optimum/habana/transformers/models/gemma/modeling_gemma.py [692:755]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
            if output_hidden_states:
                all_hidden_states += (hidden_states,)

            if self.gradient_checkpointing and self.training:
                layer_outputs = self._gradient_checkpointing_func(
                    decoder_layer.__call__,
                    hidden_states,
                    attention_mask,
                    position_ids,
                    past_key_values,
                    output_attentions,
                    use_cache,
                    cache_position,
                    None,
                    attn_softmax_bf16,
                    False,
                    use_flash_attention,
                    flash_attention_recompute,
                    flash_attention_causal_mask,
                )
            else:
                layer_outputs = decoder_layer(
                    hidden_states,
                    attention_mask=attention_mask,
                    position_ids=position_ids,
                    past_key_value=None if past_key_values is None else past_key_values[layer_idx],
                    output_attentions=output_attentions,
                    use_cache=use_cache,
                    cache_position=cache_position,
                    token_idx=token_idx,
                    attn_softmax_bf16=attn_softmax_bf16,
                    reuse_cache=reuse_cache,
                    use_flash_attention=use_flash_attention,
                    flash_attention_recompute=flash_attention_recompute,
                    flash_attention_causal_mask=flash_attention_causal_mask,
                    cache_idx=cache_idx,
                )

            hidden_states = layer_outputs[0]

            if use_cache:
                next_decoder_cache += (layer_outputs[2 if output_attentions else 1],)

            if output_attentions:
                all_self_attns += (layer_outputs[1],)

        hidden_states = self.norm(hidden_states)

        # add hidden states from the last decoder layer
        if output_hidden_states:
            all_hidden_states += (hidden_states,)

        next_cache = None
        if use_cache:
            next_cache = (
                next_decoder_cache.to_legacy_cache() if isinstance(next_decoder_cache, Cache) else next_decoder_cache
            )

        return BaseModelOutputWithPast(
            last_hidden_state=hidden_states,
            past_key_values=next_cache,
            hidden_states=all_hidden_states,
            attentions=all_self_attns,
        )
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



optimum/habana/transformers/models/starcoder2/modeling_starcoder2.py [600:662]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
            if output_hidden_states:
                all_hidden_states += (hidden_states,)

            if self.gradient_checkpointing and self.training:
                layer_outputs = self._gradient_checkpointing_func(
                    decoder_layer.__call__,
                    hidden_states,
                    attention_mask,
                    position_ids,
                    past_key_values,
                    output_attentions,
                    use_cache,
                    cache_position,
                    None,
                    attn_softmax_bf16,
                    False,
                    use_flash_attention,
                    flash_attention_recompute,
                    flash_attention_causal_mask,
                )
            else:
                layer_outputs = decoder_layer(
                    hidden_states,
                    attention_mask=attention_mask,
                    position_ids=position_ids,
                    past_key_value=None if past_key_values is None else past_key_values[layer_idx],
                    output_attentions=output_attentions,
                    use_cache=use_cache,
                    cache_position=cache_position,
                    token_idx=token_idx,
                    attn_softmax_bf16=attn_softmax_bf16,
                    reuse_cache=reuse_cache,
                    use_flash_attention=use_flash_attention,
                    flash_attention_recompute=flash_attention_recompute,
                    flash_attention_causal_mask=flash_attention_causal_mask,
                    cache_idx=cache_idx,
                )

            hidden_states = layer_outputs[0]

            if use_cache:
                next_decoder_cache += (layer_outputs[2 if output_attentions else 1],)

            if output_attentions:
                all_self_attns += (layer_outputs[1],)

        hidden_states = self.norm(hidden_states)

        # add hidden states from the last decoder layer
        if output_hidden_states:
            all_hidden_states += (hidden_states,)

        next_cache = None
        if use_cache:
            next_cache = (
                next_decoder_cache.to_legacy_cache() if isinstance(next_decoder_cache, Cache) else next_decoder_cache
            )
        return BaseModelOutputWithPast(
            last_hidden_state=hidden_states,
            past_key_values=next_cache,
            hidden_states=all_hidden_states,
            attentions=all_self_attns,
        )
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -