vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-11 01:24:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

horde workers pause themselves if recent local usage is detected (+1 squashed commits)

Browse source 
Squashed commits:

[7ebb80bc] horde workers pause themselves if recent local usage is detected
This commit is contained in:
Concedo 2024-04-07 23:15:25 +08:00
parent 81ac0e5656
commit aa5124439d
1 changed files with 23 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

23
koboldcpp.py
View file

@ -644,6 +644,7 @@ sslvalid = False
nocertify = False nocertify = False
start_time = time.time() start_time = time.time()
last_req_time = time.time() last_req_time = time.time()
last_non_horde_req_time = time.time()
class ServerRequestHandler(http.server.SimpleHTTPRequestHandler): class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
sys_version = "" sys_version = ""
@ -665,6 +666,7 @@ class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
pass pass
async def generate_text(self, genparams, api_format, stream_flag): async def generate_text(self, genparams, api_format, stream_flag):
from datetime import datetime
global friendlymodelname, chatcompl_adapter global friendlymodelname, chatcompl_adapter
is_quiet = args.quiet is_quiet = args.quiet
def run_blocking(): #api format 1=basic,2=kai,3=oai,4=oai-chat def run_blocking(): #api format 1=basic,2=kai,3=oai,4=oai-chat
@ -749,6 +751,12 @@ class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
genparams["max_length"] = 32 genparams["max_length"] = 32
genparams["prompt"] = "### Instruction: In one sentence, write a descriptive caption for this image.\n### Response:" genparams["prompt"] = "### Instruction: In one sentence, write a descriptive caption for this image.\n### Response:"
#flag instance as non-idle for a while
washordereq = genparams.get('genkey', '').startswith('HORDEREQ_')
if not washordereq:
global last_non_horde_req_time
last_non_horde_req_time = time.time()
return generate( return generate(
prompt=genparams.get('prompt', ""), prompt=genparams.get('prompt', ""),
memory=genparams.get('memory', ""), memory=genparams.get('memory', ""),
@ -792,6 +800,12 @@ class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
else: else:
recvtxt = run_blocking() recvtxt = run_blocking()
#flag instance as non-idle for a while
washordereq = genparams.get('genkey', '').startswith('HORDEREQ_')
if not washordereq:
global last_non_horde_req_time
last_non_horde_req_time = time.time()
if (args.debugmode != -1 and not is_quiet) or args.debugmode >= 1: if (args.debugmode != -1 and not is_quiet) or args.debugmode >= 1:
utfprint("\nOutput: " + recvtxt) utfprint("\nOutput: " + recvtxt)
@ -2469,6 +2483,14 @@ def run_horde_worker(args, api_key, worker_name):
else: else:
print_with_time(f"Horde Worker Exit limit reached, too many errors.") print_with_time(f"Horde Worker Exit limit reached, too many errors.")
global last_non_horde_req_time
sec_since_non_horde = time.time() - last_non_horde_req_time
no_recent_local_usage = sec_since_non_horde>20
if not no_recent_local_usage:
#print_with_time(f"Recent Local Usage - Horde Worker Waiting...")
time.sleep(1)
continue
#first, make sure we are not generating #first, make sure we are not generating
if modelbusy.locked(): if modelbusy.locked():
time.sleep(0.2) time.sleep(0.2)
@ -2516,6 +2538,7 @@ def run_horde_worker(args, api_key, worker_name):
currentjob_attempts += 1 currentjob_attempts += 1
if currentjob_attempts>5: if currentjob_attempts>5:
break break
print_with_time(f"Server Busy - Not ready to generate...") print_with_time(f"Server Busy - Not ready to generate...")
time.sleep(5) time.sleep(5)