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

allow 2048 blasbatchsize

Browse source
This commit is contained in:
Concedo 2023-05-27 17:47:18 +08:00
parent 92a0d77712
commit 94dc5c2324
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
koboldcpp.py
View file

@ -644,7 +644,7 @@ if __name__ == '__main__':
parser.add_argument("--psutil_set_threads", help="Experimental flag. If set, uses psutils to determine thread count based on physical cores.", action='store_true')
parser.add_argument("--highpriority", help="Experimental flag. If set, increases the process CPU priority, potentially speeding up generation. Use caution.", action='store_true')
parser.add_argument("--contextsize", help="Controls the memory allocated for maximum context size, only change if you need more RAM for big contexts. (default 2048)", type=int,choices=[512,1024,2048,4096,8192], default=2048)
parser.add_argument("--blasbatchsize", help="Sets the batch size used in BLAS processing (default 512)", type=int,choices=[32,64,128,256,512,1024], default=512)
parser.add_argument("--blasbatchsize", help="Sets the batch size used in BLAS processing (default 512, limit 256 for non-llama)", type=int,choices=[32,64,128,256,512,1024,2048], default=512)
parser.add_argument("--stream", help="Uses pseudo streaming when generating tokens. Only for the Kobold Lite UI.", action='store_true')
parser.add_argument("--smartcontext", help="Reserving a portion of context to try processing less frequently.", action='store_true')
parser.add_argument("--unbantokens", help="Normally, KoboldAI prevents certain tokens such as EOS and Square Brackets. This flag unbans them.", action='store_true')