Skip to content

[Feature]: Support for gfx1151 #132

@flowtyone

Description

@flowtyone

Suggestion Description

I couldn't find anywhere if iGPU is supported by this implementation, but it would be nice if we could use it with Strix Halo AI iGPUs. Currently when trying the nightly builds, I'm getting this error from XLA:

Traceback (most recent call last):
  File "/home/u/Desktop/jax-test/.venv/lib/python3.11/site-packages/jax/_src/xla_bridge.py", line 812, in backends
    backend = _init_backend(platform)
              ^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/u/Desktop/jax-test/.venv/lib/python3.11/site-packages/jax/_src/xla_bridge.py", line 896, in _init_backend
    backend = registration.factory()
              ^^^^^^^^^^^^^^^^^^^^^^
  File "/home/u/Desktop/jax-test/.venv/lib/python3.11/site-packages/jax/_src/xla_bridge.py", line 549, in make_pjrt_c_api_client
    return xla_client.make_c_api_client(plugin_name, updated_options, None)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/u/Desktop/jax-test/.venv/lib/python3.11/site-packages/jaxlib/xla_client.py", line 156, in make_c_api_client
    return _xla.get_c_api_client(
           ^^^^^^^^^^^^^^^^^^^^^^
jaxlib._jax.XlaRuntimeError: INTERNAL: no supported devices found for platform ROCM

Operating System

Ubuntu 24.04

GPU

AMD Ryzen™ Al Max+ 395 8060s

ROCm Component

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    Status

    Todo

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions