Update cubecl wgpu v28

[laggui]

Checklist

• Confirmed that cargo run-checks command has been executed.

Related Issues/PRs

tracel-ai/cubecl#1119
tracel-ai/cubek#47

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 69.03%. Comparing base (9c175cf) to head (c0288d0).
⚠️ Report is 4 commits behind head on main.

❌ Your project check has failed because the head coverage (69.03%) is below the target coverage (80.00%). You can increase the head coverage or adjust the target coverage.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4244      +/-   ##
==========================================
+ Coverage   68.98%   69.03%   +0.04%     
==========================================
  Files        1350     1409      +59     
  Lines      165725   165880     +155     
==========================================
+ Hits       114332   114520     +188     
+ Misses      51393    51360      -33     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[bot-ember]

🏷️ Workflow test-gpu enabled 🟢
🕒 Auto-run scheduled on updates
ℹ A run has been scheduled... please wait for the results

[bot-ember]

🚀️ Workflow CI GPU (run #375) started!

[bot-ember]

❌ Workflow CI GPU:#4244 (run #375) failure!

[laggui]

❌ Workflow CI GPU:#4244 (run #375) failure!

Reporting failures below, but these are all unrelated to the current wgpu v28 changes.

wgpu failures

  failures:
      tests::fusion::ops::reshape::should_support_multiple_reshapes_cloned_tensor
      tests::fusion::quantization::ops::matmul::test_matmul_2d_aligned
      tests::fusion::quantization::ops::matmul::test_matmul_2d_aligned_fused
      tests::fusion::quantization::ops::matmul::test_matmul_lhs_float_rhs_quantized
      tests::fusion::quantization::ops::matmul::test_matmul_mixed_block_scale
      tests::tensor::module::attention::test_attention_no_mask
      tests::tensor::quantization::ops::matmul::test_matmul_2d_aligned
      tests::tensor::quantization::ops::matmul::test_matmul_2d_aligned_fused
      tests::tensor::quantization::ops::matmul::test_matmul_lhs_float_rhs_quantized
      tests::tensor::quantization::ops::matmul::test_matmul_mixed_block_scale
  
  test result: FAILED. 3006 passed; 10 failed; 36 ignored; 0 measured; 0 filtered out; finished in 184.76s

are existing failures in previous versions (see this GPU run for example)

vulkan failures

  failures:
      f16_i32::tests::fusion::ops::reshape::should_support_multiple_reshapes_cloned_tensor
      f16_i32::tests::fusion::quantization::ops::matmul::test_matmul_2d_aligned
      f16_i32::tests::fusion::quantization::ops::matmul::test_matmul_2d_aligned_fused
      f16_i32::tests::fusion::quantization::ops::matmul::test_matmul_lhs_float_rhs_quantized
      f16_i32::tests::fusion::quantization::ops::matmul::test_matmul_mixed_block_scale
      f16_i32::tests::tensor::module::attention::test_attention_no_mask
      f16_i32::tests::tensor::quantization::ops::matmul::test_matmul_2d_aligned
      f16_i32::tests::tensor::quantization::ops::matmul::test_matmul_2d_aligned_fused
      f16_i32::tests::tensor::quantization::ops::matmul::test_matmul_lhs_float_rhs_quantized
      f16_i32::tests::tensor::quantization::ops::matmul::test_matmul_mixed_block_scale
      tests::fusion::ops::reshape::should_support_multiple_reshapes_cloned_tensor
      tests::fusion::quantization::ops::matmul::test_matmul_2d_aligned
      tests::fusion::quantization::ops::matmul::test_matmul_2d_aligned_fused
      tests::fusion::quantization::ops::matmul::test_matmul_lhs_float_rhs_quantized
      tests::fusion::quantization::ops::matmul::test_matmul_mixed_block_scale
      tests::tensor::module::attention::test_attention_no_mask
      tests::tensor::quantization::ops::matmul::test_matmul_2d_aligned
      tests::tensor::quantization::ops::matmul::test_matmul_2d_aligned_fused
      tests::tensor::quantization::ops::matmul::test_matmul_lhs_float_rhs_quantized
      tests::tensor::quantization::ops::matmul::test_matmul_mixed_block_scale
  
  test result: FAILED. 5924 passed; 20 failed; 72 ignored; 0 measured; 0 filtered out; finished in 906.62s

16/20 are also existing failures in previous versions (see this GPU run, while the 4/20 additional failures are related to matmul

  thread 'f16_i32::tests::fusion::quantization::ops::matmul::test_matmul_lhs_float_rhs_quantized' (9881) panicked at /home/agent/.cargo/git/checkouts/cubek-21eb4731b65c1fbd/1b26e14/crates/cubek-matmul/src/definition/base.rs:250:9:
 Invalid or non-contiguous matrix layout: shape=[4, 4], strides=[1, 1]

also fail locally on main (but no recorded failure on CI, so it must be related to changes since the last GPU run).

cuda failures

Reporting, but obviously these are totally unrelated to the wgpu upgrade

  failures:
      f16_i32::tests::fusion::ops::reshape::should_support_multiple_reshapes_cloned_tensor
      f16_i32::tests::tensor::module::attention::test_attention_no_mask
      tests::fusion::ops::reshape::should_support_multiple_reshapes_cloned_tensor
      tests::tensor::module::attention::test_attention_no_mask