Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit fd1134a6 authored by Lev Proleev's avatar Lev Proleev
Browse files

Update TRANSPOSE_CONV_2D docs 

* Add info about per-channel quantization
* Update current.txt

Test: mma
Change-Id: I197d984c8b65b4c46bf526eb137f212ad8844926
Merged-In: I197d984c8b65b4c46bf526eb137f212ad8844926
(cherry picked from commit 44015c09)
parent 85714208

current.txt

+1 −1
Original line number Diff line number Diff line
@@ -420,7 +420,7 @@ dd1ec219f5d2e2b33c6c0bcb92e63bbedb36f7c716413462848f6b6ae74fc864 android.hardwar 
92714960d1a53fc2ec557302b41c7cc93d2636d8364a44bd0f85be0c92927ff8 android.hardware.neuralnetworks@1.2::IExecutionCallback

83885d366f22ada42c00d8854f0b7e7ba4cf73ddf80bb0d8e168ce132cec57ea android.hardware.neuralnetworks@1.2::IPreparedModel

e1c734d1545e1a4ae749ff1dd9704a8e594c59aea7c8363159dc258e93e0df3b android.hardware.neuralnetworks@1.2::IPreparedModelCallback

313b341f1f6196a48cf304eaf067f67510c1ebc04df8c7cd536db5611df5c5c2 android.hardware.neuralnetworks@1.2::types

769f8650631eef7a3ceedc8cf130f4b99eb52fe698a11609d55de32985a3dddf android.hardware.neuralnetworks@1.2::types

cf7a4ba516a638f9b82a249c91fb603042c2d9ca43fd5aad9cf6c0401ed2a5d7 android.hardware.nfc@1.2::INfc

abf98c2ae08bf765db54edc8068e36d52eb558cff6706b6fd7c18c65a1f3fc18 android.hardware.nfc@1.2::types

4cb252dc6372a874aef666b92a6e9529915aa187521a700f0789065c3c702ead android.hardware.power.stats@1.0::IPowerStats

neuralnetworks/1.2/types.hal

+37 −12
Original line number Diff line number Diff line
@@ -342,7 +342,7 @@ enum OperationType : int32_t { 
     * * * input.scale * filter.scale).

     *

     * Available since API level 29:

     * * Quantized with symetric per channel quantization for the filter:

     * * Quantized with symmetric per channel quantization for the filter:

     * * * {@link OperandType::TENSOR_QUANT8_ASYMM} for input, and output.

     * * * {@link OperandType::TENSOR_QUANT8_SYMM_PER_CHANNEL} for filter.

     * * * {@link OperandType::TENSOR_INT32} for bias (scale set to 0.0,
@@ -491,7 +491,7 @@ enum OperationType : int32_t { 
     * * * input.scale * filter.scale).

     *

     * Available since API level 29:

     * * Quantized with symetric per channel quantization for the filter:

     * * Quantized with symmetric per channel quantization for the filter:

     * * * {@link OperandType::TENSOR_QUANT8_ASYMM} for input, and output.

     * * * {@link OperandType::TENSOR_QUANT8_SYMM_PER_CHANNEL} for filter.

     * * * {@link OperandType::TENSOR_INT32} for bias (scale set to 0.0,
@@ -3018,7 +3018,7 @@ enum OperationType : int32_t { 
     * * * {@link OperandType::TENSOR_INT32} for bias (with scale set to

     * * * input.scale * filter.scale).

     *

     * * Quantized with symetric per channel quantization for the filter:

     * * Quantized with symmetric per channel quantization for the filter:

     * * * {@link OperandType::TENSOR_QUANT8_ASYMM} for input, and output.

     * * * {@link OperandType::TENSOR_QUANT8_SYMM_PER_CHANNEL} for filter.

     * * * {@link OperandType::TENSOR_INT32} for bias (scale set to 0.0,
@@ -4204,10 +4204,21 @@ enum OperationType : int32_t { 
     * The output dimensions are functions of the filter dimensions, stride, and

     * padding.

     *

     * Supported tensor {@link OperandType}:

     * * {@link OperandType::TENSOR_FLOAT16}

     * * {@link OperandType::TENSOR_FLOAT32}

     * * {@link OperandType::TENSOR_QUANT8_ASYMM}

     * Supported tensor {@link OperandCode} configurations:

     * * 32 bit Floating point :

     * * * {@link OperandType::TENSOR_FLOAT32} for input, filter, output, and bias.

     *

     * * Quantized:

     * * * {@link OperandType::TENSOR_QUANT8_ASYMM} for input, filter, and output.

