Update app.py

app.py

@@ -1,32 +1,26 @@
-import torch
-from transformers import (T5ForConditionalGeneration,T5Tokenizer)
 import gradio as gr
+import torch
+from transformers import T5Tokenizer, T5ForConditionalGeneration
 
-best_model_path = "swcrazyfan/KingJamesify-T5-large"
-model = T5ForConditionalGeneration.from_pretrained(best_model_path)
-tokenizer = T5Tokenizer.from_pretrained("swcrazyfan/KingJamesify-T5-large")
-
-def tokenize_data(text):
-	# Tokenize the review body
-	input_ = "kingify: " + str(text) + ' </s>'
-	max_len = 512
-	# tokenize inputs
-	tokenized_inputs = tokenizer(input_, padding='max_length', truncation=True, max_length=max_len, return_attention_mask=True, return_tensors='pt')
+model = T5ForConditionalGeneration.from_pretrained('swcrazyfan/KingJamesify-T5-large')
+tokenizer = T5Tokenizer.from_pretrained('swcrazyfan/KingJamesify-T5-large')
 
-	inputs={"input_ids": tokenized_inputs['input_ids'],
-		"attention_mask": tokenized_inputs['attention_mask']}
-	return inputs
+def king_jamesify(input_text):
+    input_ids = tokenizer.encode(input_text, return_tensors='pt').to(torch.int64)
+    generated_ids = model.generate(input_ids=input_ids, max_length=100, num_beams=4, repetition_penalty=2.5, length_penalty=1.0, temperature=1.0, top_k=50, top_p=1.0, no_repeat_ngram_size=3)
+    result = tokenizer.decode(generated_ids[0], skip_special_tokens=True)
+    return result
 
-def generate_answers(text, temperature, num_beams, max_length):
-	inputs = tokenize_data(text)
-	results= model.generate(input_ids= inputs['input_ids'], attention_mask=inputs['attention_mask'], do_sample=True,
-						   num_beams=num_beams,
-							max_length=max_length,
-							min_length=1,
-							early_stopping=True,
-							num_return_sequences=1,
-							temperature=temperature)
-	answer = tokenizer.decode(results[0], skip_special_tokens=True)
-	return answer
+iface = gr.Interface(king_jamesify,
+                     [gr.inputs.Textbox(lines=20, label="Enter text to be King Jamesified"),
+                      gr.inputs.Slider(minimum=0.0, maximum=1.0, default=0.7, label="Temperature"),
+                      gr.inputs.Slider(minimum=1, maximum=100, default=100, label="Max Length"),
+                      gr.inputs.Slider(minimum=1, maximum=10, default=4, label="Number of Beams"),
+                      gr.inputs.Slider(minimum=0.0, maximum=10.0, default=2.5, label="Repetition Penalty"),
+                      gr.inputs.Slider(minimum=0.0, maximum=10.0, default=1.0, label="Length Penalty"),
+                      gr.inputs.Slider(minimum=1, maximum=100, default=50, label="Top K"),
+                      gr.inputs.Slider(minimum=0.0, maximum=1.0, default=1.0, label="Top P"),
+                      gr.inputs.Slider(minimum=1, maximum=10, default=3, label="No Repeat Ngram Size")],
+                     gr.outputs.Textbox(label="King Jamesified Text"))
 
-iface = gr.Interface(title="Kingify", description="Write anything below. Then, click submit to 'Kingify' it.", fn=generate_answers, inputs=[gr.inputs.Textbox(label="Original Text",lines=10), gr.inputs.Slider(label="Temperature", default=0.7, minimum=0.0, maximum=1.0, step=0.01), gr.inputs.Slider(label="Number of Beams", default=5, minimum=1, maximum=10, step=1), gr.inputs.Textbox(label="Max Length", default=512, lines=1)], outputs=["text"])
+iface.launch()