From 997dd8c4fc5a3fef5dd1d6bce829d4d241e4becb Mon Sep 17 00:00:00 2001 From: Nina Drozd Date: Tue, 21 May 2019 11:17:10 +0100 Subject: IVGCVSW-3088 Update Readme for 19.05 * Added Readme file for ArmnnQuantizer * Added section about ArmnnQuantizer in armnn Readme file * Updated ModelAccuracyTool Readme file with default values for --compute Signed-off-by: Nina Drozd Change-Id: I5fcead522b70086dcf63dfc6c77910a7d33d83f0 --- src/armnnQuantizer/README.md | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+) create mode 100644 src/armnnQuantizer/README.md (limited to 'src/armnnQuantizer/README.md') diff --git a/src/armnnQuantizer/README.md b/src/armnnQuantizer/README.md new file mode 100644 index 0000000000..cad382078c --- /dev/null +++ b/src/armnnQuantizer/README.md @@ -0,0 +1,18 @@ +# The ArmnnQuantizer + +The `ArmnnQuantizer` is a program for loading a 32-bit float network into ArmNN and converting it into a quantized asymmetric 8-bit or quantized symmetric 16-bit network. +It supports static quantization by default, dynamic quantization is enabled if CSV file of raw input tensors is provided. Run the program with no arguments to see command-line help. + + +|Cmd:||| +| ---|---|---| +| -h | --help | Display help messages | +| -f | --infile | Input file containing float 32 ArmNN Input Graph | +| -s | --scheme | Quantization scheme, "QAsymm8" or "QSymm16". Default value: QAsymm8 | +| -c | --csvfile | CSV file containing paths for raw input tensors for dynamic quantization. If unset, static quantization is used | +| -p | --preserve-data-type | Preserve the input and output data types. If unset, input and output data types are not preserved | +| -d | --outdir | Directory that output file will be written to | +| -o | --outfile | ArmNN output file name | + +Example usage:
+./ArmnnQuantizer -f /path/to/armnn/input/graph/ -s "QSymm16" -c /path/to/csv/file -p 1 -d /path/to/output -o outputFileName \ No newline at end of file -- cgit v1.2.1