convert : renamed expert_weights_func to expert_gating_func

ggerganov · Jan 2, 2025 · 93aca64 · 93aca64
1 parent a43d495
commit 93aca64
Show file tree

Hide file tree

Showing 3 changed files with 7 additions and 7 deletions.
diff --git a/convert_hf_to_gguf.py b/convert_hf_to_gguf.py
@@ -3859,9 +3859,9 @@ def set_gguf_parameters(self):
         self.gguf_writer.add_expert_weights_norm(hparams["norm_topk_prob"])
 
         if hparams["scoring_func"] == "sigmoid":
-            self.gguf_writer.add_expert_weights_func(gguf.ExpertWeightsFuncType.SIGMOID)
+            self.gguf_writer.add_expert_gating_func(gguf.ExpertGatingFuncType.SIGMOID)
         elif hparams["scoring_func"] == "softmax":
-            self.gguf_writer.add_expert_weights_func(gguf.ExpertWeightsFuncType.SOFTMAX)
+            self.gguf_writer.add_expert_gating_func(gguf.ExpertGatingFuncType.SOFTMAX)
         else:
             raise ValueError(f"Unsupported scoring_func value: {hparams['scoring_func']}")
 

diff --git a/gguf-py/gguf/constants.py b/gguf-py/gguf/constants.py
@@ -103,7 +103,7 @@ class LLM:
         EXPERT_SHARED_COUNT               = "{arch}.expert_shared_count"
         EXPERT_WEIGHTS_SCALE              = "{arch}.expert_weights_scale"
         EXPERT_WEIGHTS_NORM               = "{arch}.expert_weights_norm"
-        EXPERT_WEIGHTS_FUNC               = "{arch}.expert_weights_func"
+        EXPERT_GATING_FUNC                = "{arch}.expert_gating_func"
         POOLING_TYPE                      = "{arch}.pooling_type"
         LOGIT_SCALE                       = "{arch}.logit_scale"
         DECODER_START_TOKEN_ID            = "{arch}.decoder_start_token_id"
@@ -1581,7 +1581,7 @@ class GGMLQuantizationType(IntEnum):
     TQ2_0   = 35
 
 
-class ExpertWeightsFuncType(IntEnum):
+class ExpertGatingFuncType(IntEnum):
     SOFTMAX  = 1
     SIGMOID  = 2
 

diff --git a/gguf-py/gguf/gguf_writer.py b/gguf-py/gguf/gguf_writer.py
@@ -26,7 +26,7 @@
     RopeScalingType,
     PoolingType,
     TokenType,
-    ExpertWeightsFuncType,
+    ExpertGatingFuncType,
 )
 
 from .quants import quant_shape_from_byte_shape
@@ -719,8 +719,8 @@ def add_expert_weights_scale(self, value: float) -> None:
     def add_expert_weights_norm(self, value: bool) -> None:
         self.add_bool(Keys.LLM.EXPERT_WEIGHTS_NORM.format(arch=self.arch), value)
 
-    def add_expert_weights_func(self, value: ExpertWeightsFuncType) -> None:
-        self.add_uint32(Keys.LLM.EXPERT_WEIGHTS_FUNC.format(arch=self.arch), value.value)
+    def add_expert_gating_func(self, value: ExpertGatingFuncType) -> None:
+        self.add_uint32(Keys.LLM.EXPERT_GATING_FUNC.format(arch=self.arch), value.value)
 
     def add_swin_norm(self, value: bool) -> None:
         self.add_bool(Keys.LLM.SWIN_NORM.format(arch=self.arch), value)