Jacinto 7 TIDL Release Notes

Version: 09.01.00, Date: Nov 29, 2023

Contents

  1. Introduction
  2. Licensing
  3. What's New
  4. Documentation
  5. Upgrade and Compatibility Information
  6. Device Support
  7. Validation Information
  8. Fixed Issues
  9. Known Issues
  10. Change Request
  11. Technical Support
  12. Package Versioning

Introduction

This package contains Texas Instrument's Deep learning solution for J721E SoCs from the Jacinto 7 family

Licensing

The licensing information of this product, as well as any third-party components included which are made available under a number of other open-source licenses are enumerated as part of the manifest. Note : Below Links would work only after installing the package

Documentation

1 User's Guide [HTML]
2 Software Manifest [HTML]

What's New

Major new features in this release are:

  1. Support for quantized ONNX models with Tensor-oriented (QDQ; Quantize and DeQuantize) format
  2. Support for ONNX opset 18
  3. Support for convolution and pooling with asymmetric strides
  4. Support for 8-bit operators for Vision Transformers:
    • Support for Self Attention Blocks:
      1. Support for MatMul Operator (Signed)
      2. Support for Softmax Operator along the width axis
    • Support for Layernorm Operator (Signed) on the width axis
    • Support for GELU activation
    • Support for Patch Embedding
    • Support for Patch Merging
  5. Support for Sum and Max reduction attributes in ScatterND operator
  6. Please refer to edgeai-tidl-tools for further details and known limitations on operator support

Upgrade and Compatibility Information

NA

Device Support

Target SoC Target Plaform Build environment (OS) Target environment (OS)
J721E EVM x86_64 Linux PC x86_64 (Linux) NA

Validation Information

This release was built and validated using the following tools:

Dependencies (included in Processor SDK RTOS):

Dependencies (NOT included in Processor SDK RTOS):

Refer User's Guide for instructions to install and setup above dependencies.

Fixed Issues

ID Description Module Affected Versions Affected Platforms
TIDL-2593 OSRT : Softmax layer with non-flattened input gets delegated to C7X-MMA instead of ARM OSRT 08.02.00 J721E
TIDL-3551 Networks with multiple branches may result into corruption of padding region OSRT 09.00.00 J721E
TIDL-3605 TIDL does not transpose weights correctly when groups = 1 for transposed convolution (ONNX) OSRT 08.06.00,09.00.00 J721E
TIDL-3623 LUT Table overflow for Sigmoid & ELU Results in incorrect output OSRT 08.06.00,09.00.00 J721E

Known Issues

ID Description Module Reported in Release Affected Platforms Occurrence Workaround in this release
TIDL-1872 Import failure for deeplabv3 (mobilenetv2 backbone) based OD+SEG Model COMPUTE 08.01.00 J721E Rare
TIDL-2878 OSRT : Optimized OD post processing results in crash in case all convolution heads are not part of same subgraph OSRT 08.04.00 J721E Rare NA
TIDL-2821 16-bit convolution with pad = 0 results in hang OSRT 08.04.00 J721E Rare NA
TIDL-3416 Performance degradation observed in 9.0/9.1 release compared to 8.6 release due to certain operators TIDL 09.00.00 J721E Rare NA
TIDL-3645 MatMul not supported with two inputs of different datatype COMPUTE 09.01.00 J721E Rare NA
TIDL-3646 Known performance issue of Softmax with "axis" support COMPUTE 09.01.00 J721E Rare NA
TIDL-3651 Known performance issue of MatMul operator in TIDL COMPUTE 09.01.00 J721E Rare NA

For latest status of the issues and any new issues found post this release, please refer TI External software Incident Report. Issues on this product can be filtered with search query as project = EXT_EP AND issuetype in (Bug, Enhancement) AND Product ~ "TI Deep Learning Library"

Change Request

ID Description Original Fix Version New Fix Version Additional Information
JACINTOREQ-7450 Re-Prioritization of TIDL MRs (support for ) for SDK 9.1 09.01.00 09.02.00 Support for some transformer models like SWIN, DETR, SegFormer and few performance improvements have been descoped from 9.1

Technical Support

For technical support, please post your questions on TI E2E Forum for Processors.

For additional assistance, contact local TI Field Application Engineer

Package Versioning

Each package version is composed of 4 period-delimited numbers - represented here by the letters M, m, p and b [M.m.p.b]. The table below provides a descriptive reference regarding package version numbering.

Digit Meaning Description
1 (M=Major) Major revision Incremented when the new version is substantially different from the previous For example, a new module added or an existing module's algorithm significantly altered.
2 (m=minor) Minor revision Incremented when the new version has changed but not in a major way. For example, some minor changes in the API or feature set.
3 (p=patch) Patch number Incremented for all other source code changes. This include any packaging support code.
4 (b=build) Build number Incremented for each release delivery to CM. Reset for any change to M, m or p

© Copyright 1995-2021, Texas Instruments Incorporated. All rights reserved.
Trademarks | Privacy policy | Terms of use | Terms of sale