mmarch/text-generation-webui
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Always return only the new tokens in generation functions

Browse source
This commit is contained in:
oobabooga 2023-05-11 17:07:20 -03:00
parent c4f0e6d740
commit 0d36c18f5d
4 changed files with 16 additions and 25 deletions
Download patch file Download diff file Expand all files Collapse all files

12
extensions/openai/script.py
View file

@ -340,17 +340,14 @@ class Handler(BaseHTTPRequestHandler):
# generate reply #######################################
if debug:
print({'prompt': prompt, 'req_params': req_params, 'stopping_strings': stopping_strings})
generator = generate_reply(prompt, req_params, stopping_strings=stopping_strings, is_chat=True)
generator = generate_reply(prompt, req_params, stopping_strings=stopping_strings, is_chat=False)
answer = ''
seen_content = ''
longest_stop_len = max([len(x) for x in stopping_strings])
for a in generator:
if isinstance(a, str):
answer = a
else:
answer = a[0]
answer = a
stop_string_found = False
len_seen = len(seen_content)
@ -521,10 +518,7 @@ class Handler(BaseHTTPRequestHandler):
answer = ''
for a in generator:
if isinstance(a, str):
answer = a
else:
answer = a[0]
answer = a
completion_token_count = len(encode(answer)[0])