Skip to content

Commit

Permalink
Enable megatron core loggers for GPT pretraining (#8354) (#8384)
Browse files Browse the repository at this point in the history
* Logging changes tested for gpt_pretraining



* Additional args



* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Signed-off-by: Aishwarya Bhandare <[email protected]>
Co-authored-by: ashbhandare <[email protected]>
Co-authored-by: Aishwarya Bhandare <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Eric Harper <[email protected]>
Signed-off-by: ataghibakhsh <[email protected]>
  • Loading branch information
5 people authored and JRD971000 committed Feb 16, 2024
1 parent 02a2ad4 commit 07f9779
Show file tree
Hide file tree
Showing 4 changed files with 54 additions and 2 deletions.
7 changes: 7 additions & 0 deletions examples/nlp/language_modeling/conf/megatron_gpt_config.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -211,6 +211,13 @@ model:
## Network
sharp: False # Enable the use of SHARP for NCCL data-parallel communications. This is going to be ignored if the network doesn't support SHARP.

## Megatron timers
enable_megatron_timers: False
megatron_timer_kwargs:
log_every_n_steps: 10
log_mode: minmax
barrier: False

data:
# Path to data must be specified by the user.
# Supports List, String and Dictionary
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -37,4 +37,5 @@ normalize_attention_scores: True # Whether to scale the output Q * K^T by 1 / sq
num_moe_experts: 1 # When >1, FFNs are changed to MoE layers
moe_frequency: 1 # every Nth ffn layer will be made MoE
moe_dropout: 0.0 # Dropout value for MoE layers
use_flash_attention: false # Use flash attention in self-attention module
use_flash_attention: false # Use flash attention in self-attention module
enable_megatron_timers: false # Megatron timers
Original file line number Diff line number Diff line change
Expand Up @@ -69,6 +69,13 @@

HAVE_MEGATRON_CORE = False

try:
from megatron.core import Timers

HAVE_MEGATRON_CORE_TIMERS = True
except (ImportError, ModuleNotFoundError):
HAVE_MEGATRON_CORE_TIMERS = False

__all__ = ["MegatronBaseModel"]


Expand Down Expand Up @@ -124,6 +131,17 @@ def __init__(self, cfg: DictConfig, trainer: Trainer, no_lm_init=True):
else torch.float32
)

self.megatron_timers = None
if self.cfg.get('enable_megatron_timers', False) and HAVE_MEGATRON_CORE_TIMERS:
self.megatron_timers_cfg = dict(self.cfg.get('megatron_timer_kwargs', dict()))
if 'log_every_n_steps' not in self.megatron_timers_cfg:
self.megatron_timers_cfg['log_every_n_steps'] = self.trainer.log_every_n_steps
if 'log_option' not in self.megatron_timers_cfg:
self.megatron_timers_cfg['log_option'] = 'minmax' # minmax, max, all
if 'barrier' not in self.megatron_timers_cfg:
self.megatron_timers_cfg['barrier'] = False
self.megatron_timers = Timers(log_level=2, log_option=self.megatron_timers_cfg['log_option'])

# set the megatron core model parallel config
self.model_parallel_config: ModelParallelConfig = self.build_model_parallel_config()

Expand Down Expand Up @@ -615,6 +633,13 @@ def sync_overlap_parameters(self, params=None):
def on_train_batch_end(self, outputs, dataloader_iter: Any, batch_idx: int, unused: Optional[int] = 0) -> None:
super().on_train_batch_end(outputs, dataloader_iter, batch_idx)

# Megatron Timers
if self.megatron_timers:
if self.global_step % self.megatron_timers_cfg["log_every_n_steps"] == 0:
logging.info(
"\n " + self.megatron_timers.get_all_timers_string(barrier=self.megatron_timers_cfg["barrier"])
)

# TODO: Replace with newer override for scheduler.step() instead of
# search for plugins for fp16 GradScalar
if self.trainer.precision_plugin is not None and isinstance(
Expand Down Expand Up @@ -1044,7 +1069,7 @@ def build_model_parallel_config(self) -> ModelParallelConfig:
and megatron_amp_O2, # NeMo does not currently support fp16 training with megatron amp O2, eval and inference is supported
"bf16": self.torch_dtype == torch.bfloat16 and megatron_amp_O2,
"params_dtype": self.params_dtype,
"timers": None, # NeMo does not currently support megatron core timers
"timers": self.megatron_timers,
"async_tensor_model_parallel_allreduce": self.cfg.get('tensor_model_parallel_world_size', 1) > 1
and not self.cfg.get('sequence_parallel', False),
"pipeline_dtype": pipeline_dtype,
Expand Down Expand Up @@ -1157,3 +1182,16 @@ def configure_sharded_model(self):
# Move the CPU-initialized model (with `use_cpu_initialization=True`) to GPU, which is to avoid
# out-of-memory carash before sharding. In case of GPU-initialized model, this is no-op.
self.model = self.model.cuda(torch.cuda.current_device())

def megatron_timer_start(self, name, log_level):
if self.megatron_timers:
self.megatron_timers(name, log_level).start(barrier=False)

def megatron_timer_stop(self, name):
if self.megatron_timers:
self.megatron_timers(name).stop()

def optimizer_step(self, *args, **kwargs):
self.megatron_timer_start('optimizer', log_level=1)
super().optimizer_step(*args, **kwargs)
self.megatron_timer_stop('optimizer')
Original file line number Diff line number Diff line change
Expand Up @@ -654,8 +654,11 @@ def training_step(self, dataloader_iter, batch_idx):

# when using sequence parallelism, the sequence parallel layernorm grads must be all-reduced
if self.cfg.get('tensor_model_parallel_size', 1) > 1 and self.cfg.get('sequence_parallel', False):
self.megatron_timer_start('allreduce_sequence_parallel_gradients', log_level=1)
self.allreduce_sequence_parallel_gradients()
self.megatron_timer_stop('allreduce_sequence_parallel_gradients')

self.megatron_timer_start('gradient_allreduce', log_level=1)
if self.use_fsdp:
# Reduce the gradients omitted from FSDP-sharding
self.allreduce_fsdp_sharding_omitted_gradients()
Expand All @@ -673,12 +676,15 @@ def training_step(self, dataloader_iter, batch_idx):
# async grad allreduce is not currently implemented for O1/autocasting mixed precision training
# so we all-reduce gradients after the pipeline
self.allreduce_gradients() # @sangkug we think this is causing memory to blow up (hurts perf)
self.megatron_timer_stop('gradient_allreduce')

if self.cfg.get('pipeline_model_parallel_size', 1) > 1 and self.cfg.get(
'share_embeddings_and_output_weights', True
):
self.megatron_timer_start('allreduce_first_last_embeddings', log_level=1)
# when using pipeline parallelism the first and last stage must keep embeddings in sync
self.allreduce_first_last_embeddings()
self.megatron_timer_stop('allreduce_first_last_embeddings')

## logging
if self.log_train_loss:
Expand Down

0 comments on commit 07f9779

Please sign in to comment.