debug

modules/text_generation.py

@@ -91,6 +91,9 @@ def clear_torch_cache():
 
 def generate_reply(question, max_new_tokens, do_sample, temperature, top_p, typical_p, repetition_penalty, top_k, min_length, no_repeat_ngram_size, num_beams, penalty_alpha, length_penalty, early_stopping, eos_token=None, stopping_string=None):
     clear_torch_cache()
+    print(question)
+    print(eos_token)
+    print(stopping_string)
     t0 = time.time()
 
     # These models are not part of Hugging Face, so we handle them
@@ -234,5 +237,6 @@ def generate_reply(question, max_new_tokens, do_sample, temperature, top_p, typi
 
     finally:
         t1 = time.time()
+        print(output)
         print(f"Output generated in {(t1-t0):.2f} seconds ({(len(output)-len(original_input_ids[0]))/(t1-t0):.2f} tokens/s, {len(output)-len(original_input_ids[0])} tokens)")
         return