| Age | Commit message (Collapse) | Author | |
|---|---|---|---|
| 2020-08-02 | Bind RNG by reference in microbenchmarks | Marat Dukhan | |
| PiperOrigin-RevId: 324540550 | |||
| 2020-03-08 | Cleanup unused SoftMax benchmark presets | Marat Dukhan | |
| Leave only the "blend" presets reported in the paper PiperOrigin-RevId: 299722958 | |||
| 2020-03-06 | Optional SoftMax benchmarks vs DNNL | Marat Dukhan | |
| PiperOrigin-RevId: 299494008 | |||
| 2020-01-27 | Rename SoftArgMax operator to SoftMax | Marat Dukhan | |
| - Match naming in TensorFlow, PyTorch, and ONNX - Avoid confusion with occasionally proposed SoftArgMax operator with different semantics y[i] := exp(x[i]) * i / sum(exp(x[j])) PiperOrigin-RevId: 291812880 | |||
 |
index : external/XNNPACK.git | |
| [no description] |
| aboutsummaryrefslogtreecommitdiff |