diff --git a/scripts/train.py b/scripts/train.py new file mode 100644 index 00000000..03ffc77b --- /dev/null +++ b/scripts/train.py @@ -0,0 +1,15 @@ +from delphi.train.training import DDP,TrainingConfig, model_initialization, train_loop +from delphi.train.utils import load_config +from argparse import ArgumentParser + + +def main(): + parser = ArgumentParser() + parser.add_argument("--config", type=str, help="Path to the configuration file.") + parser.add_argument("--log_level", type=str, help="Log level to use.") + args = parser.parse_args() + + config = load_config(args.config) + TrainConf = TrainingConfig(config) + model,model_args = model_initialization(config) + train_loop(model, TrainConf) \ No newline at end of file diff --git a/src/delphi/train/training.py b/src/delphi/train/training.py index a906e004..ecbd034e 100644 --- a/src/delphi/train/training.py +++ b/src/delphi/train/training.py @@ -1,216 +1,247 @@ -""" -This training script can be run both on a single gpu in debug mode, -and also in a larger training run with distributed data parallel (ddp). -To run on a single GPU small debug run, example: -$ python -m train.py --compile=False --eval_iters=10 --batch_size=8 - -To run with DDP on 4 gpus on 1 node, example: -$ torchrun --standalone --nproc_per_node=4 train.py - -To run with DDP on 4 gpus across 2 nodes, example: -- Run on the first (master) node with example IP 123.456.123.456: -$ torchrun --nproc_per_node=8 --nnodes=2 --node_rank=0 --master_addr=123.456.123.456 --master_port=1234 train.py -- Run on the worker node: -$ torchrun --nproc_per_node=8 --nnodes=2 --node_rank=1 --master_addr=123.456.123.456 --master_port=1234 train.py -(If your cluster does not have Infiniband interconnect prepend NCCL_IB_DISABLE=1) -""" - -import math -import os +from datetime import datetime +from dataclasses import dataclass +import os +import torch import time from contextlib import nullcontext -from datetime import datetime -from functools import partial +import math +@dataclass +def TrainingConfig(config): + # ----------------------------------------------------------------------------- + # I/O + out_dir: str = "out" + eval_interval: int = 2000 + log_interval:int = 1 + eval_iters:int = 100 + eval_only:bool = False # if True, script exits right after the first eval + always_save_checkpoint:bool = False # if True, always save a checkpoint after each eval + init_from:bool = "scratch" # 'scratch' or 'resume' + # wandb logging + wandb_log:bool = False # disabled by default + wandb_project:str = "llamac" + wandb_run_name:str = "run" + datetime.now().strftime("%Y_%m_%d_%H_%M_%S") + # data + batch_size:int = 128 # if gradient_accumulation_steps > 1, this is the micro-batch size + max_seq_len:int = 256 + vocab_source:str = "llama2" # llama2|custom; use Lllama 2 vocab from Meta, or custom trained + vocab_size:str = 32000 # the Llama 2 tokenizer has 32K tokens + # model + dim:int = 288 + n_layers:int = 6 + n_heads:int = 6 + n_kv_heads:int = 6 + multiple_of:int = 32 + dropout:int = 0.0 + # adamw optimizer + gradient_accumulation_steps:int = 4 # used to simulate larger batch sizes + learning_rate:float = 5e-4 # max learning rate + max_iters:int = 100000 # total number of training iterations + weight_decay:float = 1e-1 + beta1:float = 0.9 + beta2:float = 0.95 + grad_clip:float = 1.0 # clip gradients at this value, or disable if == 0.0 + # learning rate decay settings + decay_lr:bool = True # whether to decay the learning rate + warmup_iters:int = 1000 # how many steps to warm up for + # system + device:str = "cuda" # examples: 'cpu', 'cuda', 'cuda:0', 'cuda:1' etc., or try 'mps' on macbooks + dtype:str = "bfloat16" # float32|bfloat16|float16 + compile:bool = True # use PyTorch 2.0 to compile the model to be faster + # ----------------------------------------------------------------------------- + config_keys = [ + k + for k, v in globals().items() + if not k.startswith("_") and isinstance(v, (int, float, bool, str)) + ] + exec(open("configurator.py").read()) # overrides from command line or config file + config = {k: globals()[k] for k in config_keys} # will be useful for logging -import torch -from torch.distributed import destroy_process_group, init_process_group -from torch.nn.parallel import DistributedDataParallel as DDP -from llama2 import LLaMA2, LLaMA2Args -from llama2c import model_export, Task + # ----------------------------------------------------------------------------- -from shuffle import shuffle_epoch + # fixing some hyperparams to sensible