Debug

app.py

@@ -28,7 +28,7 @@ class Interactive:
         self.tokenizer = transformers.AutoTokenizer.from_pretrained(MODEL_NAME, use_auth_token=HF_TOKEN_DOWNLOAD)
         if MODE == 'debug':
             return
-        self.model = transformers.T5EncoderModel.from_pretrained(MODEL_NAME, use_auth_token=HF_TOKEN_DOWNLOAD, low_cpu_mem_usage=True, device_map='auto', torch_dtype='auto')
+        self.model = transformers.T5EncoderModel.from_pretrained(MODEL_NAME, use_auth_token=HF_TOKEN_DOWNLOAD, low_cpu_mem_usage=True, device_map='auto', torch_dtype='auto', offload_folder='offload')
         self.model.D = self.model.shared.embedding_dim
         self.linear = torch.nn.Linear(self.model.D, 1, dtype=self.model.dtype).to(device)
         self.linear.weight = torch.nn.Parameter(self.model.shared.weight[32099, :].unsqueeze(0)) # (1, D)