     * * * {@link OperandType::TENSOR_INT32} for bias (with scale set to

     * * * input.scale * filter.scale).

     *

     * Available since API level 29:

     * * Quantized with symmetric per channel quantization for the filter:

     * * * {@link OperandType::TENSOR_QUANT8_ASYMM} for input, and output.

     * * * {@link OperandType::TENSOR_QUANT8_SYMM_PER_CHANNEL} for filter.

     * * * {@link OperandType::TENSOR_INT32} for bias (scale set to 0.0,

     * * * each value scaling is separate and equal to input.scale * filter.scales[channel]).

     *

     * Supported tensor rank: 4, with "NHWC" or "NCHW" data layout.

     * With the default data layout NHWC, the data is stored in the order of:
@@ -4221,14 +4232,20 @@ enum OperationType : int32_t { 
     *      specifying the input.

     * * 1: A 4-D tensor, of shape

     *      [depth_out, filter_height, filter_width, depth_in], specifying the

     *      filter.

     *      filter. For tensor of type

     *      {@link OperandType::TENSOR_QUANT8_SYMM_PER_CHANNEL} the channel

     *      dimension (extraParams.channelQuant.channelDim) must be set to 0.

     * * 2: A 1-D tensor, of shape [depth_out], specifying the bias. For input

     *      tensor of type {@link OperandType::TENSOR_FLOAT32} or

     *      {@link OperandType::TENSOR_FLOAT16}, the bias should be of the

     *      same type. For input tensor of type

     *      {@link OperandType::TENSOR_QUANT8_ASYMM}, the bias should be

     *      of {@link OperandType::TENSOR_INT32}, with zeroPoint of 0 and

     *      bias_scale == input_scale * filter_scale.

     *      bias_scale == input_scale * filter_scale. For filter tensor of

     *      {@link OperandType::TENSOR_QUANT8_SYMM_PER_CHANNEL}, the bias

     *      must be of {@link OperandType::TENSOR_INT32}, with zeroPoint of

     *      0 and bias_scale of 0. The actual scale of each value 'i' is equal

     *      to bias_scale[i] = input_scale * filter_scale[i].

     * * 3: An {@link OperandType::INT32} scalar, specifying the padding on

     *      the left, in the ‘width’ dimension.

     * * 4: An {@link OperandType::INT32} scalar, specifying the padding on
@@ -4252,14 +4269,20 @@ enum OperationType : int32_t { 
     *      specifying the input.

     * * 1: A 4-D tensor, of shape

     *      [depth_out, filter_height, filter_width, depth_in], specifying the

     *      filter.

     *      filter. For tensor of type

     *      {@link OperandType::TENSOR_QUANT8_SYMM_PER_CHANNEL} the channel

     *      dimension (extraParams.channelQuant.channelDim) must be set to 0.

     * * 2: A 1-D tensor, of shape [depth_out], specifying the bias. For input

     *      tensor of type {@link OperandType::TENSOR_FLOAT32} or

     *      {@link OperandType::TENSOR_FLOAT16}, the bias should be of the

     *      same type. For input tensor of type

     *      {@link OperandType::TENSOR_QUANT8_ASYMM}, the bias should be

     *      of {@link OperandType::TENSOR_INT32}, with zeroPoint of 0 and

     *      bias_scale == input_scale * filter_scale.

     *      bias_scale == input_scale * filter_scale. For filter tensor of

     *      {@link OperandType::TENSOR_QUANT8_SYMM_PER_CHANNEL}, the bias

     *      must be of {@link OperandType::TENSOR_INT32}, with zeroPoint of

     *      0 and bias_scale of 0. The actual scale of each value 'i' is equal

     *      to bias_scale[i] = input_scale * filter_scale[i].

     * * 3: An {@link OperandType::TENSOR_INT32} tensor, specifying the output

     *      tensor shape.

     * * 4: An {@link OperandType::INT32} scalar, specifying the implicit
@@ -4279,7 +4302,9 @@ enum OperationType : int32_t { 
     * * 0: The output 4-D tensor, of shape

     *      [batches, out_height, out_width, depth_out]. For output tensor of

     *      {@link OperandType::TENSOR_QUANT8_ASYMM}, the following condition

     *      must be satisfied: output_scale > input_scale * filter_scale.

     *      must be satisfied: output_scale > input_scale * filter_scale (for

     *      filter tensor of {@link OperandType::TENSOR_QUANT8_SYMM_PER_CHANNEL}

     *      this condition must be true for all filter scales).

     *

     * Available since API level 29.

     */