Update PersadianNanoV4Model.py

PersadianNanoV4Model.py

@@ -14,10 +14,10 @@ import torch.nn.functional as F
 class PersadianNanoV4Config:
     def __init__(self):
 
-        # Core architecture (MATCH CHECKPOINT)
+        # Core architecture
         self.hidden_size = 512
         self.intermediate_size = 1024
-        self.num_hidden_layers = 12
+        self.num_hidden_layers = 6
 
         # Attention
         self.num_attention_heads = 8
@@ -58,7 +58,7 @@ class PersadianNanoV4Config:
         self.attention_dropout = 0.0
         self.use_flash_attention = True
 
-        # Precision (IMPORTANT: keep as string or remove entirely)
+        # Precision
         self.torch_dtype = "float16"