Commit Graph

10 Commits

Author SHA1 Message Date
Mohsen Saffari
04d3a0ced0 Some refactoring for this 1 channel per block kernel 2025-12-09 09:13:31 +00:00
Mohsen Saffari
8d4f766f9c Add scale and bias to batch normalization 2025-12-04 16:22:32 +00:00
Mohsen Saffari
60145b3cce Add vector size management 2025-12-03 16:33:34 +00:00
Mohsen Saffari
065535cf81 some refactoring to align kernel with ck-tile kernels 2025-12-03 12:03:31 +00:00
Mohsen Saffari
c4199307ec Change from NCHW to MHWC based on old-ck and manage verifying for c > 1 2025-12-02 15:47:46 +00:00
Mohsen Saffari
3194b653f7 add tiling, pipeline, follow layernorm2d 2025-11-28 18:29:10 +00:00
Mohsen Saffari
b732595d9f add kSaveMeanInvStd, kUpdateMovingAverage in Traits 2025-11-27 15:06:13 +00:00
Mohsen Saffari
5c8e8684ec add gamma, bias to the simple kernel 2025-11-27 13:32:18 +00:00
Mohsen Saffari
c7afd540d9 correct the batch normalization algorithm to calc reduce over NxHxW 2025-11-25 16:46:00 +00:00
Mohsen Saffari
5fa3ec2b37 Some refactoring for simple batchnorm kernel and example 2025-11-25 13:14:58 +00:00