defaults + lr_decay_iters = max_iters # should be ~= max_iters per Chinchilla + min_lr = 0.0 # minimum learning rate, should be ~= learning_rate/10 per Chinchilla -# ----------------------------------------------------------------------------- -# I/O -out_dir = "out" -eval_interval = 2000 -log_interval = 1 -eval_iters = 100 -eval_only = False # if True, script exits right after the first eval -always_save_checkpoint = False # if True, always save a checkpoint after each eval -init_from = "scratch" # 'scratch' or 'resume' -# wandb logging -wandb_log = False # disabled by default -wandb_project = "llamac" -wandb_run_name = "run" + datetime.now().strftime("%Y_%m_%d_%H_%M_%S") -# data -batch_size = 128 # if gradient_accumulation_steps > 1, this is the micro-batch size -max_seq_len = 256 -vocab_source = "llama2" # llama2|custom; use Lllama 2 vocab from Meta, or custom trained -vocab_size = 32000 # the Llama 2 tokenizer has 32K tokens -# model -dim = 288 -n_layers = 6 -n_heads = 6 -n_kv_heads = 6 -multiple_of = 32 -dropout = 0.0 -# adamw optimizer -gradient_accumulation_steps = 4 # used to simulate larger batch sizes -learning_rate = 5e-4 # max learning rate -max_iters = 100000 # total number of training iterations -weight_decay = 1e-1 -beta1 = 0.9 -beta2 = 0.95 -grad_clip = 1.0 # clip gradients at this value, or disable if == 0.0 -# learning rate decay settings -decay_lr = True # whether to decay the learning rate -warmup_iters = 1000 # how many steps to warm up for -# system -device = "cuda" # examples: 'cpu', 'cuda', 'cuda:0', 'cuda:1' etc., or try 'mps' on macbooks -dtype = "bfloat16" # float32|bfloat16|float16 -compile = True # use PyTorch 2.0 to compile the model to be faster -# ----------------------------------------------------------------------------- -config_keys = [ - k - for k, v in globals().items() - if not k.startswith("_") and isinstance(v, (int, float, bool, str)) -] -exec(open("configurator.py").read()) # overrides from command line or config file -config = {k: globals()[k] for k in config_keys} # will be useful for logging -# ----------------------------------------------------------------------------- + # validating checks + assert vocab_source in ["llama2", "custom"] + assert vocab_source == "custom" or vocab_size == 32000, "The vocab from Meta has 32K tokens" -# fixing some hyperparams to sensible defaults -lr_decay_iters = max_iters # should be ~= max_iters per Chinchilla -min_lr = 0.0 # minimum learning rate, should be ~= learning_rate/10 per Chinchilla + # various inits, derived attributes, I/O setup + seed = 1337 + os.makedirs(out_dir, exist_ok=True) -# validating checks -assert vocab_source in ["llama2", "custom"] -assert vocab_source == "custom" or vocab_size == 32000, "The vocab from Meta has 32K tokens" -# various inits, derived attributes, I/O setup -seed = 1337 -ddp = int(os.environ.get("RANK", -1)) != -1 # is this a ddp run? -if ddp: - init_process_group(backend="nccl") - ddp_rank = int(os.environ["RANK"]) - ddp_local_rank = int(os.environ["LOCAL_RANK"]) - ddp_world_size = int(os.environ["WORLD_SIZE"]) - device = f"cuda:{ddp_local_rank}" - torch.cuda.set_device(device) - master_process = ddp_rank == 0 # this process will do logging, checkpointing etc. - seed_offset = ddp_rank # each process gets a different seed - # world_size number of processes will be training simultaneously, so we can scale - # down the desired gradient accumulation iterations per process proportionally - assert gradient_accumulation_steps % ddp_world_size == 0 - gradient_accumulation_steps //= ddp_world_size -else: - # if not ddp, we are running on a single gpu, and one process - master_process = True - seed_offset = 0 - ddp_world_size = 1 -tokens_per_iter = gradient_accumulation_steps * ddp_world_size * batch_size * max_seq_len -if master_process: - print(f"tokens per iteration will be: {tokens_per_iter:,}") - print(f"breaks down as: {gradient_accumulation_steps} grad accum steps * {ddp_world_size} processes * {batch_size} batch size * {max_seq_len} max seq len") +def model_initialization(config): -if master_process: - os.makedirs(out_dir, exist_ok=True) -torch.manual_seed(seed + seed_offset) -torch.backends.cuda.matmul.allow_tf32 = True # allow tf32 on matmul -torch.backends.cudnn.allow_tf32 = True # allow tf32 on cudnn -device_type = "cuda" if "cuda" in device else "cpu" # for later use in torch.autocast -# note: float16 data type will automatically use a GradScaler -ptdtype = {"float32": torch.float32, "bfloat16": torch.bfloat16, "float16": torch.float16}[dtype] -ctx = ( - nullcontext() - if device_type == "cpu" - else torch.amp.autocast(device_type=device_type, dtype=ptdtype) -) + #model + if config["model"] == "llama2": + from delphi.models.llama2 import LLaMA2, LLaMA2Args + model_args = dict( + dim=config["dim"], + n_layers=config["n_layers"], + n_heads=config["n_heads"], + n_kv_heads=config["n_kv_heads"], + vocab_size=config["vocab_size"], + multiple_of=config["multiple_of"], + max_seq_len=config["max_seq_len"], + dropout=config["dropout"], + ) + gptconf = LLaMA2Args(**model_args) + model = LLaMA2(gptconf) + elif config["model"] == "mamba": + from delphi.models.mamba import Mamba, MambaArgs + model_args = dict( + dim=config["dim"], + n_layers=config["n_layers"], + vocab_size=config["vocab_size"], + ) + mambaconf = MambaArgs(**model_args) + model = Mamba(mambaconf) + + if config["init_from"] == "resume": + print(f"Resuming training from {config['out_dir']}") + # resume training from a checkpoint. + ckpt_path = os.path.join(config['out_dir'], "ckpt.pt") + checkpoint = torch.load(ckpt_path, map_location=config['device']) + checkpoint_model_args = checkpoint["model_args"] + # force these config attributes to be equal otherwise we can't even resume training + # the rest of the attributes (e.g. dropout) can stay as desired from command line + for k in ["dim", "n_layers", "n_heads", "n_kv_heads", "vocab_size", "multiple_of", "max_seq_len"]: + model_args[k] = checkpoint_model_args[k] + # create the model + state_dict = checkpoint["model"] + # fix the keys of the state dictionary :( + # honestly no idea how checkpoints sometimes get this prefix, have to debug more + 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) + config.iter_num = checkpoint["iter_num"] + config.best_val_loss = checkpoint["best_val_loss"] + + model.to(config["device"]) + # compile the model + if config["compile"]: + print("compiling the model... (takes a ~minute)") + unoptimized_model = model + model = torch.compile(model) # requires PyTorch 2.0 + return model,model_args + +def train_loop(model, TrainConf): + torch.manual_seed(TrainConf.seed) + torch.backends.cuda.matmul.allow_tf32 = True # allow tf32 on matmul + torch.backends.cudnn.allow_tf32 = True # allow tf32 on cudnn + device_type = "cuda" if "cuda" in TrainConf.device else "cpu" # for later use in torch.autocast + # note: float16 data type will automatically use a GradScaler + ptdtype = {"float32": torch.float32, "bfloat16": torch.bfloat16, "float16": torch.float16}[TrainConf.dtype] + ctx = ( + nullcontext() + if device_type == "cpu" + else torch.amp.autocast(device_type=device_type, dtype=ptdtype) + ) + scaler = torch.cuda.amp.GradScaler(enabled=(TrainConf.dtype == "float16")) -# task-specific setup -iter_batches = partial( - Task.iter_batches, - batch_size=batch_size, - max_seq_len=max_seq_len, - vocab_size=vocab_size, - vocab_source=vocab_source, - device=device, - num_workers=0, - seed=seed -) + # optimizer + optimizer = model.configure_optimizers(TrainConf.weight_decay, TrainConf.learning_rate, (TrainConf.beta1, TrainConf.beta2), device_type) + if TrainConf.init_from == "resume" and "optimizer" in checkpoint: + optimizer.load_state_dict(checkpoint["optimizer"]) + checkpoint = None # free up memory -# init these up here, can override if init_from='resume' (i.e. from a checkpoint) -iter_num = 0 -best_val_loss = 1e9 + if TrainConf.wandb_log: + import wandb + wandb.init(project=TrainConf.wandb_project, name=TrainConf.wandb_run_name, config=TrainConf.config) -# model init -model_args = dict( - dim=dim, - n_layers=n_layers, - n_heads=n_heads, - n_kv_heads=n_kv_heads, - vocab_size=vocab_size, - multiple_of=multiple_of, - max_seq_len=max_seq_len, - dropout=dropout, -) # start with model_args from command line -if init_from == "scratch": - # init a new model from scratch - print("Initializing a new model from scratch") - gptconf = LLaMA2Args(**model_args) - model = LLaMA2(gptconf) -elif init_from == "resume": - print(f"Resuming training from {out_dir}") - # resume training from a checkpoint. - ckpt_path = os.path.join(out_dir, "ckpt.pt") - checkpoint = torch.load(ckpt_path, map_location=device) - checkpoint_model_args = checkpoint["model_args"] - # force these config attributes to be equal otherwise we can't even resume training - # the rest of the attributes (e.g. dropout) can stay as desired from command line - for k in ["dim", "n_layers", "n_heads", "n_kv_heads", "vocab_size", "multiple_of", "max_seq_len"]: - model_args[k] = checkpoint_model_args[k] - # create the