Skip to content

fix: cuda availability checker issue over web docker service with no … #89

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
IliasHad merged 1 commit into from
Feb 25, 2026
Merged

fix: cuda availability checker issue over web docker service with no … #89

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 4 additions & 4 deletions packages/embedding-core/src/services/extractors.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,7 +41,7 @@ export async function getFrameExtractor() {

const processor = await AutoProcessor.from_pretrained(VISUAL_EMBEDDING_MODEL)
const model = await CLIPVisionModelWithProjection.from_pretrained(VISUAL_EMBEDDING_MODEL, {
device: USE_GPU ? "cuda" : "auto",
device: USE_GPU ? "cuda" : "cpu",
dtype: "fp16"
})
visualModelCache = { processor, model }
Expand All @@ -53,7 +53,7 @@ export async function getAudioExtractor() {
if (!audioModelCache) {
const processor = await AutoProcessor.from_pretrained(AUDIO_EMBEDDING_MODEL)
const model = await ClapAudioModelWithProjection.from_pretrained(AUDIO_EMBEDDING_MODEL, {
device: USE_GPU ? "cuda" : "auto"
device: USE_GPU ? "cuda" : "cpu"
})

audioModelCache = { processor, model }
Expand All @@ -66,7 +66,7 @@ async function getTextToVisualExtractor() {

const tokenizer = await AutoTokenizer.from_pretrained(VISUAL_EMBEDDING_MODEL)
const model = await CLIPTextModelWithProjection.from_pretrained(VISUAL_EMBEDDING_MODEL, {
device: USE_GPU ? "cuda" : "auto",
device: USE_GPU ? "cuda" : "cpu",
dtype: "fp16"
})
textToVisualModelCache = { tokenizer, model }
Expand Down Expand Up @@ -122,7 +122,7 @@ export async function getTextExtractor() {
if (!textModelCache) {

const embed = await pipeline('feature-extraction', TEXT_EMBEDDING_MODEL, {
device: USE_GPU ? "cuda" : "auto",
device: USE_GPU ? "cuda" : "cpu",
dtype: "fp16"
})

Expand Down
11 changes: 10 additions & 1 deletion packages/shared/src/constants/gpu.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,12 @@
import { logger } from "@shared/services/logger";
import { isGPUAvailable } from "@shared/utils/gpu";

export const USE_GPU = isGPUAvailable()
export let USE_GPU = false;

(async function initGPU(): Promise<void> {
try {
USE_GPU = await isGPUAvailable();
} catch (error) {
logger.error({ error }, 'Failed to initialize GPU — defaulting to CPU');
}
})()
25 changes: 19 additions & 6 deletions packages/shared/src/utils/gpu.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,11 +1,24 @@
import { execSync } from 'child_process';
import { logger } from '@shared/services/logger';

let _gpuAvailable: boolean | null = null;

const MINIMAL_ONNX_MODEL = Buffer.from([
0x08, 0x07, 0x12, 0x00, 0x2a, 0x00, 0x3a, 0x00,
]);

export async function isGPUAvailable(): Promise<boolean> {
if (_gpuAvailable !== null) return _gpuAvailable;

export function isGPUAvailable(): boolean {
try {
// Check if nvidia-smi is available
execSync('nvidia-smi', { stdio: 'ignore' });
return true;
const ort = await import('onnxruntime-node');
await ort.InferenceSession.create(MINIMAL_ONNX_MODEL, {
executionProviders: ['CUDAExecutionProvider'],
});
_gpuAvailable = true;
logger.info('GPU available');
} catch {
return false;
_gpuAvailable = false;
logger.info('GPU unavailable — fallback to CPU');
}
return _gpuAvailable;
}