Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-18 16:55:14 +02:00
Code Issues Packages Projects Releases Wiki Activity

alias for draft

Browse Source
This commit is contained in:
Concedo
2026-06-23 18:43:28 +08:00
parent 7fe6fa6fb6
commit 4a7d6dd8a0
2 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
koboldcpp.py
+1 -1
View File
@@ -12177,7 +12177,7 @@ if __name__ == '__main__':
advparser.add_argument("--defaultgenamt", help="How many tokens to generate by default, if not specified. Must be smaller than context size. Usually, your frontend GUI will override this.", type=check_range(int,64,16384), default=default_genlen)
advparser.add_argument("--device", "-dev", metavar=('<dev1,dev2,..>'), help="Set llama.cpp compatible device selection override. Comma separated. Overrides normal device choices.", default="")
advparser.add_argument("--downloaddir", metavar=('[directory]'), help="Specify a directory that models will be downloaded to or searched from, if unset uses the working directory.", default="")
advparser.add_argument("--draftamount","--draft-max","--draft-n", metavar=('[tokens]'), help="How many tokens to draft per chunk before verifying results", type=int, default=default_draft_amount)
advparser.add_argument("--draftamount","--draft-max","--draft-n","--spec-draft-n-max", metavar=('[tokens]'), help="How many tokens to draft per chunk before verifying results", type=int, default=default_draft_amount)
advparser.add_argument("--draftgpulayers","--gpu-layers-draft","--n-gpu-layers-draft","-ngld", metavar=('[layers]'), help="How many layers to offload to GPU for the draft model (default=full offload)", type=int, default=999)
advparser.add_argument("--draftgpusplit", help="GPU layer distribution ratio for draft model (default=same as main). Only works if multi-GPUs selected for MAIN model and tensor_split is set!", metavar=('[Ratios]'), type=float, nargs='+')
advparser.add_argument("--draftmodel","--model-draft","-md", metavar=('[filename]'), help="Load a small draft model for speculative decoding. It will be fully offloaded. Vocab must match the main model.", default="")