model - gptconf = LLaMA2Args(**model_args) - model = LLaMA2(gptconf) - state_dict = checkpoint["model"] - # fix the keys of the state dictionary :( - # honestly no idea how checkpoints sometimes get this prefix, have to debug more - 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) - iter_num = checkpoint["iter_num"] - best_val_loss = checkpoint["best_val_loss"] -model.to(device) + train_batch_iter = TrainConf.iter_batches(split="train") + X, Y = next(train_batch_iter) # fetch the very first batch + t0 = time.time() + local_iter_num = 0 # number of iterations in the lifetime of this process + raw_model = model # unwrap DDP container if needed + running_mfu = -1.0 + while True: + # determine and set the learning rate for this iteration + lr = get_lr(iter_num,TrainConf) if TrainConf.decay_lr else TrainConf.learning_rate + for param_group in optimizer.param_groups: + param_group["lr"] = lr -# initialize a GradScaler. If enabled=False scaler is a no-op -scaler = torch.cuda.amp.GradScaler(enabled=(dtype == "float16")) + # evaluate the loss on train/val sets and write checkpoints + if iter_num % TrainConf.eval_interval == 0: + losses = estimate_loss() + print(f"step {iter_num}: train loss {losses['train']:.4f}, val loss {losses['val']:.4f}") + if TrainConf.wandb_log: + try: + wandb.log( + { + "iter": iter_num, + "tokens": iter_num * TrainConf.tokens_per_iter, + "loss/train": losses["train"], + "loss/val": losses["val"], + "lr": lr, + "mfu": running_mfu * 100, # convert to percentage + }, step = iter_num + ) + except Exception as e: + print(f"logging to wandb failed: {e}") + if losses["val"] < best_val_loss or TrainConf.always_save_checkpoint: + best_val_loss = losses["val"] + if iter_num > 0: + checkpoint = { + "model": raw_model.state_dict(), + "optimizer": optimizer.state_dict(), + "model_args": model_args, + "iter_num": iter_num, + "best_val_loss": best_val_loss, + "config": config, + } + print(f"saving checkpoint to {out_dir}") + torch.save(checkpoint, os.path.join(out_dir, "ckpt.pt")) + model_export(raw_model, os.path.join(out_dir, "model.bin"), version=0) + if iter_num == 0 and eval_only: + break -# optimizer -optimizer = model.configure_optimizers(weight_decay, learning_rate, (beta1, beta2), device_type) -if init_from == "resume" and "optimizer" in checkpoint: - optimizer.load_state_dict(checkpoint["optimizer"]) -checkpoint = None # free up memory + # forward backward update, with optional gradient accumulation to simulate larger batch size + # and using the GradScaler if data type is float16 + for micro_step in range(gradient_accumulation_steps): + with ctx: + logits = model(X, Y) + loss = raw_model.last_loss + loss = loss / gradient_accumulation_steps + # immediately async prefetch next batch while model is doing the forward pass on the GPU + X, Y = next(train_batch_iter) + # backward pass, with gradient scaling if training in fp16 + scaler.scale(loss).backward() + # clip the gradient + if grad_clip != 0.0: + scaler.unscale_(optimizer) + torch.nn.utils.clip_grad_norm_(model.parameters(), grad_clip) + # step the optimizer and scaler if training in fp16 + scaler.step(optimizer) + scaler.update() + # flush the gradients as soon as we can, no need for this memory anymore + optimizer.zero_grad(set_to_none=True) -# compile the model -if compile: - print("compiling the model... (takes a ~minute)") - unoptimized_model = model - model = torch.compile(model) # requires PyTorch 2.0 + # timing and logging + t1 = time.time() + dt = t1 - t0 + t0 = t1 + if iter_num % log_interval == 0: + # get loss as float, scale up due to the divide above. note: this is a CPU-GPU sync point + lossf = loss.item() * gradient_accumulation_steps + if local_iter_num >= 5: # let the training loop settle a bit + mfu = raw_model.estimate_mfu(batch_size * gradient_accumulation_steps, dt) + running_mfu = mfu if running_mfu == -1.0 else 0.9 * running_mfu + 0.1 * mfu + print( + f"{iter_num} | loss {lossf:.4f} | lr {lr:e} | {dt*1000:.2f}ms | mfu {running_mfu*100:.2f}%" + ) + iter_num += 1 + local_iter_num += 1 -# wrap model into DDP container -if ddp: - # Ignore the `freqs_cis` buffer so that DDP does not broadcast it at - # construction time since NCCL does not support `ComplexFloat` - prefix = "_orig_mod." if compile else "" - model._ddp_params_and_buffers_to_ignore = {prefix + "freqs_cis"} - model = DDP(model, device_ids=[ddp_local_rank]) + # termination conditions + if iter_num > max_iters: + break -# helps estimate an arbitrarily accurate loss over either split using many