about optimizer.step() #54
Description
train.py:
if accelerator.state.deepspeed_plugin is None:
optimizer.step()
Why is the .step() performed when the plugin is none during training? I don’t know if this is a bug.
{{ message }}
train.py:
if accelerator.state.deepspeed_plugin is None:
optimizer.step()
Why is the .step() performed when the plugin is none during training? I don’t know if this is a bug.