These 62 commits are when the Protocol Buffers files have changed:
| Commit: | e3d95d1 | |
|---|---|---|
| Author: | yuanlonglong | |
| Committer: | yuanlonglong | |
feat: Support new op where and more mode for op cast and slice 1. Support new op where 2. Support cast op "to bool" mode 3. Support slice op "multi step" mode Signed-off-by: yuanlonglong <yuanlonglong@xiaomi.com>
The documentation is generated from this commit.
| Commit: | 5500307 | |
|---|---|---|
| Author: | Wu Zhuo | |
| Committer: | yuanlonglong | |
Fix: fix rearrange_batch_to_space cause act_op output_shape incorrect
| Commit: | 0ec5e13 | |
|---|---|---|
| Author: | Wu Zhuo | |
| Committer: | Wu Zhuo | |
fix: fix dsp crash problem when dsp op tanh and sigmoid require more sapce that exceed output_shape
| Commit: | 8358b9c | |
|---|---|---|
| Author: | 吴卓 | |
| Committer: | 袁龙龙 | |
feat: Add Qnn Backend
| Commit: | dc6b644 | |
|---|---|---|
| Author: | like15 | |
| Committer: | like15 | |
refactor: Reuse OpenCL cache file for GPUs of the same type and add CRC checksum
| Commit: | d31c879 | |
|---|---|---|
| Author: | luxuhui | |
| Committer: | luxuhui | |
refactor: refactoring memory mechanism and support multi subgraph N/A Signed-off-by: Luxuhui <luxuhui@xiaomi.com>
| Commit: | 36d673e | |
|---|---|---|
| Author: | lichao18 | |
| Committer: | lichao18 | |
Add the nonlocal reshape op and optimize cpu fp16 gemm
| Commit: | 6352c29 | |
|---|---|---|
| Author: | Zhang Zhimin | |
| Committer: | Zhang Zhimin | |
feat: Integrate CMSIS5 to MACE Micro
| Commit: | 01dbeae | |
|---|---|---|
| Author: | YungChien Hsu | |
| Committer: | lichao18 | |
Integrate MediaTek APU Support on mt6873, mt6885 and mt6853 Signed-off-by: lichao18 <lichao18@xiaomi.com> Add init cache and preemption Support mnasnet and new 5g library Add GetModelSize API for model header
| Commit: | 6b65f42 | |
|---|---|---|
| Author: | luxuhui | |
| Committer: | luxuhui | |
feature: support GroupNorm op N/A Signed-off-by: Luxuhui <luxuhui@xiaomi.com>
| Commit: | b5dfcf4 | |
|---|---|---|
| Author: | luxuhui | |
| Committer: | luxuhui | |
support RELU6/ArgMax/ResizeNearestNeighbor op for Caffe, fix bug. N/A Signed-off-by: Luxuhui <luxuhui@xiaomi.com>
| Commit: | f406100 | |
|---|---|---|
| Author: | Riccardo | |
| Committer: | GitHub | |
Add support for ReLU6 op in CAFFE (#621) Add support for Clip op in CAFFE Co-authored-by: riccardo <rgall@sertiscorp.com>
| Commit: | 373f1ef | |
|---|---|---|
| Author: | luxuhui | |
| Committer: | luxuhui | |
add mace micro N/A Signed-off-by: Luxuhui <luxuhui@xiaomi.com>
| Commit: | 72a3751 | |
|---|---|---|
| Author: | luxuhui | |
| Committer: | luxuhui | |
add lpnorm、mvnorm op for caffe, enhance biasadd、reshape op N/A Signed-off-by: Luxuhui <luxuhui@xiaomi.com>
| Commit: | 23bd28c | |
|---|---|---|
| Author: | Bin Li | |
| Committer: | Bin Li | |
Dequantize weights to half
| Commit: | 9d25993 | |
|---|---|---|
| Author: | Liangliang He | |
| Committer: | Liangliang He | |
Fix gradle NDK build
| Commit: | 0c5aab6 | |
|---|---|---|
| Author: | yulianfei | |
| Committer: | liukai6 | |
Reformat fp16 gemv, add FP16_MATMUL_WEIGHT Summary: Reformat fp16 gemv, add FP16_MATMUL_WEIGHT Differential Revision: https://phabricator.d.xiaomi.net/D144134
| Commit: | 91bd5c1 | |
|---|---|---|
| Author: | 刘托 | |
Merge branch 'support-caffe-interp' into 'master' Feature: Support caffe Interp layer. See merge request !1029
| Commit: | bd7e156 | |
|---|---|---|
| Author: | Bin Li | |
| Committer: | Bin Li | |
Integrate HTA
| Commit: | 8673a20 | |
|---|---|---|
| Author: | liuqi | |
Feature: Support caffe Interp layer.
| Commit: | c180e0b | |
|---|---|---|
| Author: | lichao18 | |
| Committer: | lichao18 | |
Add SSD box predictor module
| Commit: | 78a2b5e | |
|---|---|---|
| Author: | lichao18 | |
| Committer: | lichao18 | |
Add channel shuffel caffe convert
| Commit: | e446bd6 | |
|---|---|---|
| Author: | liuqi | |
Feature: Support mixed usage of CPU and GPU 1. Support memory optimization at runtime 2. Support memory type transformation automatically at runtime. 3. Move gpu winograd convolution to Conv2D. 4. Support Data Type transformation automatically. 5. MaceTensor API support optional data type. 6. Add input_data_formats and output_data_formats at yaml 7. Related issue(#363)
| Commit: | 607a1f4 | |
|---|---|---|
| Author: | liuqi | |
Refactor: move all files in kernels directory to ops and remove kernels. 1. Move all files in kernels to ops 2. unify the op_def_registry and op_registry. 3. Support op clip: do not link the ops' code when the op is not registered.
| Commit: | e4dc6f9 | |
|---|---|---|
| Author: | 李滨 | |
Merge branch 'refactor-mace' into 'master' Ajusted the structure of MACE to support the mixing of CPU and GPU. See merge request !846
| Commit: | c0164bc | |
|---|---|---|
| Author: | liuqi | |
Refactor: Ajusted the structure of MACE to support the mixing of CPU and GPU. 1. Merge the Operation and OpKernel into Operation. 2. Add OpDefRegistry for registering the definitions of Operations. 3. Add information for MaceStatus. 4. Unify 'BatchNorm' and 'FoldedBatchNorm' into 'BatchNorm' 5. Remove unused op 'Proposal'
| Commit: | 76458b0 | |
|---|---|---|
| Author: | Bin Li | |
| Committer: | Bin Li | |
Support Mace to Hexagon
| Commit: | 8f851c1 | |
|---|---|---|
| Author: | yejianwu | |
remove name and version in NetDef in mace.proto
| Commit: | 6c65796 | |
|---|---|---|
| Author: | 李寅 | |
| Committer: | 李寅 | |
Support quantize-weights only
| Commit: | 0c6ec59 | |
|---|---|---|
| Author: | 李寅 | |
| Committer: | 李寅 | |
Move quantize info into op's output
| Commit: | 3d19517 | |
|---|---|---|
| Author: | 李寅 | |
| Committer: | 李寅 | |
Init quantize project: 1. Quantize weights 2. Add quantize-dequantze nodes
| Commit: | 6ac7c5f | |
|---|---|---|
| Author: | liuqi | |
Add device_type to MemoryBlock and fix workspace preallocation bug.
| Commit: | 8812070 | |
|---|---|---|
| Author: | liuqi | |
Add gpu avalibility check and return status to user if gpu call failed.
| Commit: | 5b35740 | |
|---|---|---|
| Author: | yejianwu | |
| Committer: | yejianwu | |
refactor arg_helper
| Commit: | 7819beb | |
|---|---|---|
| Author: | Liangliang He | |
refactor third_party
| Commit: | bd1fa3f | |
|---|---|---|
| Author: | yejianwu | |
refactor mace.proto and relevant macros
| Commit: | c761a7b | |
|---|---|---|
| Author: | yejianwu | |
remove mace.pb.h deps in public api header, Reserve() for pb resource
| Commit: | 42160b3 | |
|---|---|---|
| Author: | yejianwu | |
load model in pb
| Commit: | 3e82ad6 | |
|---|---|---|
| Author: | 李寅 | |
| Committer: | 李寅 | |
Refactor model converter and transformer
| Commit: | bcc066e | |
|---|---|---|
| Author: | Liangliang He | |
| Committer: | Liangliang He | |
Add third party license
| Commit: | 3ba07e0 | |
|---|---|---|
| Author: | liuqi | |
| Committer: | liuqi | |
Change DeviceType::OPENCL to DeviceType::GPU and fix some bugs.
| Commit: | 9985cbf | |
|---|---|---|
| Author: | liuqi | |
Support user specified input files.
| Commit: | 94a520f | |
|---|---|---|
| Author: | yejianwu | |
add lib
| Commit: | bfbe2d6 | |
|---|---|---|
| Author: | liuqi | |
| Committer: | liuqi | |
Support multiple outputs' operation.
| Commit: | e6f32c5 | |
|---|---|---|
| Author: | liuqi | |
| Committer: | liuqi | |
Support caffe model.
| Commit: | 0e5ebc1 | |
|---|---|---|
| Author: | yejianwu | |
move distribute tools from mace
| Commit: | a022085 | |
|---|---|---|
| Author: | yejianwu | |
move tools to libmace
| Commit: | e5d3bb5 | |
|---|---|---|
| Author: | liuqi | |
| Committer: | liuqi | |
Add caffe ops stat script and fix addn bugs.
| Commit: | e4860f6 | |
|---|---|---|
| Author: | Liangliang He | |
| Committer: | Liangliang He | |
Refactor OpenCL static dependencies
| Commit: | 489b091 | |
|---|---|---|
| Author: | wuchenghui | |
revert dsp convert tool to support both nn_graph & nnlib
| Commit: | 78d58e7 | |
|---|---|---|
| Author: | wuchenghui | |
fix dsp convert tool to support nn_graph
| Commit: | fb22aa7 | |
|---|---|---|
| Author: | Yin Li | |
| Committer: | Yin Li | |
Mace GPU memory sharing optimization
| Commit: | 062fa1e | |
|---|---|---|
| Author: | liuqi | |
Rename opmode to netmode.
| Commit: | 336bdbe | |
|---|---|---|
| Author: | liuqi | |
Add OpMode to support some op to run once at the beginning.
| Commit: | 9fd509d | |
|---|---|---|
| Author: | wuchenghui | |
| Committer: | wuchenghui | |
fix graph convert
| Commit: | 60a0ed6 | |
|---|---|---|
| Author: | Yin Li | |
Add output & memory related info
| Commit: | f07dd51 | |
|---|---|---|
| Author: | liuqi | |
Refactor the opencl kernel build logic.
| Commit: | 77ea99f | |
|---|---|---|
| Author: | liuqi | |
| Committer: | liuqi | |
Add dynamic build opencl kernel logic.
| Commit: | 629cb8a | |
|---|---|---|
| Author: | 李寅 | |
| Committer: | 李寅 | |
Add dsp wrapper and model convert tool
| Commit: | 9d40ffd | |
|---|---|---|
| Author: | liuqi | |
Add op run metadata statistics and benchmark framework.
| Commit: | cb871b0 | |
|---|---|---|
| Author: | Liangliang He | |
Add Neon device
| Commit: | 37e5bd4 | |
|---|---|---|
| Author: | 李寅 | |
Add Ops, Net, and Dummy Relu