Set a repetition_penalty constant as 1.8

app.py

@@ -14,11 +14,16 @@ checkpoint = "CohereForAI/aya-101"
 tokenizer = AutoTokenizer.from_pretrained(checkpoint)
 model = AutoModelForSeq2SeqLM.from_pretrained(checkpoint, torch_dtype=torch.bfloat16, low_cpu_mem_usage=True, device_map=device)
 
+#Set a the value of the repetition penalty
+#The higher the value, the less repetitive the generated text will be
+#Note that `repetition_penalty` has to be a strictly positive float
+repetition_penalty = 1.8
+
 @spaces.GPU
 def aya(text, max_new_tokens):
     model.to(device)
     inputs = tokenizer.encode(text, return_tensors="pt").to(device)
-    outputs = model.generate(inputs, max_new_tokens=max_new_tokens)
+    outputs = model.generate(inputs, max_new_tokens=max_new_tokens, repetition_penalty=repetition_penalty)
     translation = tokenizer.decode(outputs[0], skip_special_tokens=True)
     return translation