onnxruntime
07dc25e9 - Compute global gradient norm according to 'enable_grad_norm_clip' (#5728)

Commit
5 years ago
Compute global gradient norm according to 'enable_grad_norm_clip' (#5728) * Introduce PassThrough op to wait for all gradient ready before weight update * Compute gradient norm for fp32 runs * Update FE UT expected value * Respect enable_grad_norm_clip
Author
Parents
Loading