MCPcopy Create free account
hub / github.com/cli99/llm-analysis / get_num_params_per_layer_router

Method get_num_params_per_layer_router

llm_analysis/analysis.py:287–291  ·  view source on GitHub ↗
(self)

Source from the content-addressed store, hash-verified

285 return 2 * self.model_config.hidden_dim * self.model_config.ffn_embed_dim * self.model_config.moe_num_experts
286
287 def get_num_params_per_layer_router(self) -> int:
288 if self.model_config.moe_num_experts > 1:
289 return self.model_config.hidden_dim * self.model_config.moe_num_experts
290 else:
291 return 0
292
293 def get_num_params_per_layer_layernorm(self) -> int:
294 return 2 * self.model_config.hidden_dim

Calls

no outgoing calls

Tested by

no test coverage detected