Unverified Commit 214fda42 authored by mcarilli's avatar mcarilli Committed by GitHub
Browse files

Allow verbose casting information for O1

parent 84028786
...@@ -199,7 +199,7 @@ def _initialize(models, optimizers, properties): ...@@ -199,7 +199,7 @@ def _initialize(models, optimizers, properties):
if properties.patch_torch_functions: if properties.patch_torch_functions:
# handle is unused here. It's accessible later through a global value anyway. # handle is unused here. It's accessible later through a global value anyway.
handle = amp_init(loss_scale=properties.loss_scale) handle = amp_init(loss_scale=properties.loss_scale, verbose=(_amp_state.verbosity == 2))
for optimizer in optimizers: for optimizer in optimizers:
# Disable Amp casting for the optimizer step, because it should only be # Disable Amp casting for the optimizer step, because it should only be
# applied to FP32 master params anyway. # applied to FP32 master params anyway.
......
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