Add num beams

app.py

@@ -35,11 +35,12 @@ def query(payload):
     print(response)
     return json.loads(response.content.decode("utf-8"))
     
-def inference(input_sentence, max_length, no_repeat_ngram_size, temperature,top_k, top_p, greedy_decoding, seed=42):
+def inference(input_sentence, max_length, no_repeat_ngram_size, num_beams, temperature,top_k, top_p, greedy_decoding, seed=42):
     top_k = None if top_k == 0 else top_k
+    greedy = False if num_beams > 0
     payload = {"inputs": input_sentence,
                "parameters": {"max_new_tokens": max_length, "top_k": top_k, "top_p": top_p, "temperature": temperature,
-                              "do_sample": not greedy_decoding, "seed": seed, "early_stopping":no_repeat_ngram_size > 1, "no_repeat_ngram_size":no_repeat_ngram_size}}
+                              "do_sample": not greedy_decoding, "seed": seed, "early_stopping":num_beams > 0, "no_repeat_ngram_size":no_repeat_ngram_size}}
     data = query(
         payload
     )
@@ -53,6 +54,7 @@ gr.Interface(
         gr.inputs.Textbox(label="Input"),
         gr.inputs.Slider(1, 64, default=8, label="Tokens to generate"),
         gr.inputs.Slider(1, 10, default=2, step=1, label="No repeat N gram"),
+        gr.inputs.Slider(0, 10, default=5, step=1, label="Num beams"),
         gr.inputs.Slider(0.0, 1.0, default=0.1, step=0.05, label="Temperature"),
         gr.inputs.Slider(0, 64, default=0, step=1, label="Top K"),
         gr.inputs.Slider(0.0, 10, default=0.9, step=0.05, label="Top P"),