[quant] Adding zero point type check for per channel quantization #40811
Conversation
[ghstack-poisoned]
…ization" Differential Revision: [D22319417](https://our.internmc.facebook.com/intern/diff/D22319417) [ghstack-poisoned]
💊 CI failures summary and remediationsAs of commit 5ff4884 (more details on the Dr. CI page): 💚 💚 Looks good so far! There are no failures yet. 💚 💚 This comment was automatically generated by Dr. CI (expand for details).Follow this link to opt-out of these comments for your Pull Requests.Please report bugs/suggestions on the GitHub issue tracker or post in the (internal) Dr. CI Users group. This comment has been revised 11 times. |
…ization" Differential Revision: [D22319417](https://our.internmc.facebook.com/intern/diff/D22319417) [ghstack-poisoned]
| int64_t quant_min, | ||
| int64_t quant_max) { | ||
| TORCH_CHECK(self.scalar_type() == ScalarType::Float); | ||
| TORCH_CHECK(scale.scalar_type() == ScalarType::Float, |
There was a problem hiding this comment.
I think it's float, as per the iterator expectation here. Should I change them to double?
There was a problem hiding this comment.
Oh I thought this is quantize_per_tensor.
If this is float then probably zero_point should be int32_t, just to be consistent.
There was a problem hiding this comment.
The problem is that in the kernel implementation it is float for scale and int64_t for the zero_point. I will have to change all the kernels
There was a problem hiding this comment.
do you know why is that? it would be better to be consistent
There was a problem hiding this comment.
I am not sure why the inconsistency, but I see some kernels use 64 bits and some 32.
…ization" Differential Revision: [D22319417](https://our.internmc.facebook.com/intern/diff/D22319417) [ghstack-poisoned]
…ization" Differential Revision: [D22319417](https://our.internmc.facebook.com/intern/diff/D22319417) [ghstack-poisoned]
…ization" Differential Revision: [D22319417](https://our.internmc.facebook.com/intern/diff/D22319417) [ghstack-poisoned]
…ization" Differential Revision: [D22319417](https://our.internmc.facebook.com/intern/diff/D22319417) [ghstack-poisoned]
Stack from ghstack:
Differential Revision: D22319417