PowerInfer
/

TurboSparse-Mixtral

Feature Extraction

turbosparsemixtral

Model card Files Files and versions Community

yixinsong commited on May 31

Commit

f94ad98

•

1 Parent(s): 8f55951

turbo

Files changed (2) hide show

config.json +2 -2
modeling_supersparsemixtral.py +1 -1

config.json CHANGED Viewed

@@ -1,6 +1,6 @@
 {
   "architectures": [
-    "SuperSparseMixtralForCausalLM"
   ],
   "auto_map": {
       "AutoConfig": "configuration_supersparsemixtral.SuperSparseMixtralConfig",
@@ -15,7 +15,7 @@
   "initializer_range": 0.02,
   "intermediate_size": 14336,
   "max_position_embeddings": 32768,
-  "model_type": "supersparsemixtral",
   "num_attention_heads": 32,
   "num_experts_per_tok": 2,
   "num_hidden_layers": 32,

 {
   "architectures": [
+    "TurboSparseMixtralForCausalLM"
   ],
   "auto_map": {
       "AutoConfig": "configuration_supersparsemixtral.SuperSparseMixtralConfig",
   "initializer_range": 0.02,
   "intermediate_size": 14336,
   "max_position_embeddings": 32768,
+  "model_type": "trubosparsemixtral",
   "num_attention_heads": 32,
   "num_experts_per_tok": 2,
   "num_hidden_layers": 32,

modeling_supersparsemixtral.py CHANGED Viewed

@@ -541,7 +541,7 @@ if is_torch_fx_available():
 logger = logging.get_logger(__name__)
-_CONFIG_FOR_DOC = "SuperSparseMixtralConfig"
 def load_balancing_loss_func(

 logger = logging.get_logger(__name__)
+_CONFIG_FOR_DOC = "TurboSparseMixtralConfig"
 def load_balancing_loss_func(