[Bugfix] add support for 'num_attention_groups' in ModelArchConfigConvertorBase for Step3p5 (#39796)
Signed-off-by: realliujiaxu <realliujiaxu@163.com>
This commit is contained in:
@@ -356,6 +356,23 @@
|
|||||||
"is_multimodal_model": false,
|
"is_multimodal_model": false,
|
||||||
"dtype": "torch.float32"
|
"dtype": "torch.float32"
|
||||||
},
|
},
|
||||||
|
"stepfun-ai/Step-3.5-Flash": {
|
||||||
|
"architectures": [
|
||||||
|
"Step3p5ForCausalLM"
|
||||||
|
],
|
||||||
|
"model_type": "step3p5",
|
||||||
|
"text_model_type": "step3p5",
|
||||||
|
"hidden_size": 4096,
|
||||||
|
"total_num_hidden_layers": 45,
|
||||||
|
"total_num_attention_heads": 64,
|
||||||
|
"head_size": 128,
|
||||||
|
"vocab_size": 128896,
|
||||||
|
"total_num_kv_heads": 8,
|
||||||
|
"num_experts": 288,
|
||||||
|
"is_deepseek_mla": false,
|
||||||
|
"is_multimodal_model": false,
|
||||||
|
"dtype": "torch.bfloat16"
|
||||||
|
},
|
||||||
"nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16": {
|
"nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16": {
|
||||||
"architectures": [
|
"architectures": [
|
||||||
"NemotronHForCausalLM"
|
"NemotronHForCausalLM"
|
||||||
|
|||||||
@@ -16,6 +16,7 @@ BASE_TRUST_REMOTE_CODE_MODELS = {
|
|||||||
"nvidia/Llama-3_3-Nemotron-Super-49B-v1",
|
"nvidia/Llama-3_3-Nemotron-Super-49B-v1",
|
||||||
"nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16",
|
"nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16",
|
||||||
"XiaomiMiMo/MiMo-7B-RL",
|
"XiaomiMiMo/MiMo-7B-RL",
|
||||||
|
"stepfun-ai/Step-3.5-Flash",
|
||||||
# Excluded: Not available online right now
|
# Excluded: Not available online right now
|
||||||
# "FreedomIntelligence/openPangu-Ultra-MoE-718B-V1.1",
|
# "FreedomIntelligence/openPangu-Ultra-MoE-718B-V1.1",
|
||||||
"meituan-longcat/LongCat-Flash-Chat",
|
"meituan-longcat/LongCat-Flash-Chat",
|
||||||
|
|||||||
@@ -77,6 +77,8 @@ class ModelArchConfigConvertorBase:
|
|||||||
"num_key_value_heads",
|
"num_key_value_heads",
|
||||||
# For ChatGLM:
|
# For ChatGLM:
|
||||||
"multi_query_group_num",
|
"multi_query_group_num",
|
||||||
|
# For Step3p5:
|
||||||
|
"num_attention_groups",
|
||||||
]
|
]
|
||||||
# For non-grouped-query attention models, the number of KV heads is
|
# For non-grouped-query attention models, the number of KV heads is
|
||||||
# equal to the number of attention heads.
|
# equal to the number of attention heads.
|
||||||
|
|||||||
Reference in New Issue
Block a user