fix: Support building multiple models of GPU and DSP together

See merge request applied-machine-learning/sysml/mace!1314

feat: Use a standalone gitlab ci runner for Micro

See merge request applied-machine-learning/sysml/mace!1313

opt: optimize the performance of `Reduce` OP

See merge request applied-machine-learning/sysml/mace!1311

N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>

fix: Add framework_type arg in Net so that we can get it from Net besides Op

See merge request applied-machine-learning/sysml/mace!1312

fix: Extract Tensors from `params` instead of `state_dict` to be compatible with PyTorch 1.4

See merge request applied-machine-learning/sysml/mace!1310

feat: add Elu for Tensorflow and Caffe and fix bug in reverse op and net adapter

See merge request applied-machine-learning/sysml/mace!1309

feat: Integrate CMSIS5 to MACE Micro

See merge request applied-machine-learning/sysml/mace!1305

# Conflicts:
#   tools/converter.py
#   tools/python/convert.py
#   tools/python/transform/base_converter.py
#   tools/python/utils/config_parser.py

N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>

fix: Use 1 element tensor as scalar for Eltwise

See merge request applied-machine-learning/sysml/mace!1308

pytorch converter

See merge request applied-machine-learning/sysml/mace!1300

opt: reduce the so size for bazel and fix error in cmake compile

See merge request applied-machine-learning/sysml/mace!1307

N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>

fix: Fix folding MatMul BiasAdd

See merge request applied-machine-learning/sysml/mace!1306

Fix micro matmul bias access bug

See merge request applied-machine-learning/sysml/mace!1304

fix: fix `Transpose` op's computing error on batch tensor.

See merge request applied-machine-learning/sysml/mace!1302

perf: opt `reduce` op's performance on GPU

See merge request applied-machine-learning/sysml/mace!1301

N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>

Co-authored-by: Ngasgallo <rgall@sertiscorp.com>

feat: Add bf16 kernels for MobileNet

See merge request applied-machine-learning/sysml/mace!1298

N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>

perf: Prefer GPU if input ops on different devices

See merge request applied-machine-learning/sysml/mace!1297

feat: support FusedBatchNormV2/FusedBatchNormV3 for tensorflow

See merge request applied-machine-learning/sysml/mace!1296

N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>

add micro concat op

See merge request applied-machine-learning/sysml/mace!1295

Add onnx's Elu operator

See merge request applied-machine-learning/sysml/mace!1294