Skip to content

GitLab

Commit 35410da7 authored by lukovnikov's avatar lukovnikov
Browse files

added warning

parent 4d79e0d3
Hide whitespace changes
Inline Side-by-side
Showing with 6 additions and 6 deletions
pytorch_pretrained_bert/optimization.py
View file @ 35410da7
...@@ -159,8 +159,8 @@ class BertAdam(Optimizer): ...@@ -159,8 +159,8 @@ class BertAdam(Optimizer):
# warning for exceeding t_total (only active with warmup_linear # warning for exceeding t_total (only active with warmup_linear
if group['schedule'] == "warmup_linear" and progress > 1. and not warned_for_t_total: if group['schedule'] == "warmup_linear" and progress > 1. and not warned_for_t_total:
logger.warning( logger.warning(
"Training beyond specified 't_total' steps. Learning rate set to {}. " "Training beyond specified 't_total' steps with schedule '{}'. Learning rate set to {}. "
"Please set 't_total' of {} correctly.".format(lr_scheduled, self.__class__.__name__)) "Please set 't_total' of {} correctly.".format(group['schedule'], lr_scheduled, self.__class__.__name__))
warned_for_t_total = True warned_for_t_total = True
# end warning # end warning
else: else:
......
pytorch_pretrained_bert/optimization_openai.py
View file @ 35410da7
...@@ -29,14 +29,14 @@ def warmup_cosine(x, warmup=0.002): ...@@ -29,14 +29,14 @@ def warmup_cosine(x, warmup=0.002):
return 0.5 * (1.0 + torch.cos(math.pi * x)) return 0.5 * (1.0 + torch.cos(math.pi * x))
def warmup_constant(x, warmup=0.002): def warmup_constant(x, warmup=0.002):
""" Linearly increases learning rate over `warmup`*`t_total` (as provided to BertAdam) training steps. """ Linearly increases learning rate over `warmup`*`t_total` (as provided to OpenAIAdam) training steps.
Learning rate is 1. afterwards. """ Learning rate is 1. afterwards. """
if x < warmup: if x < warmup:
return x/warmup return x/warmup
return 1.0 return 1.0
def warmup_linear(x, warmup=0.002): def warmup_linear(x, warmup=0.002):
""" Specifies a triangular learning rate schedule where peak is reached at `warmup`*`t_total`-th (as provided to BertAdam) training step. """ Specifies a triangular learning rate schedule where peak is reached at `warmup`*`t_total`-th (as provided to OpenAIAdam) training step.
After `t_total`-th training step, learning rate is zero. """ After `t_total`-th training step, learning rate is zero. """
if x < warmup: if x < warmup:
return x/warmup return x/warmup
...@@ -142,8 +142,8 @@ class OpenAIAdam(Optimizer): ...@@ -142,8 +142,8 @@ class OpenAIAdam(Optimizer):
# warning for exceeding t_total (only active with warmup_linear # warning for exceeding t_total (only active with warmup_linear
if group['schedule'] == "warmup_linear" and progress > 1. and not warned_for_t_total: if group['schedule'] == "warmup_linear" and progress > 1. and not warned_for_t_total:
logger.warning( logger.warning(
"Training beyond specified 't_total' steps. Learning rate set to {}. " "Training beyond specified 't_total' steps with schedule '{}'. Learning rate set to {}. "
"Please set 't_total' of {} correctly.".format(lr_scheduled, self.__class__.__name__)) "Please set 't_total' of {} correctly.".format(group['schedule'], lr_scheduled, self.__class__.__name__))
warned_for_t_total = True warned_for_t_total = True
# end warning # end warning
else: else:
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment