提交 · 7f2c41fe5087b71c703af9d60608f5b6a9e4cb09 · Xiaomi / Mace

27 10月, 2020 1 次提交
- L
  
  fix: Extract Tensors from `params` instead of `state_dict` to be compatible with PyTorch 1.4 · 97624cf2
  由 like15 提交于 10月 27, 2020
  
  97624cf2
23 10月, 2020 3 次提交
- L
  feat: add Elu for Tensorflow and Caffe and fix bug in reverse op and net adapter · dd2cdb01
  由 luxuhui 提交于 10月 21, 2020
```
N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>
```
  dd2cdb01
- L
  
  feat: Support Densenet and MobilenetV2 conversion from PyTorch · 9efe5dc5
  由 like15 提交于 9月 30, 2020
  
  9efe5dc5
- B
  
  fix: Use 1 element tensor as scalar for Eltwise · b39b5399
  由 Bin Li 提交于 10月 23, 2020
  
  b39b5399
22 10月, 2020 1 次提交
- Z
  
  feat: Integrate CMSIS5 to MACE Micro · 6352c291
  由 Zhang Zhimin 提交于 10月 10, 2020
  
  6352c291
20 10月, 2020 1 次提交
- B
  
  fix: Fix folding MatMul BiasAdd, Reshape · 49659ddb
  由 Bin Li 提交于 10月 20, 2020
  
  49659ddb
21 9月, 2020 1 次提交
- L
  feat: support FusedBatchNormV2/FusedBatchNormV3 for tensorflow · 44f7d259
  由 luxuhui 提交于 9月 21, 2020
```
N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>
```
  44f7d259
27 8月, 2020 1 次提交
- L
  
  Add onnx's Elu operator · c6118a22
  由 lichao18 提交于 8月 27, 2020
  
  c6118a22
19 8月, 2020 1 次提交
- B
  
  Add quantized Relu/BiasAdd/Pad/Transpose for CPU · 30cff381
  由 Bin Li 提交于 8月 18, 2020
  
  30cff381
22 7月, 2020 1 次提交

Integrate MediaTek APU Support on mt6873, mt6885 and mt6853 (#662) · 63feaf50

由 Yi-Kai-Chen 提交于 7月 22, 2020

* Margaux dev.

* Support Pad, Activation, Mul operator for mnasnet

* Support document enhance model

* Apply add-init-cache-and-preemption patch

* Refine code

* Add APU int16 quantization and fix macro bugs

* Refine code
Co-authored-by: NYungChien Hsu <yungchien.hsu@mediatek.com>
Co-authored-by: NEric Chen <eric-yk.chen@mediatek.com>

63feaf50

17 7月, 2020 2 次提交
- L
  feat: support extract_image_patches op and support dynamic filter of conv2d/deconv2d · 9472dcee
  由 luxuhui 提交于 6月 24, 2020
```
N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>
```
  9472dcee
- Y
  Integrate MediaTek APU Support on mt6873, mt6885 and mt6853 · 01dbeae3
  由 YungChien Hsu 提交于 3月 11, 2020
```
Signed-off-by: Nlichao18 <lichao18@xiaomi.com>

Add init cache and preemption

Support mnasnet and new 5g library

Add GetModelSize API for model header
```
  01dbeae3
15 7月, 2020 1 次提交
- B
  
  Fix ONNX output shape bug · 19089a52
  由 Bin Li 提交于 7月 15, 2020
  
  19089a52
03 7月, 2020 1 次提交
- Y
  
  add Linear and NoOp ops to support more Kaldi Components(layers) (#660) · aec84219
  由 Yiming Wang 提交于 7月 03, 2020
  
  aec84219
30 6月, 2020 1 次提交
- L
  feat: support splitV op for tensorflow · 041a5d2e
  由 luxuhui 提交于 6月 29, 2020
```
N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>
```
  041a5d2e
29 6月, 2020 1 次提交
- Y
  
  Add MegEngine converter for MACE (#658) · a28e8128
  由 yzchenmonkey 提交于 6月 29, 2020
  
  a28e8128
17 6月, 2020 1 次提交
- L
  feat: support `half_pixel_centers` arg for `resize` type ops · d24a3ddd
  由 luxuhui 提交于 6月 16, 2020
```
N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>
```
  d24a3ddd
01 6月, 2020 1 次提交
- B
  
