Spaces:
Running on L40S
Running on L40S
File size: 4,035 Bytes
9f818c5 | 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 | # SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: OpenMDW-1.1
import time
import torch
import wandb
from torch import Tensor
from cosmos_framework.callbacks.every_n import EveryN
from cosmos_framework.model._base import ImaginaireModel
from cosmos_framework.trainer import ImaginaireTrainer
from cosmos_framework.utils import log
from cosmos_framework.utils.distributed import rank0_only
from cosmos_framework.utils.easy_io import easy_io
class IterSpeed(EveryN):
"""
Args:
hit_thres (int): Number of iterations to wait before logging.
save_s3 (bool): Whether to save to S3.
save_s3_every_log_n (int): Save to S3 every n log iterations, which means save_s3_every_log_n n * every_n global iterations.
"""
def __init__(self, *args, hit_thres: int = 5, save_s3: bool = True, save_s3_every_log_n: int = 10, **kwargs):
super().__init__(*args, **kwargs)
self.time = None
self.hit_counter = 0
self.hit_thres = hit_thres
self.save_s3 = save_s3
self.save_s3_every_log_n = save_s3_every_log_n
self.name = self.__class__.__name__
self.last_hit_time = time.time()
def on_training_step_end(
self,
model: ImaginaireModel,
data_batch: dict[str, torch.Tensor],
output_batch: dict[str, torch.Tensor],
loss: torch.Tensor,
iteration: int = 0,
) -> None:
if self.hit_counter < self.hit_thres:
log.info(
f"Iteration {iteration}: "
f"Hit counter: {self.hit_counter + 1}/{self.hit_thres} | "
f"Loss: {loss.detach().item():.4f} | "
f"Time: {time.time() - self.last_hit_time:.2f}s",
rank0_only=False,
)
self.hit_counter += 1
self.last_hit_time = time.time()
#! useful for large scale training and avoid oom crash in the first two iterations!!!
torch.cuda.synchronize()
return
super().on_training_step_end(model, data_batch, output_batch, loss, iteration)
@rank0_only
def every_n_impl(
self,
trainer: ImaginaireTrainer,
model: ImaginaireModel,
data_batch: dict[str, Tensor],
output_batch: dict[str, Tensor],
loss: Tensor,
iteration: int,
) -> None:
if self.time is None:
self.time = time.time()
return
cur_time = time.time()
iter_speed = (cur_time - self.time) / self.every_n / self.step_size
log.info(
f"{iteration} : iter_speed {iter_speed:.2f} seconds per iteration | Loss: {loss.detach().item():.4f}",
rank0_only=False,
)
per_sample_batch_counter = dict()
if hasattr(model, "is_image_batch"):
is_image_batch = model.is_image_batch(data_batch)
if is_image_batch:
image_batch_size = len(data_batch[model.input_image_key])
per_sample_batch_counter["image_batch_size"] = image_batch_size
else:
video_batch_size = len(data_batch[model.input_video_key])
per_sample_batch_counter["video_batch_size"] = video_batch_size
if wandb.run:
sample_counter = getattr(trainer, "sample_counter", iteration)
wandb.log(
{
"timer/iter_speed": iter_speed,
"sample_counter": sample_counter,
}
| per_sample_batch_counter,
step=iteration,
)
self.time = cur_time
if self.save_s3:
if iteration % (self.save_s3_every_log_n * self.every_n) == 0:
easy_io.dump(
{
"iter_speed": iter_speed,
"iteration": iteration,
},
f"s3://rundir/{self.name}/iter_{iteration:09d}.yaml",
)
|