vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

SSE emoji fix

Browse source
This commit is contained in:
Concedo 2023-07-30 22:31:20 +08:00
parent 82d0695f0f
commit 5ea5d19d6a
1 changed files with 9 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

13
koboldcpp.py
View file

@ -400,6 +400,7 @@ class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
current_token = 0 current_token = 0
incomplete_token_buffer = bytearray()
while not handle.has_finished(): while not handle.has_finished():
if current_token < handle.get_stream_count(): if current_token < handle.get_stream_count():
token = handle.new_token(current_token) token = handle.new_token(current_token)
@ -409,10 +410,14 @@ class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
current_token += 1 current_token += 1
tokenStr = ctypes.string_at(token).decode("UTF-8","ignore") newbyte = ctypes.string_at(token)
event_data = {"token": tokenStr} incomplete_token_buffer += bytearray(newbyte)
event_str = json.dumps(event_data) tokenStr = incomplete_token_buffer.decode("UTF-8","ignore")
await self.send_sse_event("message", event_str) if tokenStr!="":
incomplete_token_buffer.clear()
event_data = {"token": tokenStr}
event_str = json.dumps(event_data)
await self.send_sse_event("message", event_str)
await asyncio.sleep(0) await asyncio.sleep(0)