[NPU] Support GW for NPU C++ (#12450)

This commit is contained in:
Ruonan Wang 2024-11-26 01:46:40 -08:00 committed by GitHub
parent c2efa264d9
commit 7b40f9b372
No known key found for this signature in database
GPG key ID: B5690EEEBB952194

View file

@ -445,7 +445,9 @@ def convert_llm_for_deploy(model: torch.nn.Module,
"qkv_bias": True,
"use_prefill_sdp": False,
"weight_num": 7,
"weight_idx": 8}
"weight_idx": 8,
"n_splits_linear": n_splits_linear,
"n_splits_down_proj": n_splits_down_proj}
model.config.update(update_dict)
model.config.save_pretrained(save_directory)
@ -495,7 +497,9 @@ def convert_llm_for_deploy(model: torch.nn.Module,
"weight_num": 7,
"weight_idx": 5,
"embedding_post": embedding_post,
"cos_sin_input": cos_sin_input}
"cos_sin_input": cos_sin_input,
"n_splits_linear": n_splits_linear,
"n_splits_down_proj": n_splits_down_proj}
model.config.update(update_dict)
model.config.save_pretrained(save_directory)
@ -530,7 +534,9 @@ def convert_llm_for_deploy(model: torch.nn.Module,
"weight_num": 7,
"weight_idx": 5,
"model_type": "minicpm",
"embedding_post": True}
"embedding_post": True,
"n_splits_linear": n_splits_linear,
"n_splits_down_proj": n_splits_down_proj}
model.config.update(update_dict)
model.config.save_pretrained(save_directory)