batches @torch.no_grad() def estimate_loss(): out = {} @@ -229,118 +260,17 @@ def estimate_loss(): return out # learning rate decay scheduler (cosine with warmup) -def get_lr(it): +def get_lr(it,TrainConf): # 1) linear warmup for warmup_iters steps - if it < warmup_iters: - return learning_rate * it / warmup_iters + if it < TrainConf.warmup_iters: + return TrainConf.learning_rate * it / TrainConf.warmup_iters # 2) if it > lr_decay_iters, return min learning rate - if it > lr_decay_iters: - return min_lr + if it > TrainConf.lr_decay_iters: + return TrainConf.min_lr # 3) in between, use cosine decay down to min learning rate - decay_ratio = (it - warmup_iters) / (lr_decay_iters - warmup_iters) + decay_ratio = (it - TrainConf.warmup_iters) / (TrainConf.lr_decay_iters - TrainConf.warmup_iters) assert 0 <= decay_ratio <= 1 coeff = 0.5 * (1.0 + math.cos(math.pi * decay_ratio)) # coeff ranges 0..1 - return min_lr + coeff * (learning_rate - min_lr) - -# logging -if wandb_log and master_process: - import wandb - wandb.init(project=wandb_project, name=wandb_run_name, config=config) - -# training loop -train_batch_iter = iter_batches(split="train") -X, Y = next(train_batch_iter) # fetch the very first batch -t0 = time.time() -local_iter_num = 0 # number of iterations in the lifetime of this process -raw_model = model.module if ddp else model # unwrap DDP container if needed -running_mfu = -1.0 -while True: - # determine and set the learning rate for this iteration - lr = get_lr(iter_num) if decay_lr else learning_rate - for param_group in optimizer.param_groups: - param_group["lr"] = lr - - # evaluate the loss on train/val sets and write checkpoints - if iter_num % eval_interval == 0 and master_process: - losses = estimate_loss() - print(f"step {iter_num}: train loss {losses['train']:.4f}, val loss {losses['val']:.4f}") - if wandb_log: - try: - wandb.log( - { - "iter": iter_num, - "tokens": iter_num * tokens_per_iter, - "loss/train": losses["train"], - "loss/val": losses["val"], - "lr": lr, - "mfu": running_mfu * 100, # convert to percentage - }, step = iter_num - ) - except Exception as e: - print(f"logging to wandb failed: {e}") - if losses["val"] < best_val_loss or always_save_checkpoint: - best_val_loss = losses["val"] - if iter_num > 0: - checkpoint = { - "model": raw_model.state_dict(), - "optimizer": optimizer.state_dict(), - "model_args": model_args, - "iter_num": iter_num, - "best_val_loss": best_val_loss, - "config": config, - } - print(f"saving checkpoint to {out_dir}") - torch.save(checkpoint, os.path.join(out_dir, "ckpt.pt")) - model_export(raw_model, os.path.join(out_dir, "model.bin"), version=0) - if iter_num == 0 and eval_only: - break - - # forward backward update, with optional gradient accumulation to simulate larger batch size - # and using the GradScaler if data type is float16 - for micro_step in range(gradient_accumulation_steps): - if ddp: - # in DDP training we only need to sync gradients at the last micro step. - # the official way to do this is with model.no_sync() context manager, but - # I really dislike that this bloats the code and forces us to repeat code - # looking at the source of that context manager, it just toggles this variable - model.require_backward_grad_sync = micro_step == gradient_accumulation_steps - 1 - with ctx: - logits = model(X, Y) - loss = raw_model.last_loss - loss = loss / gradient_accumulation_steps - # immediately async prefetch next batch while model is doing the forward pass on the GPU - X, Y = next(train_batch_iter) - # backward pass, with gradient scaling if training in fp16 - scaler.scale(loss).backward() - # clip the gradient - if grad_clip != 0.0: - scaler.unscale_(optimizer) - torch.nn.utils.clip_grad_norm_(model.parameters(), grad_clip) - # step the optimizer and scaler if training in fp16 - scaler.step(optimizer) - scaler.update() - # flush the gradients as soon as we can, no need for this memory anymore - optimizer.zero_grad(set_to_none=True) - - # timing and logging - t1 = time.time() - dt = t1 - t0 - t0 = t1 - if iter_num % log_interval == 0 and master_process: - # get loss as float, scale up due to the divide above. note: this is a CPU-GPU sync point - lossf = loss.item() * gradient_accumulation_steps - if local_iter_num >= 5: # let the training loop settle a bit - mfu = raw_model.estimate_mfu(batch_size * gradient_accumulation_steps, dt) - running_mfu = mfu if running_mfu == -1.0 else 0.9 * running_mfu + 0.1 * mfu - print( - f"{iter_num} | loss {lossf:.4f} | lr {lr:e} | {dt*1000:.2f}ms | mfu {running_mfu*100:.2f}%" - ) - iter_num += 1 - local_iter_num += 1 + return TrainConf.min_lr + coeff * (TrainConf.learning_rate - TrainConf.min_lr) - # termination conditions - if iter_num > max_iters: - break -if ddp: - destroy_process_group() diff --git a/src/delphi/train/training_old.py b/src/delphi/train/training_old.py new file mode 100644 index 00000000..d2d5e0e1 --- /dev/null +++ b/src/delphi/train/training_old.py @@ -0,0 +1,353 @@ +""" +This training script can be run both on a single gpu in debug mode, +and also in a larger training run with distributed data parallel (ddp). + +To run on a single GPU small debug run, example: +$ python -m train.py --compile=False --eval_iters=10 --batch_size=8 + +To run with DDP on 4 gpus on 1 node, example: +$ torchrun --standalone --nproc_per_node=4 train.py + +To run with DDP on 4 gpus across 2 nodes, example: +- Run on the first (master) node with example IP 123.456.123.456: +$ torchrun --nproc_per_node=8 --nnodes=2 --node_rank=0 --master_addr=123.456.123.456 --master_port=1234 train.py +- Run on the worker node: +$ torchrun --nproc_per_node=8 --nnodes=2 --node_rank=1 --master_addr=123.456.123.456 --master_port=1234 train.py +(If your cluster does not have Infiniband interconnect prepend NCCL_IB_DISABLE=1) +""" + +import math +import os +import time +from contextlib import nullcontext +from datetime import datetime +from functools import partial + +import torch +from torch.distributed import destroy_process_group, init_process_group +from torch.nn.parallel import DistributedDataParallel as DDP + +from llama2 import LLaMA2, LLaMA2Args +from llama2c import model_export, Task + +from shuffle import shuffle_epoch + +# ----------------------------------------------------------------------------- +# I/O +out_dir = "out" +eval_interval = 2000 +log_interval = 1 +eval_iters = 100 +eval_only = False # if True, script exits right after the first eval +always_save_checkpoint = False # if True, always save a checkpoint after each eval +init_from = "scratch" # 'scratch' or 'resume' +# wandb logging +wandb_log = False # disabled by default +wandb_project = "llamac" +wandb_run_name = "run" + datetime.now().strftime("%Y_%m_%d_%H_%M_%S") +# data +batch_size = 128 # if gradient_accumulation_steps > 1, this is the micro-batch size +max_seq_len = 256 +vocab_source = "llama2" # llama2|custom; use Lllama 2 vocab from Meta, or custom trained +vocab_size = 32000 # the Llama 2 tokenizer has 32K tokens +# model +dim = 288 +n_layers = 6 +n_heads = 6 +n_kv_heads = 6 +multiple_of = 32 +dropout = 0.0 +# adamw optimizer +gradient_accumulation_steps = 4 # used to simulate larger batch sizes +learning_rate = 5e-4 # max learning rate +max_iters = 100000 # total number of training iterations +weight_decay = 1e-1 +beta1 = 0.9 +beta2 = 0.95 +grad_clip = 1.0 # clip gradients at this value, or disable if == 0.0 +# learning rate decay settings +decay_lr = True # whether to decay the learning rate +warmup_iters = 1000 # how many steps to warm up for +# system +device = "cuda" # examples: 'cpu', 'cuda', 'cuda:0', 'cuda:1' etc., or try 'mps' on macbooks +dtype = "bfloat16" # float32|bfloat16|float16 +compile = True # use PyTorch 2.0 to compile the model to be faster +# ----------------------------------------------------------------------------- +config_keys = [ + k + for k, v in globals().items() + if not k.startswith("_") and isinstance(v, (int, float, bool, str)) +] +exec(open("configurator.py").read()) # overrides from command line or config file +config = {k: globals()[k] for k in config_keys} # will be useful for logging + + +# ----------------------------------------------------------------------------- + +# fixing some hyperparams to sensible defaults +lr_decay_iters = max_iters # should be ~= max_iters per Chinchilla +min_lr = 0.0 # minimum learning rate, should be ~= learning_rate/10 per Chinchilla + +# validating checks +assert vocab_source in ["llama2", "custom"] +assert vocab_source == "custom" or vocab_size == 32000, "The vocab from Meta has 32K tokens" + +# various inits, derived attributes, I/O setup +seed = 1337 +ddp = int(os.environ.get("RANK", -1)) != -1 # is this a ddp run? +if ddp: + init_process_group(backend="nccl") + ddp_rank = int(os.environ["RANK"]) + ddp_local_rank = int(os.environ["LOCAL_RANK"]) + ddp_world_size = int(os.environ["WORLD_SIZE"]) + device = f"cuda:{ddp_local_rank}" + torch.cuda.set_device(device) + master_process = ddp_rank == 0 # this process will do logging, checkpointing etc. + seed_offset = ddp_rank # each process gets a different seed + # world_size number of processes will be training simultaneously, so we can scale + # down the desired gradient accumulation iterations per process proportionally + assert gradient_accumulation_steps % ddp_world_size == 0 + gradient_accumulation_steps //= ddp_world_size +else: + # if not ddp, we are running on a single gpu, and one process + master_process = True + seed_offset = 0 + ddp_world_size = 1 +tokens_per_iter = gradient_accumulation_steps * ddp_world_size * batch_size * max_seq_len +if master_process: + print(f"tokens per iteration will be: {tokens_per_iter:,}") + print(f"breaks down as: {gradient_accumulation_steps} grad accum steps * {ddp_world_size} processes * {batch_size} batch size * {max_seq_len} max seq len") + +if master_process: + os.makedirs(out_dir, exist_ok=True) +torch.manual_seed(seed + seed_offset) +torch.backends.cuda.matmul.allow_tf32 = True # allow tf32 on matmul +torch.backends.cudnn.allow_tf32 = True # allow tf32 on cudnn +device_type = "cuda" if "cuda" in device else "cpu" # for later use in torch.autocast +# note: float16 data type will automatically use a GradScaler +ptdtype = {"float32": torch.float32, "bfloat16": torch.bfloat16, "float16": torch.float16}[dtype] +ctx = ( + nullcontext() + if device_type == "cpu" + else torch.amp.autocast(device_type=device_type, dtype=ptdtype) +) + +# task-specific setup +iter_batches = partial( + Task.iter_batches, + batch_size=batch_size, + max_seq_len=max_seq_len, + vocab_size=vocab_size, + vocab_source=vocab_source, + device=device, + num_workers=0, + seed=seed +) + +# init these up here, can override if init_from='resume' (i.e. from a checkpoint) +iter_num = 0 +best_val_loss = 1e9 + +# model init +model_args = dict( + dim=dim, + n_layers=n_layers, + n_heads=n_heads, + n_kv_heads=n_kv_heads, + vocab_size=vocab_size, + multiple_of=multiple_of, + max_seq_len=max_seq_len, + dropout=dropout, +) # start with model_args from command line +if init_from == "scratch": + # init a new model from scratch + print("Initializing a new model from scratch") + gptconf = LLaMA2Args(**model_args) + model = LLaMA2(gptconf) +elif init_from == "resume": + print(f"Resuming training from {out_dir}") + # resume training from a checkpoint. + ckpt_path = os.path.join(out_dir, "ckpt.pt") + checkpoint = torch.load(ckpt_path, map_location=device) + checkpoint_model_args = checkpoint["model_args"] + # force these config attributes to be equal otherwise we can't even resume training + # the rest of the attributes (e.g. dropout) can stay as desired from command line + for k in ["dim", "n_layers", "n_heads", "n_kv_heads", "vocab_size", "multiple_of", "max_seq_len"]: + model_args[k] = checkpoint_model_args[k] + # create the model + gptconf = LLaMA2Args(**model_args) + model = LLaMA2(gptconf) + state_dict = checkpoint["model"] + # fix the keys of the state dictionary :( + # honestly no idea how checkpoints sometimes get this prefix, have to debug more + 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) + iter_num = checkpoint["iter_num"] + best_val_loss = checkpoint["best_val_loss"] +model.to(device) + +# initialize a GradScaler. If enabled=False scaler is a no-op +scaler = torch.cuda.amp.GradScaler(enabled=(dtype == "float16")) + +# optimizer +optimizer = model.configure_optimizers(weight_decay, learning_rate, (beta1, beta2), device_type) +if init_from == "resume" and "optimizer" in checkpoint: + optimizer.load_state_dict(checkpoint["optimizer"]) +checkpoint = None # free up memory + +# compile the model +if compile: + print("compiling the model... (takes a ~minute)") + unoptimized_model = model + model = torch.compile(model) # requires PyTorch 2.0 + +# wrap model into DDP container +if ddp: + # Ignore the `freqs_cis` buffer so that DDP does not broadcast it at + # construction time since NCCL does not support `ComplexFloat` + prefix = "_orig_mod." if compile else "" + model._ddp_params_and_buffers_to_ignore = {prefix + "freqs_cis"} + model = DDP(model, device_ids=[ddp_local_rank]) + +# helps estimate an arbitrarily accurate loss over either split using many batches +@torch.no_grad() +def estimate_loss(): + out = {} + model.eval() + for split in ["train", "val"]: + batch_iter = iter_batches(split=split) + losses = torch.zeros(eval_iters) # keep on CPU + for k in range(eval_iters): + X, Y = next(batch_iter) + with ctx: + logits = model(X, Y) + loss = raw_model.last_loss + losses[k] = loss.item() + out[split] = losses.mean() + model.train() + return out + +# learning