Update app.py

app.py

@@ -207,7 +207,7 @@ def beam_search_decode(model, src, src_mask, max_len, start_symbol, beam_size ,l
     best_beam = beams[0][0]
     return best_beam
 
-def translate(model: torch.nn.Module, strategy:str = 'greedy' , src_sentence: str, lenght_extend :int = 5, beam_size: int = 5,  length_penalty:float = 0.6):
+def translate(model: torch.nn.Module, src_sentence: str, strategy:str = 'greedy' , lenght_extend :int = 5, beam_size: int = 5,  length_penalty:float = 0.6):
     assert strategy in ['greedy','beam search'], 'the strategy for decoding has to be either greedy or beam search'
     # Tokenize the source sentence
     src = source_tokenizer(src_sentence, **token_config)['input_ids']
@@ -249,7 +249,19 @@ model.eval()
 
 import gradio as gr
 
-x = lambda text : translate(x)
-
-iface = gr.Interface(fn=x, inputs="text", outputs="text")
+iface = gr.Interface(
+    fn=translate,
+    inputs=[
+        gr.inputs.Textbox("Enter a sentence to translate"),
+        gr.inputs.Radio(['greedy', 'beam search'], label="Decoding Strategy"),
+        gr.inputs.Number(label="Length Extend (for greedy)", default=5),
+        gr.inputs.Number(label="Beam Size (for beam search)", default=5),
+        gr.inputs.Number(label="Length Penalty (for beam search)", default=0.6)
+    ],
+    outputs=gr.outputs.Textbox("Translation"),
+    title="Translation Interface",
+    description="Translate text using a pre-trained model.",
+)
+
+# Launch the Gradio interface
 iface.launch()