MLSDK documentation

MLSDK is an SDK for utilizing MN-Core series processors from PyTorch programs.

Table of Contents:

• 1. Getting Started
  • 1.1. Prerequisites
  • 1.2. Directory Structure
  • 1.3. Running Sample Programs
  • 1.4. Sample Program Explanation
  • 1.5. Advanced Topics
• 2. Machine Learning Tutorial
  • 2.1. Prerequisites
  • 2.2. Running the Sample Program
  • 2.3. Sample Program Explanation
  • 2.4. Advanced Topics
• 3. Porting Tutorial
  • 3.1. Porting Process
  • 3.2. Advanced Topics
• 4. Advanced Features
  • 4.1. Features of gpfn3-smi
  • 4.2. Compile Options
  • 4.3. Directly loading codegen_dir
  • 4.4. Codegen Dashboard
  • 4.5. Data Transfer via TensorProxy
• 5. FAQ
  • 5.1. About MLSDK
  • 5.2. Common Errors and Solutions
• 6. Technical Notes
  • 6.1. MLSDK for MN-Core
  • 6.2. MLSDK Pipeline
  • 6.3. codegen Glossary
  • 6.4. Appendix
• 7. Hardware Specification
  • 7.1. Hierarchical Architecture
  • 7.2. Memory Word
  • 7.3. Floating Point Format
• 8. Gallery
  • 8.1. API Examples
  • 8.2. Advanced Examples
  • 8.3. Use Cases
• 9. API Reference
  • MNDevice
  • Context
  • CompiledFunction
  • TensorProxy
  • mlsdk.TensorLike
  • CacheOptions
  • MNCoreOptimizer
  • MNCoreSGD
  • MNCoreAdam
  • MNCoreAdamW
  • MNCoreLRScheduler
  • set_buffer_name_in_optimizer()
  • get_tensor_name()
  • set_tensor_name()
  • set_tensor_name_in_module()
  • storage.path()
  • path()
  • trace_event()
  • trace_scope()
  • LintLevel
  • lint()
  • LintResult