgraded bugfix now you can build custom personalities updated. now we can use other personalities Bugfix added return changed colors added protection added back to personality installation bugfix typo fixed autogptq fixed autogptq gptq upgraded gptq changed version upgraded console typo Added send file updated send file upgraded personality upgraded image analysis tool updated upgraded version upgraded tool updated gpt4all is now working version update upgraded naming scheme hapen Upgraded path data upgraded version updated upgraded version upgraded install procedures personal path can be changed online upgraded chatgpt upgraded upgraded updated version bugfix upgraded personalities upgraded version enhanced enhanced update bugfix version update Added reset functionality Added settings upgraded enhanced library upgraded models Upgraded upgraded rebased upgraded code fixed gpt4all updated version
2023-06-02 10:46:41 +00:00
import os
from transformers import AutoModelForCausalLM, AutoTokenizer, get_scheduler, LlamaForCausalLM
import torch
from torch.optim import AdamW
from argparse import ArgumentParser
from read import read_config
from accelerate import Accelerator
from accelerate.utils import DummyScheduler, DummyOptim, set_seed
from peft import get_peft_model, LoraConfig, TaskType
from data import load_data
from torchmetrics import MeanMetric
from tqdm import tqdm
import wandb
torch.backends.cuda.matmul.allow_tf32 = True
def format_metrics(metrics, split, prefix=""):
log = f"[{split}]" + prefix
log += " ".join([f"{key}: {value:.4f}" for key, value in metrics.items()])
return log
def evaluate(model, val_dataloader):
model.eval()
val_loss = MeanMetric(nan_strategy="error").to(model.device)
with torch.no_grad():
for batch in tqdm(val_dataloader):
loss = model(**batch).loss
loss_values = accelerator.gather_for_metrics({"loss": loss.detach()})
val_loss.update(loss_values["loss"])
return val_loss
def train(accelerator, config):
set_seed(config['seed'])
accelerator.print(config)
accelerator.print(f"Using {accelerator.num_processes} GPUs")
tokenizer = AutoTokenizer.from_pretrained(config['tokenizer_name'], model_max_length=config['max_length'])
# if no pad token, set it to eos
if tokenizer.pad_token is None:
tokenizer.pad_token = tokenizer.eos_token
with accelerator.main_process_first():
train_dataloader, val_dataloader = load_data(config, tokenizer)
checkpoint = config["gradient_checkpointing"]
model = AutoModelForCausalLM.from_pretrained(config["model_name"],
use_cache=False if checkpoint else True,
trust_remote_code=True)
if checkpoint:
model.gradient_checkpointing_enable()
if config["lora"]:
peft_config = LoraConfig(
# should R be configurable?
task_type=TaskType.CAUSAL_LM, inference_mode=False, r=8, lora_alpha=32, lora_dropout=0.1
)
model = get_peft_model(model, peft_config)
model.print_trainable_parameters()
optimizer_cls = (
AdamW
if accelerator.state.deepspeed_plugin is None
or "optimizer" not in accelerator.state.deepspeed_plugin.deepspeed_config
else DummyOptim
)
# karpathy doesn't decay embeddding, maybe we should exclude
# https://github.com/karpathy/minGPT/commit/bbbdac74fa9b2e55574d70056163ffbae42310c1#diff-2075fa9c224b395be5bda85544dd36572b59c76c54562819eadadbf268602834R157s
optimizer = optimizer_cls(model.parameters(), lr=config["lr"], weight_decay=config["weight_decay"])
if accelerator.state.deepspeed_plugin is not None:
gradient_accumulation_steps = accelerator.state.deepspeed_plugin.deepspeed_config[
"gradient_accumulation_steps"
]
# decay to min_lr instead of 0
lr_ratio = config["min_lr"] / config["lr"]
accelerator.print(f"Len of train_dataloader: {len(train_dataloader)}")
total_num_steps = (len(train_dataloader) / gradient_accumulation_steps) * config["num_epochs"]
# instead of decaying to zero, decay to ratio of min_lr / lr
total_num_steps += int(total_num_steps * lr_ratio) + config["warmup_steps"]
accelerator.print(f"Total training steps: {total_num_steps}")
# Creates Dummy Scheduler if `scheduler` was specified in the config file else creates `args.lr_scheduler_type` Scheduler
if (
accelerator.state.deepspeed_plugin is None
or "scheduler" not in accelerator.state.deepspeed_plugin.deepspeed_config
