[NPU] fix transpose_value = False for NPU optimize_model=True (#12525)

2024-12-10 23:51:39 -08:00 · 2024-12-10 23:51:39 -08:00 · 41ef4974ab
commit 41ef4974ab
parent 588bfa24dc
1 changed files with 1 additions and 0 deletions
--- a/python/llm/src/ipex_llm/transformers/npu_models/mp_models_base.py
+++ b/python/llm/src/ipex_llm/transformers/npu_models/mp_models_base.py
@ -225,6 +225,7 @@ class LLMBaseNNFactory(NNFactory):
            head_dim=head_dim,
        )
        new_key_states = key_states
+        new_value_states = value_states

        if mode == "decode":
            key_states = self.concat(past_key, key_states, axis=-2)