[TIDL-4273] Correct documentation errors

docs/supported_ops_rts_versions.md

@@ -11,7 +11,7 @@
 | 3  | TIDL_PoolingLayer              | MaxPool<br>AveragePool<br>GlobalAveragePool | MAX_POOL_2D<br>AVERAGE_POOL_2D<br>MEAN     | <ul><li>Pooling has been validated for the following kernel sizes: 3x3,2x2s,1x1 with stride 1 and stride 2 (both horizontal and vertical dimensions)</li><li> Max pooling supports 1x1 filters with asymmetric stride</li><li> Max pooling additionally supports 1x2,1x3 filters with a stride of 2 (Along the horizontal direction) & 2x1,3x1 filters with a stride of 2 (Along the vertical direction)</li></ul> |
 | 4  | TIDL_EltWiseLayer              | Add<br>Mul                                  | ADD<br>MUL                                 | <ul><li>Support for 2 input tensors validated extensively, multiple input tensors have limited validation</li><li>Supports broadcasting of dimensions above width</li></ul> |
 | 5  | TIDL_InnerProductLayer         | Gemm, MatMul                                | FULLY_CONNECTED                            | <ul><li> Broadcast is only supported in channel dimension </li><li>For TDA4VM variable input case, doesn’t support unsigned input </li><li>Higher dimensional matmuls can be realized by reshaping the dimensions higher than 3rd dimension into the 3rd dimension</li></ul>|
-| 6  | TIDL_SoftMaxLayer              | Softmax                                     | SOFTMAX                                    | <ul><li>Supports 8-bit inputs with 8-bit outputs with axis support for width (axis=-1) for any NxCxHxW tensor</li><li>Supports integer (8/16-bit) to float softmax only for flattened inputs</li></ul> |
+| 6  | TIDL_SoftMaxLayer              | Softmax                                     | SOFTMAX                                    | <ul><li>Supports 8-bit(/16-bit) inputs with 8-bit(/16-bit) outputs (both input and output are of the same bit-depth) with axis support for width (axis=-1) for any NxCxHxW tensor</li><li>Supports integer (8/16-bit) to float softmax only for flattened inputs</li></ul> |
 | 7  | TIDL_Deconv2DLayer             | ConvTranspose                               | TRANSPOSE_CONV                             | <ul><li>Only 8x8, 4x4 and 2x2 kernel with 2x2 stride is supported. It is recommended to use Resize/Upsample to get better performance. This layer is not supported in 16-bit for AM62A/AM67A</li></ul>|
 | 8  | TIDL_ConcatLayer               | Concat                                      | CONCATENATION                              | <ul><li>Concat is supported on channel, height or width axis</li></ul>|
 | 9  | TIDL_SliceLayer                | Split<br>Slice                                       | NA                                         | <ul><li>Slice is supported on all axes except for the batch axis & only one axis can be sliced per operator</li><li>[Patch merging](./tidl_fsg_vtfr.md#patch-merging) expressed with strided slice will be transformed into a transpose layer</li></ul>|

examples/osrt_python/README.md

@@ -193,7 +193,8 @@ Please refer [Quantization](../../docs/tidl_fsg_quantization.md) for more detail
 |       Name      |                      Description                        |  Supported values/range  |       Default values      |  Option Type  |  Additional details  
 |:-------------------|:--------------------------------------------------------|:--------------|:------------|:--------------|:------------------------|
 | advanced_options:quantization_scale_type   | This option specifies type of quantization style to be used for model quantization | 0 - non-power-of-2,<br> 1 - power-of-2 <br> 3 - TF-Lite pre-quantized model <br> 4 - Asymmetric, Per-channel Quantization       | 0    | Model compilation | Refer [Quantization](../../docs/tidl_fsg_quantization.md) for more details |
-| advanced_options:pre-quantized_model    | This option enables reading of scales and zero points from an ONNX QDQ model and bypasses the need for calibration | 0 - disable, <br> 1 enable    | 0       | Model compilation | This impacts only ONNX models, for TF-Lite models quantization_scale_type=3 has the same effect |
+| advanced_options:quant_params_proto_path   | This option allows you to configure quantization scales manually by specifying the min/max values of outputs |  String      | ""  | Model compilation | Refer to [Quantization Parameters](../../docs/tidl_quantParams.md) for further details |
+| advanced_options:prequantized_model    | This option enables reading of scales and zero points from an ONNX QDQ model and bypasses the need for calibration | 0 - disable, <br> 1 enable    | 0       | Model compilation | This impacts only ONNX models, for TF-Lite models quantization_scale_type=3 has the same effect |
 | advanced_options:high_resolution_optimization    | This option enables performance optimization for high resolution models  | 0 - disable, <br> 1 enable    | 0       | Model compilation | |
 | advanced_options:add_data_convert_ops   | This option embeds input and output format conversions (layout, data type, etc.) as part of model and performs the same in DSP instead of ARM    | 0 - disable, <br> 1 - Input format conversion <br> 2 - output format conversion <br> 3 - Input and output format conversion |    0      | Model compilation | This is currently an experimental feature |
 | advanced_options:network_name   | This option allows the user to set the network name (used for the name of the subgraph being delegated to C7x/MMA). If your model contains a network name, it will get used by default    | String |    "Subgraph"      | Model compilation |  |