cerebras.modelzoo.common.run_cstorch_flow.optimizer_step_with_summaries#

cerebras.modelzoo.common.run_cstorch_flow.optimizer_step_with_summaries(loss, optimizer, grad_scaler, max_gradient_norm=None, max_gradient_value=None, log_summaries=False, model=None)[source]#

Customized equivalent to cstorch.amp.optimizer_step additionally featuring grad norm summaries.