Remove duplicate max_new_tokens parameter

modules/chat.py

@@ -24,16 +24,16 @@ def clean_chat_message(text):
     text = text.strip()
     return text
 
-def generate_chat_prompt(user_input, tokens, name1, name2, context, chat_prompt_size, impersonate=False):
+def generate_chat_prompt(user_input, max_new_tokens, name1, name2, context, chat_prompt_size, impersonate=False):
     user_input = clean_chat_message(user_input)
     rows = [f"{context.strip()}\n"]
 
     if shared.soft_prompt:
        chat_prompt_size -= shared.soft_prompt_tensor.shape[1]
-    max_length = min(get_max_prompt_length(tokens), chat_prompt_size)
+    max_length = min(get_max_prompt_length(max_new_tokens), chat_prompt_size)
 
     i = len(shared.history['internal'])-1
-    while i >= 0 and len(encode(''.join(rows), tokens)[0]) < max_length:
+    while i >= 0 and len(encode(''.join(rows), max_new_tokens)[0]) < max_length:
         rows.insert(1, f"{name2}: {shared.history['internal'][i][1].strip()}\n")
         if not (shared.history['internal'][i][0] == '<|BEGIN-VISIBLE-CHAT|>'):
             rows.insert(1, f"{name1}: {shared.history['internal'][i][0].strip()}\n")
@@ -47,7 +47,7 @@ def generate_chat_prompt(user_input, tokens, name1, name2, context, chat_prompt_
         rows.append(f"{name1}:")
         limit = 2
 
-    while len(rows) > limit and len(encode(''.join(rows), tokens)[0]) >= max_length:
+    while len(rows) > limit and len(encode(''.join(rows), max_new_tokens)[0]) >= max_length:
         rows.pop(1)
 
     prompt = ''.join(rows)
@@ -95,7 +95,7 @@ def generate_chat_picture(picture, name1, name2):
 def stop_everything_event():
     shared.stop_everything = True
 
-def chatbot_wrapper(text, tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, min_length, no_repeat_ngram_size, num_beams, penalty_alpha, length_penalty, early_stopping, name1, name2, context, check, chat_prompt_size, picture=None):
+def chatbot_wrapper(text, 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, name1, name2, context, check, chat_prompt_size, picture=None):
     shared.stop_everything = False
     just_started = True
     eos_token = '\n' if check else None
@@ -110,10 +110,10 @@ def chatbot_wrapper(text, tokens, do_sample, max_new_tokens, temperature, top_p,
         if shared.args.chat:
             visible_text = visible_text.replace('\n', '<br>')
     text = apply_extensions(text, "input")
-    prompt = generate_chat_prompt(text, tokens, name1, name2, context, chat_prompt_size)
+    prompt = generate_chat_prompt(text, max_new_tokens, name1, name2, context, chat_prompt_size)
 
     # Generate
-    for reply in generate_reply(prompt, tokens, do_sample, max_new_tokens, 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=eos_token, stopping_string=f"\n{name1}:"):
+    for reply in generate_reply(prompt, 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=eos_token, stopping_string=f"\n{name1}:"):
 
         # Extracting the reply
         reply, next_character_found, substring_found = extract_message_from_reply(prompt, reply, name2, name1, check, extensions=True)
@@ -138,15 +138,15 @@ def chatbot_wrapper(text, tokens, do_sample, max_new_tokens, temperature, top_p,
             break
     yield shared.history['visible']
 
-def impersonate_wrapper(text, tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, min_length, no_repeat_ngram_size, num_beams, penalty_alpha, length_penalty, early_stopping, name1, name2, context, check, chat_prompt_size, picture=None):
+def impersonate_wrapper(text, 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, name1, name2, context, check, chat_prompt_size, picture=None):
     eos_token = '\n' if check else None
 
     if 'pygmalion' in shared.model_name.lower():
         name1 = "You"
 
-    prompt = generate_chat_prompt(text, tokens, name1, name2, context, chat_prompt_size, impersonate=True)
+    prompt = generate_chat_prompt(text, max_new_tokens, name1, name2, context, chat_prompt_size, impersonate=True)
 
-    for reply in generate_reply(prompt, tokens, do_sample, max_new_tokens, 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=eos_token, stopping_string=f"\n{name2}:"):
+    for reply in generate_reply(prompt, 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=eos_token, stopping_string=f"\n{name2}:"):
         reply, next_character_found, substring_found = extract_message_from_reply(prompt, reply, name1, name2, check, extensions=False)
         if not substring_found:
             yield reply
@@ -154,11 +154,11 @@ def impersonate_wrapper(text, tokens, do_sample, max_new_tokens, temperature, to
             break
     yield reply
 
-def cai_chatbot_wrapper(text, tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, min_length, no_repeat_ngram_size, num_beams, penalty_alpha, length_penalty, early_stopping, name1, name2, context, check, chat_prompt_size, picture=None):
-    for _history in chatbot_wrapper(text, tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, min_length, no_repeat_ngram_size, num_beams, penalty_alpha, length_penalty, early_stopping, name1, name2, context, check, chat_prompt_size, picture):
+def cai_chatbot_wrapper(text, 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, name1, name2, context, check, chat_prompt_size, picture=None):
+    for _history in chatbot_wrapper(text, 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, name1, name2, context, check, chat_prompt_size, picture):
         yield generate_chat_html(_history, name1, name2, shared.character)
 
-def regenerate_wrapper(text, tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, min_length, no_repeat_ngram_size, num_beams, penalty_alpha, length_penalty, early_stopping, name1, name2, context, check, chat_prompt_size, picture=None):
+def regenerate_wrapper(text, 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, name1, name2, context, check, chat_prompt_size, picture=None):
     if shared.character != 'None' and len(shared.history['visible']) == 1:
         if shared.args.cai_chat:
             yield generate_chat_html(shared.history['visible'], name1, name2, shared.character)
@@ -168,7 +168,7 @@ def regenerate_wrapper(text, tokens, do_sample, max_new_tokens, temperature, top
         last_visible = shared.history['visible'].pop()
         last_internal = shared.history['internal'].pop()
 
-        for _history in chatbot_wrapper(last_internal[0], tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, min_length, no_repeat_ngram_size, num_beams, penalty_alpha, length_penalty, early_stopping, name1, name2, context, check, chat_prompt_size, picture):
+        for _history in chatbot_wrapper(last_internal[0], 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, name1, name2, context, check, chat_prompt_size, picture):
             if shared.args.cai_chat:
                 shared.history['visible'][-1] = [last_visible[0], _history[-1][1]]
                 yield generate_chat_html(shared.history['visible'], name1, name2, shared.character)

modules/text_generation.py

@@ -72,14 +72,14 @@ def formatted_outputs(reply, model_name):
     else:
         return reply
 
-def generate_reply(question, tokens, do_sample, max_new_tokens, 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):
+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):
     original_question = question
     if not (shared.args.chat or shared.args.cai_chat):
         question = apply_extensions(question, "input")
     if shared.args.verbose:
         print(f"\n\n{question}\n--------------------\n")
 
-    input_ids = encode(question, tokens)
+    input_ids = encode(question, max_new_tokens)
     cuda = "" if (shared.args.cpu or shared.args.deepspeed or shared.args.flexgen) else ".cuda()"
     if not shared.args.flexgen:
         n = shared.tokenizer.eos_token_id if eos_token is None else shared.tokenizer.encode(eos_token, return_tensors='pt')[0][-1]
@@ -126,7 +126,7 @@ def generate_reply(question, tokens, do_sample, max_new_tokens, temperature, top
     if shared.args.deepspeed:
         generate_params.append("synced_gpus=True")
     if shared.args.no_stream:
-        generate_params.append("max_new_tokens=tokens")
+        generate_params.append("max_new_tokens=max_new_tokens")
     else:
         generate_params.append("max_new_tokens=8")
 
@@ -156,7 +156,7 @@ def generate_reply(question, tokens, do_sample, max_new_tokens, temperature, top
     # Generate the reply 8 tokens at a time
     else:
         yield formatted_outputs(original_question, shared.model_name)
-        for i in tqdm(range(tokens//8+1)):
+        for i in tqdm(range(max_new_tokens//8+1)):
             with torch.no_grad():
                 output = eval(f"shared.model.generate({', '.join(generate_params)}){cuda}")[0]
             if shared.soft_prompt:

server.py

@@ -252,7 +252,7 @@ if shared.args.chat or shared.args.cai_chat:
             with gr.Tab("Extensions"):
                 extensions_module.create_extensions_block()
 
-        input_params = [shared.gradio[i] for i in ['textbox', 'max_new_tokens', 'do_sample', 'max_new_tokens', 'temperature', 'top_p', 'typical_p', 'repetition_penalty', 'top_k', 'min_length', 'no_repeat_ngram_size', 'num_beams', 'penalty_alpha', 'length_penalty', 'early_stopping', 'name1', 'name2', 'context', 'check', 'chat_prompt_size_slider']]
+        input_params = [shared.gradio[i] for i in ['textbox', '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', 'name1', 'name2', 'context', 'check', 'chat_prompt_size_slider']]
         if shared.args.picture:
             input_params.append(shared.gradio['picture_select'])
         function_call = "chat.cai_chatbot_wrapper" if shared.args.cai_chat else "chat.chatbot_wrapper"
@@ -312,7 +312,7 @@ elif shared.args.notebook:
         if shared.args.extensions is not None:
             extensions_module.create_extensions_block()
 
-        input_params = [shared.gradio[k] for k in ('textbox', 'max_new_tokens', 'do_sample', 'max_new_tokens', 'temperature', 'top_p', 'typical_p', 'repetition_penalty', 'top_k', 'min_length', 'no_repeat_ngram_size', 'num_beams', 'penalty_alpha', 'length_penalty', 'early_stopping')]
+        input_params = [shared.gradio[k] for k in ('textbox', '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')]
         output_params = [shared.gradio[k] for k in ["textbox", "markdown", "html"]]
         gen_events.append(shared.gradio["Generate"].click(generate_reply, input_params, output_params, show_progress=shared.args.no_stream, api_name="textgen"))
         gen_events.append(shared.gradio['textbox'].submit(generate_reply, input_params, output_params, show_progress=shared.args.no_stream))
@@ -344,7 +344,7 @@ else:
                 with gr.Tab('HTML'):
                     shared.gradio['html'] = gr.HTML()
 
-        input_params = [shared.gradio[k] for k in ['textbox', 'max_new_tokens', 'do_sample', 'max_new_tokens', 'temperature', 'top_p', 'typical_p', 'repetition_penalty', 'top_k', 'min_length', 'no_repeat_ngram_size', 'num_beams', 'penalty_alpha', 'length_penalty', 'early_stopping']]
+        input_params = [shared.gradio[k] for k in ['textbox', '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']]
         output_params = [shared.gradio[k] for k in ['output_textbox', 'markdown', 'html']]
         gen_events.append(shared.gradio['Generate'].click(generate_reply, input_params, output_params, show_progress=shared.args.no_stream, api_name="textgen"))
         gen_events.append(shared.gradio['textbox'].submit(generate_reply, input_params, output_params, show_progress=shared.args.no_stream))