Update app.py

app.py

@@ -9,6 +9,7 @@ os.system('pip install llama-cpp-python transformers torch')
 
 from llama_cpp import Llama
 from transformers import AutoTokenizer
+import torch
 from transformers.generation.logits_process import ExponentialDecayLengthPenalty
 model_id = "Elijahbodden/eliGPTv1.1"
 
@@ -98,7 +99,7 @@ def respond(
         max_tokens=128,
         frequency_penalty=frequency_penalty,
         presence_penalty=presence_penalty,
-        logits_processor=ExponentialDecayLengthPenalty((lp_start, lp_decay), tokenizer.eos_token_id, len(convo))
+        logits_processor=lambda ids, logits: ExponentialDecayLengthPenalty((lp_start, lp_decay), tokenizer.eos_token_id, len(convo))(ids, torch.from_numpy(logits))
     ):
         token = message["choices"][0]["text"]