Skip to content

Add Wasm fallback while running llms #97

Closed
#302
Closed
Add Wasm fallback while running llms#97
#302
Assignees
sauravpanda

Description

@sauravpanda
sauravpanda
opened on Feb 9, 2025

Currently we use MLC which needs webgpu and also force webgpu in TTS and whisper. But we can fallback to onnx models and wasm device mode if we cant find the relevant items.

Activity

  1. sauravpanda
    self-assigned this
    on Feb 9, 2025
  2. sauravpanda
    added a commit that references this issue on Apr 15, 2026

    fix: WebGPU fallback, Gemma 3 1B registry, mobile/desktop HTML dedup

    c31653a

  3. sauravpanda commented on Apr 15, 2026

    @sauravpanda
    sauravpanda
    on Apr 15, 2026
    OwnerAuthor

    Fixed in #302. TransformersEngineWrapper now calls detectBestDevice() before loading — probes navigator.gpu.requestAdapter() and falls back to 'cpu' (ONNX WASM) automatically when WebGPU is unavailable. No API changes; callers can still pass options.device to override.

  4. sauravpanda
    closed this as completedon Apr 15, 2026
  5. sauravpanda
    added a commit that references this issue on Apr 16, 2026

    fix: WebGPU CPU fallback, Gemma 3 1B registry, HTML mobile dedup (#97 #…

    e970683
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

No labels
No labels

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions