increase debounce time for hf seaches to 600ms, updates for latest kcpp

This commit is contained in:
lone-cloud 2026-01-03 19:43:49 -08:00
parent 2395dc4c9e
commit ad5347c36d
6 changed files with 44 additions and 65 deletions

File diff suppressed because one or more lines are too long

Binary file not shown.

View file

@ -1,7 +1,7 @@
{ {
"name": "gerbil", "name": "gerbil",
"productName": "Gerbil", "productName": "Gerbil",
"version": "1.18.3", "version": "1.18.4",
"description": "Run Large Language Models locally", "description": "Run Large Language Models locally",
"main": "out/main/index.js", "main": "out/main/index.js",
"homepage": "./", "homepage": "./",
@ -58,7 +58,7 @@
"rehype-raw": "^7.0.0", "rehype-raw": "^7.0.0",
"rehype-sanitize": "^6.0.0", "rehype-sanitize": "^6.0.0",
"remark-gfm": "^4.0.1", "remark-gfm": "^4.0.1",
"systeminformation": "^5.28.7", "systeminformation": "^5.28.10",
"winston": "^3.19.0", "winston": "^3.19.0",
"winston-daily-rotate-file": "^5.0.0", "winston-daily-rotate-file": "^5.0.0",
"yauzl": "^3.2.0", "yauzl": "^3.2.0",
@ -85,7 +85,7 @@
"eslint-plugin-react": "^7.37.5", "eslint-plugin-react": "^7.37.5",
"eslint-plugin-react-hooks": "^7.0.1", "eslint-plugin-react-hooks": "^7.0.1",
"eslint-plugin-sonarjs": "^3.0.5", "eslint-plugin-sonarjs": "^3.0.5",
"globals": "^16.5.0", "globals": "^17.0.0",
"jiti": "^2.6.1", "jiti": "^2.6.1",
"prettier": "^3.7.4", "prettier": "^3.7.4",
"rollup-plugin-visualizer": "^6.0.5", "rollup-plugin-visualizer": "^6.0.5",

View file

@ -69,6 +69,9 @@ const UI_COVERED_ARGS = new Set([
'--lowvram', '--lowvram',
'--smartcache', '--smartcache',
'--pipelineparallel', '--pipelineparallel',
'--quantkv',
'--usecpu',
'--autofit',
] as const) as ReadonlySet<string>; ] as const) as ReadonlySet<string>;
const IGNORED_ARGS = new Set([ const IGNORED_ARGS = new Set([
@ -184,7 +187,7 @@ const COMMAND_LINE_ARGUMENTS = [
{ {
flag: '--ropeconfig', flag: '--ropeconfig',
description: description:
'If set, uses customized RoPE scaling from configured frequency scale and frequency base (e.g. --ropeconfig 0.25 10000). Otherwise, uses NTK-Aware scaling set automatically based on context size.', 'If set, uses customized RoPE scaling from configured frequency scale and frequency base (e.g. --ropeconfig 0.25 10000). Otherwise, uses NTK-Aware scaling set automatically based on context size. For linear rope, simply set the freq-scale and ignore the freq-base',
metavar: '[rope-freq-scale] [rope-freq-base]', metavar: '[rope-freq-scale] [rope-freq-base]',
default: '0.0 10000.0', default: '0.0 10000.0',
type: 'float[]', type: 'float[]',
@ -365,16 +368,6 @@ const COMMAND_LINE_ARGUMENTS = [
type: 'float[]', type: 'float[]',
category: 'Speculative Decoding', category: 'Speculative Decoding',
}, },
{
flag: '--quantkv',
description:
'Sets the KV cache data type quantization, 0=f16, 1=q8, 2=q4. Requires Flash Attention for full effect, otherwise only K cache is quantized.',
metavar: '[quantization level 0/1/2]',
type: 'int',
choices: ['0', '1', '2'],
default: 0,
category: 'Performance',
},
{ {
flag: '--defaultgenamt', flag: '--defaultgenamt',
description: description:
@ -383,20 +376,6 @@ const COMMAND_LINE_ARGUMENTS = [
default: 896, default: 896,
category: 'Performance', category: 'Performance',
}, },
{
flag: '--smartcache',
description:
'Enables intelligent context switching by saving KV cache snapshots to RAM. Requires fast forwarding.',
type: 'boolean',
category: 'Performance',
},
{
flag: '--pipelineparallel',
description:
'Enable Pipeline Parallelism for faster multigpu speeds but using more memory, only active for multigpu.',
type: 'boolean',
category: 'Performance',
},
{ {
flag: '--nobostoken', flag: '--nobostoken',
description: description:

View file

@ -43,7 +43,7 @@ export const HuggingFaceSearchModal = ({
const [searchQuery, setSearchQuery] = useState( const [searchQuery, setSearchQuery] = useState(
initialSearchParams.search || '' initialSearchParams.search || ''
); );
const [debouncedQuery] = useDebouncedValue(searchQuery, 300); const [debouncedQuery] = useDebouncedValue(searchQuery, 600);
const scrollAreaRef = useRef<HTMLDivElement>(null); const scrollAreaRef = useRef<HTMLDivElement>(null);
const prevOpenedRef = useRef(false); const prevOpenedRef = useRef(false);

View file

@ -4152,7 +4152,7 @@ __metadata:
eslint-plugin-react-hooks: "npm:^7.0.1" eslint-plugin-react-hooks: "npm:^7.0.1"
eslint-plugin-sonarjs: "npm:^3.0.5" eslint-plugin-sonarjs: "npm:^3.0.5"
execa: "npm:^9.6.1" execa: "npm:^9.6.1"
globals: "npm:^16.5.0" globals: "npm:^17.0.0"
jiti: "npm:^2.6.1" jiti: "npm:^2.6.1"
lucide-react: "npm:^0.562.0" lucide-react: "npm:^0.562.0"
mime-types: "npm:^3.0.2" mime-types: "npm:^3.0.2"
@ -4165,7 +4165,7 @@ __metadata:
rehype-sanitize: "npm:^6.0.0" rehype-sanitize: "npm:^6.0.0"
remark-gfm: "npm:^4.0.1" remark-gfm: "npm:^4.0.1"
rollup-plugin-visualizer: "npm:^6.0.5" rollup-plugin-visualizer: "npm:^6.0.5"
systeminformation: "npm:^5.28.7" systeminformation: "npm:^5.28.10"
typescript: "npm:^5.9.3" typescript: "npm:^5.9.3"
vite: "npm:^7.3.0" vite: "npm:^7.3.0"
winston: "npm:^3.19.0" winston: "npm:^3.19.0"
@ -4310,10 +4310,10 @@ __metadata:
languageName: node languageName: node
linkType: hard linkType: hard
"globals@npm:^16.5.0": "globals@npm:^17.0.0":
version: 16.5.0 version: 17.0.0
resolution: "globals@npm:16.5.0" resolution: "globals@npm:17.0.0"
checksum: 10c0/615241dae7851c8012f5aa0223005b1ed6607713d6813de0741768bd4ddc39353117648f1a7086b4b0fa45eae733f1c0a0fe369aa4e543bb63f8de8990178ea9 checksum: 10c0/e3c169fdcb0fc6755707b697afb367bea483eb29992cfc0de1637382eb893146e17f8f96db6d7453e3696b478a7863ae2000e6c71cd2f4061410285106d3847a
languageName: node languageName: node
linkType: hard linkType: hard
@ -7839,12 +7839,12 @@ __metadata:
languageName: node languageName: node
linkType: hard linkType: hard
"systeminformation@npm:^5.28.7": "systeminformation@npm:^5.28.10":
version: 5.28.7 version: 5.28.10
resolution: "systeminformation@npm:5.28.7" resolution: "systeminformation@npm:5.28.10"
bin: bin:
systeminformation: lib/cli.js systeminformation: lib/cli.js
checksum: 10c0/dbf7fd3f893ad84032399e4c81147715933f80e3b11bc0c2c59a14948be488404a0de49407253bcc1b9fae6ef2b8c25a9798ed4424158c70730a44e4b138aa71 checksum: 10c0/c0396edac846a3a15750c3299434eca5fa782b494a6bf185f075292de2556bb17ab2dd95883f32bd9ab6d3a7e90df7c28716eb1b19c5994cc4bab4fb3d884325
conditions: (os=darwin | os=linux | os=win32 | os=freebsd | os=openbsd | os=netbsd | os=sunos | os=android) conditions: (os=darwin | os=linux | os=win32 | os=freebsd | os=openbsd | os=netbsd | os=sunos | os=android)
languageName: node languageName: node
linkType: hard linkType: hard