making model first

app.py

@@ -17,7 +17,7 @@ import torch
 sys.path.insert(1, "../")
 # from utils import model_utils, train_utils, data_utils, run_utils
 # from model_utils import jason_regnet_maker, jason_efficientnet_maker
-from model_utils.efficientnet_config import EfficientNetConfig, EfficientNetPreTrained
+from model_utils.efficientnet_config import EfficientNetConfig, EfficientNetPreTrained, EfficientNet
 
 model_path = 'chlab/'
 # model_path = './models/'
@@ -223,13 +223,22 @@ def predict_and_analyze(model_name, num_channels, dim, input_channel, image):
         )
         EfficientNetConfig.model_type = "efficientnet_%s_planet_detection" % (hparams.num_channels)
     
-    config.save_pretrained(save_directory=model_loading_name)
+    # config.save_pretrained(save_directory=model_loading_name)
     # config = EfficientNetConfig.from_pretrained(model_loading_name)
     
     # model = EfficientNetPreTrained.from_pretrained(model_loading_name)
     # model = AutoModel.from_pretrained(model_loading_name, trust_remote_code=True)
     
-    model = cached_download(hf_hub_url(model_loading_name, filename="pytorch_model.bin"))
+    model = EfficientNet(dropout=hparams.dropout,
+                                    num_channels=hparams.num_channels,
+                                    num_classes=hparams.num_classes,
+                                    size=hparams.size,
+                                    stochastic_depth_prob=hparams.stochastic_depth_prob,
+                                    width_mult=hparams.width_mult,
+                                    depth_mult=hparams.depth_mult,)
+    
+    model_url = cached_download(hf_hub_url(model_loading_name, filename="pytorch_model.bin"))
+    model.load_state_dict(torch.load(model_url, map_location='cpu'))
     
     print(model)