Created HuggingFace Space

app.py

@@ -0,0 +1,15 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Wed Oct 11 10:26:15 2023
+
+@author: peter
+"""
+
+import gradio as gr
+
+def greet(name):
+    return "Hello " + name + "!!"
+
+iface = gr.Interface(fn=greet, inputs="text", outputs="text")
+iface.launch()

qarac/corpora/CombinedCorpus.py

@@ -145,11 +145,11 @@ class CombinedCorpus(torch.utils.data.IterableDataset):
         
         X={key:self.pad(value,self.max_lengths[key])
            for (key,value) in X.items()}
-        Y={key:torch.tensor(value).float() if key=='consistency' else self.pad(value,
+        Y={key:torch.tensor(value,device='cuda').float() if key=='consistency' else self.pad(value,
                                                                        self.max_lengths[key],
                                                                        False)
            for (key,value) in Y.items()}
-        Y['question_answering'] = torch.zeros((n,768))
+        Y['question_answering'] = torch.zeros((n,768),device='cuda')
         return (X,
                 tuple([Y[key] 
                          for key in ('encode_decode',
@@ -175,7 +175,8 @@ class CombinedCorpus(torch.utils.data.IterableDataset):
         for sample in batch:
             sample.pad(maxlen,pad_id=self.pad_token)
         input_ids = torch.tensor([sample.ids
-                                  for sample in batch])
+                                  for sample in batch],
+                                 device='cuda')
         result = input_ids
         if inputs:
             attention_mask = torch.not_equal(input_ids,

scripts.py

@@ -135,7 +135,9 @@ def train_models(path):
     tokenizer = tokenizers.Tokenizer.from_pretrained('roberta-base')
     trainer = qarac.models.QaracTrainerModel.QaracTrainerModel('roberta-base', 
                                                                tokenizer)
+    trainer.cuda()
     loss_fn = CombinedLoss()
+    loss_fn.cuda()
     optimizer = torch.optim.NAdam(trainer.parameters(),lr=5.0e-5)
     scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer,gamma=0.9)
     training_data = qarac.corpora.CombinedCorpus.CombinedCorpus(tokenizer,