):
scheduler = get_scheduler(
name="cosine",
optimizer=optimizer,
num_warmup_steps=config["warmup_steps"] * accelerator.num_processes,
num_training_steps=total_num_steps,
)
else:
scheduler = DummyScheduler(
optimizer, total_num_steps=config["warmup_steps"], warmup_num_steps=config["warmup_steps"]
)
model, optimizer, train_dataloader, val_dataloader, scheduler = accelerator.prepare(
model, optimizer, train_dataloader, val_dataloader, scheduler
)
# setup for saving training states in case preemption
accelerator.register_for_checkpointing(scheduler)
if config["checkpoint"]:
accelerator.load_state(config["checkpoint"])
accelerator.print(f"Resumed from checkpoint: {config['checkpoint']}")
path = os.path.basename(config["train_args"]["resume_from_checkpoint"])
training_difference = os.path.splitext(path)[0]
resume_step = int(training_difference.replace("step_", ""))
accelerator.skip_first_batches(train_dataloader, resume_step)
accelerator.print(f"Resuming from step {resume_step}")
# log gradients
if accelerator.is_main_process and config["wandb"]:
wandb.watch(model, log_freq=config["log_grads_every"], log="all")
for epoch in range(config["num_epochs"]):
train_loss = MeanMetric(nan_strategy="error").to(model.device)
for step, batch in enumerate(tqdm(train_dataloader)):
model.train()
outputs = model(**batch)
loss = outputs.loss
# gather loss before backprop in case of gradient accumulation
loss_values = accelerator.gather_for_metrics({"loss": loss.detach().float()})
train_loss.update(loss_values["loss"])
loss = loss / gradient_accumulation_steps
accelerator.backward(loss)
# get gradient norm of all params
# log LR in case something weird happens
if step > 0 and step % (config["eval_every"] // 10) == 0:
if config["wandb"]:
curr_step = step + epoch * len(train_dataloader)
accelerator.log({"lr": scheduler.get_last_lr()[0]}, step=curr_step)
if (step + 1) % gradient_accumulation_steps == 0 or step == len(train_dataloader) - 1:
optimizer.step()
scheduler.step()
optimizer.zero_grad()
if step > 0 and step % config["save_every"] == 0:
curr_step = step + epoch * len(train_dataloader)
accelerator.save_state(f"{config['output_dir']}/step_{curr_step}")
if step > 0 and (step % config["eval_every"] == 0 or step == len(train_dataloader) - 1):
val_loss = evaluate(model, val_dataloader)
log_train = {
"train_loss": train_loss.compute()
}
log_val = {
"val_loss": val_loss.compute()
}
if config["wandb"]:
curr_step = step + epoch * len(train_dataloader)
accelerator.log({**log_train, **log_val}, step=curr_step)
accelerator.print(f"Current LR: {scheduler.get_last_lr()[0]}")
accelerator.print(format_metrics(log_train, "train", f" step {step} "))
accelerator.print(format_metrics(log_val, "val", f" step {step} "))
train_loss.reset()
accelerator.print(f"Epoch {epoch} finished")
accelerator.print(f"Pushing to HF hub")
accelerator.wait_for_everyone()
unwrapped_model = accelerator.unwrap_model(model)
try:
if accelerator.is_main_process:
unwrapped_model.push_to_hub(config["save_name"] + f"-epoch_{epoch}", private=True)
except Exception as e:
accelerator.print(e)
accelerator.print(f"Failed to push to hub")
unwrapped_model.save_pretrained(
f"{config['output_dir']}/epoch_{epoch}",
is_main_process=accelerator.is_main_process,
save_function=accelerator.save,
state_dict=accelerator.get_state_dict(model),
)
accelerator.wait_for_everyone()
unwrapped_model = accelerator.unwrap_model(model)
unwrapped_model.save_pretrained(
f"{config['output_dir']}/final",
is_main_process=accelerator.is_main_process,
save_function=accelerator.save,
state_dict=accelerator.get_state_dict(model),
)
accelerator.end_training()
if __name__ == "__main__":
# parse arguments by reading in a config
parser = ArgumentParser()
parser.add_argument("--config", type=str, default="config.yaml")
args = parser.parse_args()
config = read_config(args.config)
if config["wandb"]:
accelerator = Accelerator(log_with="wandb")
accelerator.init_trackers(
project_name=config["wandb_project_name"],
config=config,
init_kwargs={"wandb": {"entity": config["wandb_entity"]}},
)
else:
accelerator = Accelerator()
train(accelerator, config=config)