log debug temp

custom_generate/generate.py

@@ -41,11 +41,14 @@ def generate(model, input_ids, generation_config=None, n_sigma:float=1.0, **kwar
                                       and max_new_tokens.
         n_sigma (float): The number of standard deviations to use for topN-sigma sampling.
         **kwargs: Additional keyword arguments.
-    """    
+    """
     generation_config = generation_config or model.generation_config  # default to the model generation config
     cur_length = input_ids.shape[1]
-    max_length = generation_config.max_length or cur_length + generation_config.max_new_tokens
-    print(f"Starting generation with max_length: {max_length}, current length: {cur_length}")
+    if generation_config.max_length is None:
+        max_length = cur_length + generation_config.max_new_tokens
+    else:
+        max_length = generation_config.max_length
+    print(f"Starting generation with max_length: {max_length}, current length: {cur_length} and temperature: {generation_config.temperature}")
 
     while cur_length < max_length:
         logits = model(input_ids).logits
@@ -56,6 +59,5 @@ def generate(model, input_ids, generation_config=None, n_sigma:float=1.0, **kwar
         next_tokens = torch.multinomial(torch.softmax(filtered_logits, dim=-1), num_samples=1)
         input_ids = torch.cat((input_ids, next_tokens), dim=-1)
         cur_length += 1
-        print(f"Current length: {cur_length}, Next token: {next_tokens.item()}")
 
     return input_ids