- 13fe02a Create RUN_KERNEL macro for DepthwiseConvS8 by Alex Van Damme · 8 months ago
- 088841b Specialize 3x3, stride 1 DepthwiseConv by Alex Van Damme · 8 months ago
- 5106685 Vectorized Logistic for int8 by Alex Van Damme · 8 months ago
- 8ef6f11 Use 1x1 d=32n kernel more aggressively. by Derek Chow · 8 months ago
- 1ad0284 Point wise convolution input_depth = 16 by Derek Chow · 8 months ago
- 5160bc6 Add specialized pointwise convolution. by Derek Chow · 8 months ago
- 78538b3 Use dwconv on input_depth == 1, 5x5 by Alex Van Damme · 8 months ago
- 0d75a42 Refactor kernel selection. by Derek Chow · 8 months ago
- 2aeaaa8 Enable O2 optimizations. by Derek Chow · 8 months ago
- 5971487 Improvements to elementwise_add_{s8,s16} by Alex Van Damme · 8 months ago
- ca72b40 Improve ConvS8D4 by Alex Van Damme · 8 months ago
- cd3d0e3 3x3 DepthwiseConv w/ adwconv by Alex Van Damme · 9 months ago
- b1afda6 Specialize 5x5 DepthwiseConv by Alex Van Damme · 9 months ago
- fd54c9e Relax output_depth for ConvS8D4 by Alex Van Damme · 9 months ago
- 86806b3 Add input_depth == 1 specialized convolution. by Derek Chow · 9 months ago
- 40a8300 Fix swizzling and rounding for depthwise conv by Alex Van Damme · 9 months ago
- e617f61 Relax restrictions on aconv-based Conv2D by Alex Van Damme · 9 months ago
- 3b8d3cb sw/kelvin: Refactoring TFLM Kelvin kernel optimization codes by Lun Dong · 9 months ago
- b11ad17 Specialized int8 conv kernels by Alex Van Damme · 9 months ago
- be4ab97 WIP Depthwise conv for 2d kernels by Naveen Dodda · 9 months ago
- 3285c6a Changes to optimize max_pool for Kelvin int8 dtype by Naveen-Dodda · 1 year, 2 months ago
- 903c311 Consolidate kernel selection behind generic conv_per_channel_b64. by Derek Chow · 1 year, 2 months ago
- a5f129c Optimize 1x1 group conv. by Derek Chow · 1 year, 3 months ago
- 43879e4 OSS cleanup by Cindy Liu · 1 year, 3 months ago
- 766e5af Add dilation support for depthwise convolution. by Derek Chow · 1 year, 3 months ago
- 16ffb60 8x16 Depthwise convolution for Kelvin. by Derek Chow · 1 year, 4 months ago
- f39cadd Helper methods for Conv2D kernel by Alex Van Damme · 1 year, 5 months ago
- 5706c8a Refactor crt targets. by Cindy Liu · 1 year, 4 months ago
- 2c72b28 Fix another rounding issue in elementwise_add by Alex Van Damme · 1 year, 5 months ago
- 7a2eaa4 Support functions for LeakyRelu by Alex Van Damme · 1 year, 5 months ago
- 288b9e9 Switch addition to vdmulh, fix rounding by Alex Van Damme · 1 year, 5 months ago
- 59112ad Add support functions for an optimized Add kernel by Alex Van Damme · 1 year, 5 months ago
- d555403 Optimized TFLM Reshape for kelvin by Alex Van Damme · 1 year, 6 months ago