transformers

0af2381f - feat(granitemoe*): Remove logits upcast when computing loss (#42753)

feat(granitemoe*): Remove logits upcast when computing loss (#42753) * feat: Remove logits upcast when computing loss When the CausalLM loss is used, the upcast is done in the loss function utils, so this is redundant. https://github.com/huggingface/transformers/issues/42709 Branch: GraniteOptionalUpcast-42709 Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * chore: make fix-copies https://github.com/huggingface/transformers/issues/42709 Branch: GraniteOptionalUpcast-42709 Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> --------- Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>