vrr/kvcache-ai-ktransformers
2
0
Fork 0
mirror of https://github.com/kvcache-ai/ktransformers.git synced 2026-04-29 20:29:48 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 6

Kt minimax (#1742)

Browse source 
[feat]: fp8 kernel and kt-cli support
This commit is contained in:
ErvinXie 2025-12-24 15:39:44 +08:00 • committed by GitHub
parent e7d277d163
commit d8046e1bb4
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
65 changed files with 12111 additions and 2502 deletions
Download patch file Download diff file Expand all files Collapse all files

153
kt-kernel/python/cli/completions/_kt Normal file
View file

@ -0,0 +1,153 @@
#compdef kt
# Zsh completion for kt command
# This is a static completion script that doesn't require Python startup
_kt() {
local -a commands
commands=(
'version:Show version information'
'run:Start model inference server'
'chat:Interactive chat with running model'
'quant:Quantize model weights'
'bench:Run full benchmark'
'microbench:Run micro-benchmark'
'doctor:Diagnose environment issues'
'model:Manage models and storage paths'
'config:Manage configuration'
'sft:Fine-tuning with LlamaFactory'
)
local -a run_opts
run_opts=(
'--host[Server host]:host:'
'--port[Server port]:port:'
'--gpu-experts[Number of GPU experts]:count:'
'--cpu-threads[Number of CPU threads]:count:'
'--tensor-parallel-size[Tensor parallel size]:size:'
'--kt-method[KT method]:method:(AMXINT4 FP8 RAWINT4)'
'--attention-backend[Attention backend]:backend:(triton flashinfer)'
'--max-total-tokens[Maximum total tokens]:tokens:'
'--dry-run[Show command without executing]'
'--help[Show help message]'
)
local -a chat_opts
chat_opts=(
'--host[Server host]:host:'
'--port[Server port]:port:'
'--model[Model name]:model:'
'--temperature[Sampling temperature]:temp:'
'--max-tokens[Maximum tokens]:tokens:'
'--system[System prompt]:prompt:'
'--save-history[Save conversation history]'
'--no-save-history[Do not save history]'
'--history-file[History file path]:path:_files'
'--stream[Enable streaming output]'
'--no-stream[Disable streaming output]'
'--help[Show help message]'
)
local -a model_cmds
model_cmds=(
'download:Download a model from HuggingFace'
'list:List available models'
'path-list:List all model storage paths'
'path-add:Add a new model storage path'
'path-remove:Remove a model storage path'
'search:Search for models in the registry'
)
local -a config_cmds
config_cmds=(
'show:Show all configuration'
'get:Get configuration value'
'set:Set configuration value'
'reset:Reset to defaults'
'path:Show configuration file path'
'init:Re-run first-time setup wizard'
)
local -a sft_cmds
sft_cmds=(
'train:Train model'
'chat:Chat with model'
'export:Export model'
)
_arguments -C \
'1: :->command' \
'*::arg:->args'
case $state in
command)
_describe 'kt commands' commands
_arguments \
'--help[Show help message]' \
'--version[Show version]'
;;
args)
case $words[1] in
run)
_arguments $run_opts \
'1:model:'
;;
chat)
_arguments $chat_opts
;;
quant)
_arguments \
'--method[Quantization method]:method:' \
'--output[Output directory]:path:_files -/' \
'--help[Show help message]' \
'1:model:_files -/'
;;
bench|microbench)
_arguments \
'--model[Model name or path]:model:' \
'--config[Config file path]:path:_files' \
'--help[Show help message]'
;;
doctor)
_arguments \
'--verbose[Verbose output]' \
'--help[Show help message]'
;;
model)
_arguments \
'1: :->model_cmd' \
'*::arg:->model_args'
case $state in
model_cmd)
_describe 'model commands' model_cmds
;;
esac
;;
config)
_arguments \
'1: :->config_cmd' \
'*::arg:->config_args'
case $state in
config_cmd)
_describe 'config commands' config_cmds
;;
esac
;;
sft)
_arguments \
'1: :->sft_cmd' \
'*::arg:->sft_args'
case $state in
sft_cmd)
_describe 'sft commands' sft_cmds
;;
esac
;;
esac
;;
esac
}
_kt "$@"