Create app.py

app.py

@@ -0,0 +1,31 @@
+import streamlit as st
+
+st.title("Grammar Corrector")
+st.write("Paste or type text, submit and the machine will attempt to correct your text's grammar.")
+
+default_text = "This should working"
+sent = st.text_area("Text", default_text, height=40)
+num_correct_options = st.number_input('Number of Correction Options', min_value=1, max_value=3, value=1, step=1)
+
+from transformers import T5ConditionalGeneration, T5Tokenizer
+import torch
+torch_device = 'cuda' if torch.cude.is_available() else 'cpu'
+tokenizer = T5Tokenizer.from_pretrained('deep-learning-analytics/GrammarCorrector')
+model = T5ConditionalGeneration.from_pretrained('deep-learning-analytics/GrammarCorrector').to(torch_device)
+
+def correct_grammar(input_text, num_correct_options=num_correct_options):
+  batch = tokenizer([input_text], truncation=True, padding = 'max length', max_length = 64, return_tensors = 'pt').to(torch_device)
+  results = model.generate(**batch, max_length = 64, num_beams = 2, num_correct_options = num_correct_options, temperature = 1.5)
+  
+  return results
+  
+results = correct_grammar(sent, num_correct_options)
+
+generated_options = []
+for generated_option_idx, generated_option in enumerate(results):
+    
+    text = tokenizer.decode(generated_option, clean_up_tokenization_spaces = True, skip_special_tokens = True)
+    generated_options.append(text)
+    
+st.write(generated options)
+