Huawei uses machine translation combined with human proofreading to translate this document to different languages in order to help you better understand the content of this document.
Note: Even the most advanced machine translation cannot match the quality of professional translators.
Huawei shall not bear any responsibility for translation accuracy and it is recommended that you refer to the English document (a link for which has been provided).
Introduction- ATC Tool Instructions
- Introduction
- Getting Started with ATC
- Command-line Options
- Feature Configuration
- Operator Specifications (Caffe)
- Operator Specifications (TensorFlow)
- Error Codes
- Command Line Errors
- Insufficient Free Memory in Tuning
- Model Parsing Errors (Caffe)
- Model Parsing Errors (TensorFlow)
- Operator Information Library Verification Errors
- Invalid GE Operator infershape
- Model Parsing Errors
- Memory Errors
- Environment Variable Configuration Errors
- Unknown Errors
- Warnings
- Model Optimization Errors
- Model Build Errors
- Failure to Initialize Modules
- Operator Parsing Errors
- Operator Build Errors
- Connection Failures Between Auto Tune and Device
- RTS Verification Failures
- FAQs
- Ascend Model Compression Toolkit Instructions (Caffe)
- Introduction
- AMCT Installation
- Calibration-based Quantization
- Retrain-based Quantization
- Tensor Decomposition
- AMCT Update
- AMCT Uninstallation
- API Description
- FAQs
- Failed to Install the Caffe Environment in CLI Mode
- An Error Message Is Displayed During python3-tk Installation
- Proto Merging Error
- "RuntimeError: Cannot find scale_d of layer '**' in record file" Is Displayed During Quantization
- "UserWarning: Matplotlib is currently using agg, which is a non-GUI backend, so cannot show the figure." Is Displayed When Quantizing an Object Detection Network
- "IfmrQuantWithOffset scale is illegal" Is Displayed During Calibration
- Appendixes
- Ascend Model Compression Toolkit Instructions (TensorFlow)
- Introduction
- AMCT Installation
- Calibration-based Quantization
- Retrain-based Quantization
- Tensor Decomposition
- AMCT Update
- AMCT Uninstallation
- API Description
- FAQs
- Appendixes
- Sample Package Content
- Simplified Quantization Configuration File
- Simplified Retrain Configuration File
- Quantization Factor Record File
- Tensor Decomposition Specification Reference of Open-Source Networks
- What Do I Do If My TensorFlow Network Output Node Is Changed by AMCT?
- How Do I Restore the Model Training Parameters After Quantization Operators Are Inserted?
- Model Accuracy Analyzer Instructions
- Auto Tune Tool Instructions
- Profiling Tool Instructions
- Environment Preparation
- Profiling by Using the hiprof Command
- Profiling by Calling the acl.json File
- Profiling by Calling the AscendCL API
- Appendixes
- Metric Query
- Scripts and Interfaces
- Description of ai_core_op_summary_{device_id}.csv
- AI Core Metrics
- Exporting Profiling Results
- Profiling Workflow and Maintenance & Testing Description
- Profiling Fails Due to App Execution Error
- Profiling Fails with Exit Code
- Profiling Fails with AI Core Metrics
- ada Fails Due to Profiling Initialization Failure
- Profiling Fails Due to Runtime Error
- Profiling Fails with Prompt "CFUNCTYPE(c_int)(lambda:None)"
- Profiling Fails with Prompt "Unable to get ai_core_pmu_events data."
- AICore Error Analyzer Instructions
- Black Box Instructions
- ADC Instructions
- ADC Command Reference
- Introduction
- Environment Preparation
- Delivering a Command to Host
- Transferring a File to Host
- Fetching a File from Host
- Setting Log Levels
- Collecting Profile Data
- Obtaining the Information of Devices Connected to Host
- Generating Files with Encrypted Character Strings
- Obtaining the List of Model Dump File Paths
- Detecting the Heartbeat Between ADC and ADA
- Synchronizing the Time of ADC and ADA
- Viewing the Help Information
- Reference
- ADX Certificate Replacement Guide
- Appendix