Add sparse compressed fake tensor support #120920
Conversation
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/120920
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (4 Unrelated Failures)As of commit 968a6cf with merge base 2e50566 ( BROKEN TRUNK - The following jobs failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
pearu
added
module: sparse
As in the title. cc alexsamardzic nikitaved cpuhrsch amjames bhosmer jcaip [ghstack-poisoned]
As in the title. cc alexsamardzic nikitaved cpuhrsch amjames bhosmer jcaip [ghstack-poisoned]
As in the title. cc alexsamardzic nikitaved cpuhrsch amjames bhosmer jcaip voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 aakhundov kadeng [ghstack-poisoned]
As in the title. cc alexsamardzic nikitaved cpuhrsch amjames bhosmer jcaip voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 aakhundov kadeng [ghstack-poisoned]
As in the title. cc alexsamardzic nikitaved cpuhrsch amjames bhosmer jcaip voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 aakhundov kadeng [ghstack-poisoned]
As in the title. cc alexsamardzic nikitaved cpuhrsch amjames bhosmer jcaip voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 aakhundov kadeng [ghstack-poisoned]
| elif is_sparse_compressed(t): | ||
| is_leaf = safe_is_leaf(t) | ||
|
|
||
| def mk_meta(): |
There was a problem hiding this comment.
This function is weirdly a lot more complicated than the one in sparse_coo above, not sure if it's needed or not yet
There was a problem hiding this comment.
I guess the problem is you don't have a constructor that just directly takes sparse/dense dim and allocates those tensor under the hood, do you...
There was a problem hiding this comment.
Exactly, we don't have _sparse_coo_tensor_with_dims-alike for sparse compressed tensors. I could expose one as an aten operator but does it matter if it is either implemented in Python (as above) or in C++? IIRC, their specification will be the same: construct as sparse compressed tensor with indices and values as empty tensors with device='meta' and nnz==0.
What do you think?
... not sure if it's needed or not yet
It surely appears to be needed as without it a bunch of dynamo tests is failing.
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Merge failedReason: Command Details for Dev Infra teamRaised by workflow job |
As in the title. cc alexsamardzic nikitaved cpuhrsch amjames bhosmer jcaip voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang aakhundov [ghstack-poisoned]
|
@pytorchbot merge -f "The 4 CI failures are not related to the changes in the PR" |
Merge startedYour change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Please use Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
As in the title.
Stack from ghstack (oldest at bottom):
cc @alexsamardzic @nikitaved @cpuhrsch @amjames @bhosmer @jcaip @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @chenyang78 @kadeng @chauhang @aakhundov