Add DeepSeek3.2 and GlmMoeDsa into moe tune (#18876)
Co-authored-by: luoyuan.luo <luoyuan.luo@antgroup.com>
This commit is contained in:
@@ -86,7 +86,9 @@ def get_model_config(
|
|||||||
elif architecture in [
|
elif architecture in [
|
||||||
"DeepseekV2ForCausalLM",
|
"DeepseekV2ForCausalLM",
|
||||||
"DeepseekV3ForCausalLM",
|
"DeepseekV3ForCausalLM",
|
||||||
|
"DeepseekV32ForCausalLM",
|
||||||
"Glm4MoeForCausalLM",
|
"Glm4MoeForCausalLM",
|
||||||
|
"GlmMoeDsaForCausalLM",
|
||||||
"MistralLarge3ForCausalLM",
|
"MistralLarge3ForCausalLM",
|
||||||
]:
|
]:
|
||||||
E = (config.n_routed_experts // ep_size) + (
|
E = (config.n_routed_experts // ep_size) + (
|
||||||
@@ -95,7 +97,9 @@ def get_model_config(
|
|||||||
or architecture
|
or architecture
|
||||||
not in [
|
not in [
|
||||||
"DeepseekV3ForCausalLM",
|
"DeepseekV3ForCausalLM",
|
||||||
|
"DeepseekV32ForCausalLM",
|
||||||
"Glm4MoeForCausalLM",
|
"Glm4MoeForCausalLM",
|
||||||
|
"GlmMoeDsaForCausalLM",
|
||||||
"MistralLarge3ForCausalLM",
|
"MistralLarge3ForCausalLM",
|
||||||
]
|
]
|
||||||
else 1
|
else 1
|
||||||
|
|||||||
Reference in New Issue
Block a user