  Fix ONNX Upsample · 5660644a
  由 Bin Li 提交于 6月 01, 2020
  
  5660644a
29 5月, 2020 1 次提交
- L
  feature: support GroupNorm op · 6b65f42b
  由 luxuhui 提交于 5月 12, 2020
```
N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>
```
  6b65f42b
28 5月, 2020 1 次提交
- L
  support RELU6/ArgMax/ResizeNearestNeighbor op for Caffe, fix bug. · b5dfcf4d
  由 luxuhui 提交于 5月 11, 2020
```
N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>
```
  b5dfcf4d
18 5月, 2020 1 次提交
- B
  
  Fix onnx depthspace and cmake host · b53a461c
  由 Bin Li 提交于 5月 12, 2020
  
  b53a461c
09 5月, 2020 2 次提交
- B
  
  Fix ONNX padding and add transpose half · 2530c7ee
  由 Bin Li 提交于 5月 09, 2020
  
  2530c7ee
- L
  remove openmp, fix scratch bug in micro and opt some codes · 2ea7b6b1
  由 luxuhui 提交于 5月 08, 2020
```
N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>
```
  2ea7b6b1
08 5月, 2020 1 次提交
- R
  Add support for AddV2 op in TF (#637) · ab7499a5
  由 Riccardo 提交于 5月 08, 2020
```
Co-authored-by: Nriccardo <rgall@sertiscorp.com>
```
  ab7499a5
07 5月, 2020 1 次提交
- B
  
  Fix squeeze if not transposable · e21e8d19
  由 Bin Li 提交于 5月 07, 2020
  
  e21e8d19
29 4月, 2020 1 次提交
- B
  
  Fix caffe quantized fc · 123e17d7
  由 Bin Li 提交于 4月 29, 2020
  
  123e17d7
21 4月, 2020 3 次提交
- B
  
  Fix hexagon deconv and stridedslice · 27757d88
  由 Bin Li 提交于 4月 14, 2020
  
  27757d88
- B
  
  Support HTA custom API · 28954099
  由 Bin Li 提交于 3月 26, 2020
  
  28954099
- R
  Add support for ReLU6 op in CAFFE (#621) · f4061000
  由 Riccardo 提交于 4月 21, 2020
```
Add support for Clip op in CAFFE
Co-authored-by: Nriccardo <rgall@sertiscorp.com>
```
  f4061000
30 3月, 2020 1 次提交
- Y
  
  fix: Caffe filter format for quantization · e66d5ab6
  由 yejianwu 提交于 3月 30, 2020
  
  e66d5ab6
27 3月, 2020 1 次提交
- L
  fix validate bug on reshape layer and duplicate name bug in encrypt.py · 87041d07
  由 luxuhui 提交于 3月 24, 2020
```
issue595
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>
```
  87041d07
10 3月, 2020 1 次提交
- B
  
  Support mixing quantization-aware training and post-quantization ranges · b79f9f5a
  由 Bin Li 提交于 3月 09, 2020
  
  b79f9f5a
09 3月, 2020 1 次提交
- B
  
  Support unet for hexagon · 7c10b286
  由 Bin Li 提交于 3月 04, 2020
  
  7c10b286
02 3月, 2020 1 次提交
- L
  fix prelu&validate bug · 36009059
  由 luxuhui 提交于 3月 02, 2020
```
N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>
```
  36009059
19 2月, 2020 1 次提交
- L
  fix bug on reshape op & tar command & merge_duplicate_nodes · 020455ce
  由 luxuhui 提交于 2月 18, 2020
```
issue:593
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>
```
  020455ce
14 1月, 2020 1 次提交
- L
  fix bug in converter and dana · c37d017a
  由 luxuhui 提交于 1月 14, 2020
```
N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>
```
  c37d017a
11 1月, 2020 1 次提交
- L
  fix 'op not unique' error in converting model from tf · bc1f0a56
  由 luxuhui 提交于 1月 11, 2020
```
N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>
```
  bc1f0a56
07 1月, 2020 2 次提交
- L
  opt the performance of ResizeNearestNeighbor&Deconv OP · c8d5c88e
  由 luxuhui 提交于 1月 06, 2020
```
N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>
```
  c8d5c88e
- L
  fix bug on converting LeakyRelu from ONNX · 716963c9
  由 luxuhui 提交于 1月 07, 2020
```
N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>
```
  716963c9
02 1月, 2020 1 次提交
- B
  
  Support Hexagon TransposeConv2d and multiple float inputs/outputs for HTA · 68cfff5f
  由 Bin Li 提交于 11月 27, 2019
  
  68cfff5f