app update

app.py

@@ -74,9 +74,7 @@ def generate(
     eos_id:int = None,
 ) -> torch.Tensor:
     """Takes a conditioning sequence (prompt) as input and continues to generate as many tokens as requested.
-
     The implementation of this function is modified from A. Karpathy's nanoGPT.
-
     Args:
         model: The model to use.
         idx: Tensor of shape (T) with indices of the prompt sequence.
@@ -131,7 +129,7 @@ def generate(
 
 device = 'cuda' if torch.cuda.is_available() else 'cpu'
 
-def generate_dialogue(input_text, temperature=0.8, max_tokens=200, top_k=1):
+def generate_text(input_text, temperature=0.8, max_tokens=200, top_k=None):
     encoded = tokenizer.encode(input_text, device=fabric.device)
     max_returned_tokens = encoded.size(0) + max_tokens
 
@@ -148,4 +146,18 @@ def generate_dialogue(input_text, temperature=0.8, max_tokens=200, top_k=1):
 
     return(tokenizer.decode(y))
 
+import gradio as gr
+
+title = "GPT from scratch"
+
+description1 = "GPT implementation taken from <a href='https://github.com/Lightning-AI/lit-gpt'>Lit-GPT</a>. It is trained on a samples of the <a href='https://huggingface.co/datasets/togethercomputer/RedPajama-Data-1T-Sample'>RedPajama 1 trillion dataset</a> to understand how GPT's are trained and built. The github link can be found <a href='https://github.com/mkthoma/gpt_from_scratch'>here.</a>"
+
+demo = gr.Interface(generate_text,
+                    inputs=[gr.Textbox(label="Enter any prompt ", type="text", value="Once upon a time,"),
+                            gr.Slider(minimum=0, maximum=1, step=0.1, value=0.8, label="Temperature"),
+                            gr.Slider(minimum=200, maximum=1000, step=50, value=300, label="Max Tokens"),
+                            gr.Slider(minimum=10, maximum=100, step=5, value=20, label="Top K")],
+                    outputs=gr.Textbox(label="Text generated", type="text"), description=description1)
+
 
+demo.launch()