rate decay scheduler (cosine with warmup) +def get_lr(it): + # 1) linear warmup for warmup_iters steps + if it < warmup_iters: + return learning_rate * it / warmup_iters + # 2) if it > lr_decay_iters, return min learning rate + if it > lr_decay_iters: + return min_lr + # 3) in between, use cosine decay down to min learning rate + 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)) # coeff ranges 0..1 + return min_lr + coeff * (learning_rate - min_lr) + + + +# logging +if wandb_log and master_process: + import wandb + wandb.init(project=wandb_project, name=wandb_run_name, config=config) + + + + +# training loop +train_batch_iter = iter_batches(split="train") +X, Y = next(train_batch_iter) # fetch the very first batch +t0 = time.time() +local_iter_num = 0 # number of iterations in the lifetime of this process +raw_model = model.module if ddp else model # unwrap DDP container if needed +running_mfu = -1.0 +while True: + # determine and set the learning rate for this iteration + lr = get_lr(iter_num) if decay_lr else learning_rate + for param_group in optimizer.param_groups: + param_group["lr"] = lr + + # evaluate the loss on train/val sets and write checkpoints + if iter_num % eval_interval == 0 and master_process: + losses = estimate_loss() + print(f"step {iter_num}: train loss {losses['train']:.4f}, val loss {losses['val']:.4f}") + if wandb_log: + try: + wandb.log( + { + "iter": iter_num, + "tokens": iter_num * tokens_per_iter, + "loss/train": losses["train"], + "loss/val": losses["val"], + "lr": lr, + "mfu": running_mfu * 100, # convert to percentage + }, step = iter_num + ) + except Exception as e: + print(f"logging to wandb failed: {e}") + if losses["val"] < best_val_loss or always_save_checkpoint: + best_val_loss = losses["val"] + if iter_num > 0: + checkpoint = { + "model": raw_model.state_dict(), + "optimizer": optimizer.state_dict(), + "model_args": model_args, + "iter_num": iter_num, + "best_val_loss": best_val_loss, + "config": config, + } + print(f"saving checkpoint to {out_dir}") + torch.save(checkpoint, os.path.join(out_dir, "ckpt.pt")) + model_export(raw_model, os.path.join(out_dir, "model.bin"), version=0) + if iter_num == 0 and eval_only: + break + + # forward backward update, with optional gradient accumulation to simulate larger batch size + # and using the GradScaler if data type is float16 + for micro_step in range(gradient_accumulation_steps): + if ddp: + # in DDP training we only need to sync gradients at the last micro step. + # the official way to do this is with model.no_sync() context manager, but + # I really dislike that this bloats the code and forces us to repeat code + # looking at the source of that context manager, it just toggles this variable + model.require_backward_grad_sync = micro_step == gradient_accumulation_steps - 1 + with ctx: + logits = model(X, Y) + loss = raw_model.last_loss + loss = loss / gradient_accumulation_steps + # immediately async prefetch next batch while model is doing the forward pass on the GPU + X, Y = next(train_batch_iter) + # backward pass, with gradient scaling if training in fp16 + scaler.scale(loss).backward() + # clip the gradient + if grad_clip != 0.0: + scaler.unscale_(optimizer) + torch.nn.utils.clip_grad_norm_(model.parameters(), grad_clip) + # step the optimizer and scaler if training in fp16 + scaler.step(optimizer) + scaler.update() + # flush the gradients as soon as we can, no need for this memory anymore + optimizer.zero_grad(set_to_none=True) + + # timing and logging + t1 = time.time() + dt = t1 - t0 + t0 = t1 + if iter_num % log_interval == 0 and master_process: + # get loss as float, scale up due to the divide above. note: this is a CPU-GPU sync point + lossf = loss.item() * gradient_accumulation_steps + if local_iter_num >= 5: # let the training loop settle a bit + mfu = raw_model.estimate_mfu(batch_size * gradient_accumulation_steps, dt) + running_mfu = mfu if running_mfu == -1.0 else 0.9 * running_mfu + 0.1 * mfu + print( + f"{iter_num} | loss {lossf:.4f} | lr {lr:e} | {dt*1000:.2f}ms | mfu {running_mfu*100:.2f}%" + ) + iter_num += 1 + local_iter_num += 1 + + # termination conditions + if iter_num > max_iters: + break + +if ddp: + destroy_process_group() diff --git a/src/delphi/train/utils.py b/src/delphi/train/utils.py new file mode 100644 index 00000000..73c3e9e6 --- /dev/null +++ b/src/delphi/train/utils.py @@ -0,0 +1,6 @@ +import json + +def load_config(config_path): + with open(config_path, 'r') as file: + return json.load(file) + \ No newline at end of file