File size: 7,058 Bytes
7cbb55b | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 | """
Training script for MHD64 dataset using neural operators.
This script trains a neural operator on the 3D magnetohydrodynamics (MHD)
dataset with 64x64x64 resolution. The model learns to predict the next time step
in the MHD simulation, supporting both next-step prediction and autoregressive
evaluation modes.
"""
import sys
from configmypy import ConfigPipeline, YamlConfig, ArgparseConfig
from pathlib import Path
import torch
from torch.utils.data import DataLoader, DistributedSampler
import wandb
from neuralop import H1Loss, LpLoss, Trainer, get_model
from neuralop.data.datasets.the_well_dataset import MHD64Dataset
from neuralop.data.transforms.data_processors import MGPatchingDataProcessor
from neuralop.training import setup, AdamW
from neuralop.mpu.comm import get_local_rank
from neuralop.utils import get_wandb_api_key, count_model_params
# Configuration setup
config_name = "default"
from zencfg import make_config_from_cli
import sys
sys.path.insert(0, "../")
from config.the_well.mhd_64_config import Default
config = make_config_from_cli(Default)
config = config.to_dict()
# Distributed training setup, if enabled
device, is_logger = setup(config)
# Set up WandB logging
wandb_args = None
if config.wandb.log and is_logger:
wandb.login(key=get_wandb_api_key())
if config.wandb.name:
wandb_name = config.wandb.name
else:
wandb_name = "_".join(
f"{var}"
for var in [
config.model.model_arch,
config.model.n_layers,
config.model.n_modes,
config.model.hidden_channels,
]
)
wandb_args = dict(
config=config,
name=wandb_name,
group=config.wandb.group,
project=config.wandb.project,
entity=config.wandb.entity,
)
if config.wandb.sweep:
for key in wandb.config.keys():
config.params[key] = wandb.config[key]
wandb.init(**wandb_args)
# Make sure we only print information when needed
config.verbose = config.verbose and is_logger
# Print configuration details
if config.verbose and is_logger:
print(f"##### CONFIG #####\n\n{config}\n")
sys.stdout.flush()
# Load the MHD64 dataset
dataset = MHD64Dataset(
root_dir=Path(config.data.root).expanduser(),
train_task="next_step",
eval_tasks=["next_step", "autoregression"],
first_only=True,
)
# Print dataset metadata
print(dataset.train_db.metadata.n_steps_per_trajectory)
# Create data loaders
train_loader = DataLoader(dataset.train_db, batch_size=config.data.batch_size)
test_loaders = {}
for mode, db in dataset.test_dbs.items():
test_loaders[mode] = DataLoader(db, batch_size=config.data.test_batch_size)
# Get data processor from dataset
data_processor = dataset.data_processor
print(data_processor)
# Model initialization
model = get_model(config)
# convert dataprocessor to an MGPatchingDataprocessor if patching levels > 0
if config.patching.levels > 0:
data_processor = MGPatchingDataProcessor(
model=model,
in_normalizer=data_processor.normalizer,
out_normalizer=data_processor.normalizer,
padding_fraction=config.patching.padding,
stitching=config.patching.stitching,
levels=config.patching.levels,
use_distributed=config.distributed.use_distributed,
device=device,
)
# Reconfigure DataLoaders to use a DistributedSampler
# if in distributed data parallel mode
"""if config.distributed.use_distributed:
train_sampler = DistributedSampler(dataset.train_db, rank=get_local_rank())
train_loader = DataLoader(dataset=dataset.train_db,
batch_size=config.data.batch_size,
sampler=train_sampler)
for (res, loader), batch_size in zip(test_loaders.items(), config.data.test_batch_sizes):
test_db = loader.dataset
test_sampler = DistributedSampler(test_db, rank=get_local_rank())
test_loaders[res] = DataLoader(dataset=test_db,
batch_size=batch_size,
shuffle=False,
sampler=test_sampler)"""
# Create the optimizer
optimizer = AdamW(
model.parameters(),
lr=config.opt.learning_rate,
weight_decay=config.opt.weight_decay,
)
if config.opt.scheduler == "ReduceLROnPlateau":
scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(
optimizer,
factor=config.opt.gamma,
patience=config.opt.scheduler_patience,
mode="min",
)
elif config.opt.scheduler == "CosineAnnealingLR":
scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(
optimizer, T_max=config.opt.scheduler_T_max
)
elif config.opt.scheduler == "StepLR":
scheduler = torch.optim.lr_scheduler.StepLR(
optimizer, step_size=config.opt.step_size, gamma=config.opt.gamma
)
else:
raise ValueError(f"Got scheduler={config.opt.scheduler}")
# Loss function configuration
l2loss = LpLoss(d=3, p=2)
h1loss = H1Loss(d=3)
if config.opt.training_loss == "l2":
train_loss = l2loss
elif config.opt.training_loss == "h1":
train_loss = h1loss
else:
raise ValueError(
f"Got training_loss={config.opt.training_loss}"
f'but expected one of ["l2", "h1"]'
)
eval_losses = {"h1": h1loss, "l2": l2loss}
if config.verbose and is_logger:
print("\n### MODEL ###\n", model)
print("\n### OPTIMIZER ###\n", optimizer)
print("\n### SCHEDULER ###\n", scheduler)
print("\n### LOSSES ###")
print(f"\n * Train: {train_loss}")
print(f"\n * Test: {eval_losses}")
print(f"\n### Beginning Training...\n")
sys.stdout.flush()
trainer = Trainer(
model=model,
n_epochs=config.opt.n_epochs,
device=device,
data_processor=data_processor,
mixed_precision=config.opt.mixed_precision,
wandb_log=config.wandb.log,
eval_interval=config.opt.eval_interval,
log_output=config.wandb.log_output,
use_distributed=config.distributed.use_distributed,
verbose=config.verbose and is_logger,
)
# Log model parameter count
if is_logger:
n_params = count_model_params(model)
if config.verbose:
print(f"\nn_params: {n_params}")
sys.stdout.flush()
if config.wandb.log:
to_log = {"n_params": n_params}
if config.n_params_baseline is not None:
to_log["n_params_baseline"] = (config.n_params_baseline,)
to_log["compression_ratio"] = (config.n_params_baseline / n_params,)
to_log["space_savings"] = 1 - (n_params / config.n_params_baseline)
wandb.log(to_log, commit=False)
wandb.watch(model)
# Start training process with autoregressive evaluation
trainer.train(
train_loader=train_loader,
test_loaders=test_loaders,
eval_modes={"autoregression": "autoregression"},
optimizer=optimizer,
scheduler=scheduler,
regularizer=False,
training_loss=train_loss,
eval_losses=eval_losses,
)
# Finalize WandB logging
if config.wandb.log and is_logger:
wandb.finish()
|