Dudi Lester

Encapsulate FSDPA in GaudiLlamaAttention

1

* Done to allow quantization using HQT * Added use_flash_attention, flash_attention_causal_mask and flash_attention_recompute to run_lm_eval * Enforce recompute flag on fsdpa quantization

Quantization for FSDPA

1

Added use_flash_attention, flash_attention_causal_mask and flash_attention_recompute to run_lm_eval Enforce recompute flag on fsdpa quantization Allow quantization using HQT Document FusedScaledDotProductAttention quantization

update quantization configuration files

1

update quantization configuration files