Mirrors/text-generation-webui
1
0
Fork 0
mirror of https://github.com/oobabooga/text-generation-webui.git synced 2024-11-22 08:07:56 +01:00
Code Issues Packages Projects Releases Wiki Activity

Add greed parameter

Browse Source
This commit is contained in:
oobabooga 2023-02-25 01:31:01 -03:00
parent 13f2688134
commit 2dfb999bf1
2 changed files with 31 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
modules/chat.py
View File

@ -84,7 +84,7 @@ def extract_message_from_reply(question, reply, current, other, check, extension
def stop_everything_event():
shared.stop_everything = True
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):
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, greed=1):
shared.stop_everything = False
just_started = True
eos_token = '\n' if check else None
@ -112,7 +112,9 @@ def chatbot_wrapper(text, max_new_tokens, do_sample, temperature, top_p, typical
prompt = custom_prompt_generator(text, max_new_tokens, name1, name2, context, chat_prompt_size)
# Generate
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}:"):
reply = ''
for i in range(greed):
for reply in generate_reply(prompt+reply, 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)
@ -136,6 +138,7 @@ def chatbot_wrapper(text, max_new_tokens, do_sample, temperature, top_p, typical
if next_character_found:
break
yield shared.history['visible']
print(i, reply)
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):
eos_token = '\n' if check else None
@ -153,11 +156,11 @@ def impersonate_wrapper(text, max_new_tokens, do_sample, temperature, top_p, typ
break
yield reply
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):
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):
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, greed=1):
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, greed):
yield generate_chat_html(_history, name1, name2, shared.character)
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):
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, greed=1):
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)
@ -167,7 +170,7 @@ def regenerate_wrapper(text, max_new_tokens, do_sample, temperature, top_p, typi
last_visible = shared.history['visible'].pop()
last_internal = shared.history['internal'].pop()
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):
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, greed):
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)

3
server.py
View File

@ -241,9 +241,10 @@ if shared.args.chat or shared.args.cai_chat:
shared.gradio['max_new_tokens'] = gr.Slider(minimum=shared.settings['max_new_tokens_min'], maximum=shared.settings['max_new_tokens_max'], step=1, label='max_new_tokens', value=shared.settings['max_new_tokens'])
with gr.Column():
shared.gradio['chat_prompt_size_slider'] = gr.Slider(minimum=shared.settings['chat_prompt_size_min'], maximum=shared.settings['chat_prompt_size_max'], step=1, label='Maximum prompt size in tokens', value=shared.settings['chat_prompt_size'])
shared.gradio['greed'] = gr.Slider(minimum=1, maximum=5, value=1, step=1)
create_settings_menus()
shared.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', 'name1', 'name2', 'context', 'check', 'chat_prompt_size_slider']]
shared.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', 'name1', 'name2', 'context', 'check', 'chat_prompt_size_slider', 'greed']]
if shared.args.extensions is not None:
with gr.Tab('Extensions'):
extensions_module.create_extensions_block()