clean up

intel-analytics · Aug 23, 2024 · 147f163 · 147f163
1 parent 4dc044b
commit 147f163
Show file tree

Hide file tree

Showing 3 changed files with 2 additions and 9 deletions.
diff --git a/python/llm/src/ipex_llm/transformers/npu_model.py b/python/llm/src/ipex_llm/transformers/npu_model.py
@@ -149,6 +149,7 @@ def from_pretrained(cls, *args, **kwargs):
                 ),
             )
             from ipex_llm.transformers.npu_models.convert_mp import optimize_llm
+
             with torch.no_grad():
                 cls.load_convert(qtype, model, "cpu", *args, **kwargs)
                 create_npu_kernels(model)

diff --git a/python/llm/src/ipex_llm/transformers/npu_models/convert_mp.py b/python/llm/src/ipex_llm/transformers/npu_models/convert_mp.py
@@ -24,13 +24,6 @@ def convert_forward(m, target_m, new_forward):
         convert_forward(sub_m, target_m, new_forward)
 
 
-def optimize_llm_pre(model: torch.nn.Module):
-    if model.config.model_type == "qwen2" and model.config.intermediate_size == 18944:
-        # for qwen2-7B
-        from ipex_llm.transformers.npu_models.qwen2_mp import split_mlp
-        model.apply(split_mlp)
-
-
 def optimize_llm(
     model: torch.nn.Module,
     max_output_len=1024,

diff --git a/python/llm/src/ipex_llm/transformers/npu_models/qwen2_mp.py b/python/llm/src/ipex_llm/transformers/npu_models/qwen2_mp.py
@@ -39,7 +39,6 @@
 from transformers.modeling_outputs import BaseModelOutputWithPast
 from ipex_llm.transformers.npu_models.mp_models_base import run_model
 from ipex_llm.transformers.npu_models.mp_models_base import LLMBaseNNFactory
-from ipex_llm.transformers.npu_models.common import reshape_lm_head_input
 from transformers.modeling_outputs import CausalLMOutputWithPast
 from torch.nn import CrossEntropyLoss
 
@@ -198,7 +197,7 @@ def __init__(
             new_value_states = self.convert_to_fp16(curr_key_values[i][1])
 
         self.compile()
-    
+
     def build_decoder(
         self,
         hidden_states,