Skip to content

Commit a3aa5cd

Browse files
authored
use cl100k_base as the default tiktoken encoding (#2974)
Signed-off-by: bjwswang <[email protected]>
1 parent dc26514 commit a3aa5cd

File tree

1 file changed

+12
-2
lines changed

1 file changed

+12
-2
lines changed

fastchat/serve/openai_api_server.py

Lines changed: 12 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -229,10 +229,20 @@ def process_input(model_name, inp):
229229
inp = [inp]
230230
elif isinstance(inp, list):
231231
if isinstance(inp[0], int):
232-
decoding = tiktoken.model.encoding_for_model(model_name)
232+
try:
233+
decoding = tiktoken.model.encoding_for_model(model_name)
234+
except KeyError:
235+
logger.warning("Warning: model not found. Using cl100k_base encoding.")
236+
model = "cl100k_base"
237+
decoding = tiktoken.get_encoding(model)
233238
inp = [decoding.decode(inp)]
234239
elif isinstance(inp[0], list):
235-
decoding = tiktoken.model.encoding_for_model(model_name)
240+
try:
241+
decoding = tiktoken.model.encoding_for_model(model_name)
242+
except KeyError:
243+
logger.warning("Warning: model not found. Using cl100k_base encoding.")
244+
model = "cl100k_base"
245+
decoding = tiktoken.get_encoding(model)
236246
inp = [decoding.decode(text) for text in inp]
237247

238248
return inp

0 commit comments

Comments
 (0)