updating flatten/unflatten functions #3282
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/3282
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ✅ No FailuresAs of commit fd731af with merge base 9266734 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
liangel-02
added
the
topic: improvement
liangel-02
force-pushed
the
update_metadata_structure
branch
3 times, most recently
from
83fdbdc to
bd6bbac
Compare
jerryzh168
left a comment
jerryzh168
left a comment
There was a problem hiding this comment.
maybe add a test for the format as well? it's a bit surprising to see that format changes like this doesn't require any changes to the test
liangel-02
force-pushed
the
update_metadata_structure
branch
3 times, most recently
from
1539983 to
22d8c3a
Compare
liangel-02
force-pushed
the
update_metadata_structure
branch
2 times, most recently
from
1c695a1 to
cb2b13e
Compare
| result = {} | ||
|
|
||
| for tensor_name in tensor_names: | ||
| prefix = f"{tensor_name.rsplit('.', 1)[0]}._{tensor_name.rsplit('.', 1)[1]}_" |
There was a problem hiding this comment.
nit: split the tensor_name separate beforehand?
module_fqn, weight_name = tensor_name.rsplit(".", 1)
...
jerryzh168
left a comment
jerryzh168
left a comment
There was a problem hiding this comment.
looks good, although I think ideally this should be 2 PRs, future PRs should only contain a single change for easy review and clear purpose
liangel-02
force-pushed
the
update_metadata_structure
branch
from
cb2b13e to
fd731af
Compare
Summary
Updating
unflatten_tensor_state_dictandflatten_tensor_state_dictto only modify tensor subclasses. Also changing naming format so that we can directly assign module parameters.Testing
python test/prototype/safetensors/test_safetensors_support.py