Upload README.md with huggingface_hub

README.md

@@ -7,12 +7,13 @@ license: mit
  
 # dm_qwen4b_emulator
  
-2-layer MLP.
+MLP with 3 hidden Linear->LayerNorm->ReLU block(s) and a linear output head.
  
 ## Config
 - `input_dim`: 6
-- `hidden_dim`: 256
+- `hidden_dim`: 512
 - `output_dim`: 3
+- `n_layers`: 3
  
 ## Usage
  
@@ -32,13 +33,16 @@ class MLP(nn.Module):
             nn.Linear(hidden_dim, hidden_dim),
             nn.LayerNorm(hidden_dim),
             nn.ReLU(),
+            nn.Linear(hidden_dim, hidden_dim),
+            nn.LayerNorm(hidden_dim),
+            nn.ReLU(),
             nn.Linear(hidden_dim, output_dim),
         )
     def forward(self, x):
         return self.mlp(x)
  
 path = hf_hub_download("anonom799/dm_qwen4b_emulator", "model.safetensors")
-model = MLP(input_dim=6, hidden_dim=256, output_dim=3)
+model = MLP(input_dim=6, hidden_dim=512, output_dim=3)
 model.load_state_dict(load_file(path))
 model.eval()