Conversation
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
|
run-slow: compressed_tensors_integration, fouroversix_integration |
|
This comment contains models: [] |
| self.assertTrue( | ||
| torch.allclose( |
There was a problem hiding this comment.
let's use torch.testing.assert_close
| f"Weight mismatch for module '{name}' in quantized-only or stacked model.", | ||
| ) | ||
|
|
||
| def test_outputs_match(self): |
There was a problem hiding this comment.
just a short comment (on PR page like here) to say why this is removed would be nice. So we keep a "for the record")
| def test_default_run_compressed__True(self): | ||
| from compressed_tensors.linear.compressed_linear import CompressedLinear | ||
| from compressed_tensors.quantization.utils import iter_named_leaf_modules | ||
| from compressed_tensors import QuantizationStatus |
There was a problem hiding this comment.
I guess the changes here is also because sth is deprecated (?)
|
[For maintainers] Suggested jobs to run (before merge) run-slow: compressed_tensors_integration, fouroversix_integration |
* update * Use torch.testing.assert_close in compressed_tensors weight check
3 participants
What does this PR do?
This PR fixes quantization tests. A few things were deprecated when compressed-tensors had their latest release, so i'm updating the tests. For fouroversix, it's just that the model was a bit too big for the CI.
We removed some tests in compressed tensors as we don't test sparse model anymore.