IntenseWebs/openvino
3
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

remove quantization config for vpu (#17507)

Browse Source 
* remove quantization config for vpu

* remove VPU config testcases from test unify scales

---------

Co-authored-by: shokinal <alexander.shokin@intel.com>
Co-authored-by: Alexander Kozlov <alexander.kozlov@intel.com>
This commit is contained in:
Roman Zubarev 2023-06-05 11:14:48 +01:00 committed by GitHub
parent 0d9109acf3
commit 8e97010595
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 0 additions and 310 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

307
tools/pot/openvino/tools/pot/configs/hardware/vpu.json
View File

@ -1,307 +0,0 @@
{
"target_device": "VPU",
"config": {
"quantization": {
"q8_tn": {
"bits": 8,
"mode": [
"asymmetric"
],
"granularity": "pertensor"
},
"q8_ch": {
"bits": 8,
"mode": [
"symmetric"
],
"granularity": "perchannel"
},
"q8_a_ch": {
"bits": 8,
"mode": [
"asymmetric"
],
"unified_zeropoint": true,
"granularity": "perchannel"
},
"q4_tn": {
"bits": 4,
"mode": "symmetric",
"granularity": "pertensor"
},
"q4_ch": {
"bits": 4,
"mode": "symmetric",
"granularity": "perchannel"
},
"q2_ch": {
"bits": 2,
"mode": "symmetric",
"granularity": "perchannel"
}
}
},
"operations": [
{
"type": "Convolution",
"quantization": {
"activations": ["q8_tn", "q4_tn", "q8_tn"],
"weights": ["q8_ch", "q4_ch", "q2_ch", "q8_a_ch"]
}
},
{
"type": "GroupConvolutionBackpropData",
"quantization": {
"activations": "q8_tn",
"weights": ["q8_ch", "q4_ch", "q2_ch", "q8_a_ch"]
}
},
{
"type": "DepthWiseConvolution",
"quantization": {
"activations": "q8_ch",
"weights": ["q8_ch", "q4_ch", "q2_ch", "q8_a_ch"]
}
},
{
"type": "MatMul",
"quantization": {
"activations": ["q8_tn", "q4_tn", "q8_tn"],
"weights": ["q8_ch", "q4_ch", "q2_ch", "q8_a_ch"]
}
},
{
"type": "Add",
"quantization": {
"activations": ["q8_tn", "q4_tn", "q8_tn"],
"weights": ["q8_ch", "q4_tn", "q8_ch"]
},
"attributes": {
"unified_scales": "all"
}
},
{
"type": "Multiply",
"quantization": {
"activations": ["q8_tn", "q4_tn", "q8_tn"],
"weights": ["q8_ch", "q4_tn", "q8_ch"]
},
"attributes": {
"unified_scales": "all"
}
},
{
"type": "Maximum",
"quantization": {
"activations": ["q8_tn", "q4_tn", "q8_tn"],
"weights": ["q8_ch", "q4_tn", "q8_ch"]
},
"attributes": {
"unified_scales": "all"
}
},
{
"type": "Less",
"quantization": {
"activations": "q8_tn",
"weights": "q8_ch"
},
"attributes": {
"unified_scales": "all"
}
},
{
"type": "LessEqual",
"quantization": {
"activations": "q8_tn",
"weights": "q8_ch"
},
"attributes": {
"unified_scales": "all"
}
},
{
"type": "Greater",
"quantization": {
"activations": "q8_tn",
"weights": "q8_ch"
},
"attributes": {
"unified_scales": "all"
}
},
{
"type": "GreaterEqual",
"quantization": {
"activations": "q8_tn",
"weights": "q8_ch"
},
"attributes": {
"unified_scales": "all"
}
},
{
"type": "Divide",
"quantization": {
"activations": "q8_tn",
"weights": "q8_ch"
},
"attributes": {
"unified_scales": "all"
}
},
{
"type": "Minimum",
"quantization": {
"activations": ["q8_tn", "q4_tn", "q8_tn"],
"weights": ["q8_ch", "q4_tn", "q8_ch"],
},
"attributes": {
"unified_scales": "all"
}
},
{
"type": "Equal",
"quantization": {
"activations": "q8_tn",
"weights": "q8_ch"
},
"attributes": {
"unified_scales": "all"
}
},
{
"type": "Subtract",
"quantization": {
"activations": ["q8_tn", "q4_tn", "q8_tn"],
"weights": ["q8_ch", "q4_tn", "q8_ch"],
},
"attributes": {
"unified_scales": "all"
}
},
{
"type": "NotEqual",
"quantization": {
"activations": "q8_tn",
"weights": "q8_ch"
},
"attributes": {
"unified_scales": "all"
}
},
{
"type": "FloorMod",
"quantization": {
"activations": "q8_tn",
"weights": "q8_ch"
},
"attributes": {
"unified_scales": "all"
}
},
{
"type": "LogicalOr",
"quantization": {
"activations": "q8_tn",
"weights": "q8_ch"
},
"attributes": {
"unified_scales": "all"
}
},
{
"type": "LogicalXor",
"quantization": {
"activations": "q8_tn",
"weights": "q8_ch"
},
"attributes": {
"unified_scales": "all"
}
},
{
"type": "LogicalAnd",
"quantization": {
"activations": "q8_tn",
"weights": "q8_ch"
},
"attributes": {
"unified_scales": "all"
}
},
{
"type": "LogicalNot",
"quantization": {
"activations": "q8_tn",
"weights": "q8_ch"
}
},
{
"type": "Power",
"quantization": {
"activations": "q8_tn"
}
},
{
"type": "AvgPool",
"quantization": {
"activations": "q8_ch"
}
},
{
"type": "NormalizeL2",
"quantization": {
"activations": "q8_tn"
}
},
{
"type": "ReduceMean",
"quantization": {
"activations": "q8_ch"
}
},
{
"type": "MaxPool"
},
{
"type": "ReduceMax"
},
{
"type": "Interpolate",
"quantization": {
"activations": "q8_tn"
}
},
{
"type": "Concat",
"attributes": {
"unified_scales": "all"
},
"quantization": {
"activations": "q8_tn"
}
},
{
"type": "MVN",
"quantization": {
"activations": "q8_tn"
}
},
{"type": "Reshape"},
{"type": "Flatten"},
{"type": "Squeeze"},
{"type": "Unsqueeze"},
{"type": "Split"},
{"type": "VariadicSplit"},
{"type": "Crop"},
{"type": "Transpose"},
{"type": "Tile"},
{"type": "StridedSlice"},
{"type": "ShuffleChannels"},
{"type": "Broadcast"},
{"type": "Pad"},
{"type": "ConvertLike"},
{"type": "DepthToSpace"}
]
}

3
tools/pot/tests/test_unify_scales.py
View File

@ -18,9 +18,6 @@ from .utils.path import TEST_ROOT
from .utils.data_helper import load_json
TEST_MODELS = [
('mobilenet-v2-pytorch', 'pytorch', 'MinMaxQuantization', 'performance', 'VPU'),
('resnet-50-tf', 'tf', 'DefaultQuantization', 'performance', 'VPU'),
('octave-resnet-26-0.25', 'mxnet', 'DefaultQuantization', 'accuracy', 'VPU'),
('concat_depthwise_model', 'pytorch', 'MinMaxQuantization', 'accuracy', 'CPU'),
]