_load_for_executorch pybinding cannot load model for training

[dvorjackz]

🐛 Describe the bug

_load_for_executorch pybinding cannot load joint graph for phi-3-mini-lora because load_into is not implemented in MmapLoader, and load_into is used by Program's load_mutable_subsegment_into to load weights for training in TensorParser.

Versions

Collecting environment information... PyTorch version: 2.5.0.dev20240821 Is debug build: False CUDA used to build PyTorch: None ROCM used to build PyTorch: N/A

OS: macOS 14.6.1 (arm64) GCC version: Could not collect Clang version: 15.0.0 (clang-1500.3.9.4) CMake version: version 3.30.0-rc3 Libc version: N/A

Python version: 3.10.0 (default, Mar 3 2022, 03:54:28) [Clang 12.0.0 ] (64-bit runtime) Python platform: macOS-14.6.1-arm64-arm-64bit Is CUDA available: False CUDA runtime version: No CUDA CUDA_MODULE_LOADING set to: N/A GPU models and configuration: No CUDA Nvidia driver version: No CUDA cuDNN version: No CUDA HIP runtime version: N/A MIOpen runtime version: N/A Is XNNPACK available: True

CPU: Apple M1 Pro

Versions of relevant libraries: [pip3] executorch==0.4.0a0+3ad2959 [pip3] numpy==1.21.3 [pip3] torch==2.5.0.dev20240821 [pip3] torchaudio==2.4.0.dev20240821 [pip3] torchsr==1.0.4 [pip3] torchvision==0.20.0.dev20240821 [conda] executorch 0.4.0a0+3ad2959 pypi_0 pypi [conda] numpy 1.21.3 pypi_0 pypi [conda] torch 2.5.0.dev20240821 pypi_0 pypi [conda] torchaudio 2.4.0.dev20240821 pypi_0 pypi [conda] torchsr 1.0.4 pypi_0 pypi [conda] torchvision 0.20.0.dev20240821 pypi_0 pypi

[JacobSzwejbka]

This is because the mmap_data_loader does not currently implement load_into.