leftyer/mindformers
1
0
Fork 0
mirror of https://gitee.com/mindspore/mindformers.git synced 2025-12-06 19:42:57 +08:00
Code Issues Packages Projects Releases Wiki Activity

hide monitor config

Browse Source
This commit is contained in:
魏琢艺
2025-07-24 19:44:29 +08:00
parent 5b875ecf5f
commit c8bab7f145
12 changed files with 52 additions and 53 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
research/deepseek3/deepseek2_model.py
View File

@@ -1609,14 +1609,8 @@ class DeepseekV2ForCausalLM(DeepseekV2PreTrainedModel):
vocab_size, mp)
logger.warning("Now, the model_parallel num of Loss will be changed: mp = 1")
loss_parallel_config.model_parallel = 1
monitor_config = getattr(config, "monitor_config", None)
monitor_on = getattr(monitor_config, "monitor_on", False)
check_for_nan_in_loss_and_grad = monitor_on and bool(getattr(monitor_config, "local_loss_format", None))
monitor_device_local_loss = monitor_on and bool(getattr(monitor_config, "device_local_loss_format", None))
calculate_per_token_loss = getattr(config, "calculate_per_token_loss", False)
self.loss = CrossEntropyLoss(parallel_config=loss_parallel_config,
check_for_nan_in_loss_and_grad=check_for_nan_in_loss_and_grad,
monitor_device_local_loss=monitor_device_local_loss,
calculate_per_token_loss=calculate_per_token_loss,
seq_split_num=config.parallel_config.seq_split_num)

8
research/telechat2/telechat.py
View File

@@ -319,13 +319,7 @@ class TelechatForCausalLM(TelechatPreTrainedModel):
logger.warning("Now, the model_parallel num of Loss will be changed: mp = 1")
loss_parallel_config.model_parallel = 1
loss_parallel_config.data_parallel *= loss_parallel_config.context_parallel
monitor_config = getattr(config, "monitor_config", None)
monitor_on = getattr(monitor_config, "monitor_on", False)
check_for_nan_in_loss_and_grad = monitor_on and bool(getattr(monitor_config, "local_loss_format", None))
monitor_device_local_loss = monitor_on and bool(getattr(monitor_config, "device_local_loss_format", None))
self.loss = CrossEntropyLoss(parallel_config=loss_parallel_config,
check_for_nan_in_loss_and_grad=check_for_nan_in_loss_and_grad,
monitor_device_local_loss=monitor_device_local_loss)
self.loss = CrossEntropyLoss(parallel_config=loss_parallel_config)
dp = config.parallel_config.data_parallel
mp = config.parallel_config.model_parallel