What is the relationship between Vela versions and Ethos‑U software versions?

I would like to understand the relationship between Vela compiler versions and Arm Ethos‑U software components.

  1. What is the version relationship between Vela and Ethos‑U?

    • Is there an officially recommended or validated mapping between:
      • Vela versions
      • Ethos‑U IP versions (e.g. Ethos‑U55)
      • Ethos‑U Core Driver
      • Ethos‑U Core Software
      • Ethos‑U Core Platform
    • If such a relationship exists, is there a compatibility or version mapping table that can be shared?
  2. Specific question for the latest Vela release

    • The current latest Vela version is Vela 5.0.0.
    • For a system based on Ethos‑U55, which versions of the following components are recommended or required to be used together with Vela 5.0.0?
      • Ethos‑U Core Driver
      • Ethos‑U Core Software
      • Ethos‑U Core Platform

Any official guidance, documentation, or best‑practice recommendations would be greatly appreciated.

Parents
  • Hi Ben,

    Thanks for your reply.

    I am currently using Vela 4.2.0 with Ethos‑U55, and the inference processing works well with correct results.

    Recently, I am trying to upgrade my inference workflow to the latest Vela 5.0.0. During this process, I noticed that Vela 5.0.0 introduces new parameters in the vela.ini configuration file compared with Vela 4.2.0.

    When I reuse my existing vela.ini file from Vela 4.2.0 directly with Vela 5.0.0, I encounter the following error (see below). This seems to indicate that my current configuration file is no longer fully compatible and needs to be updated.

    My questions are:

    • What are the required changes when migrating a vela.ini file from Vela 4.2.0 to 5.0.0?
    • Is there any guideline or example vela.ini for Vela 5.0.0 with Ethos‑U55 that can be referenced?

    Thanks in advance for your help.

    Vela version: 5.0.0
    Target NPU: Ethos-U55
    Previous Vela version: 4.2.0
    Error message:
    (py_312) C:\xxxx>vela model\tf_fp32\xxx_int8.tflite --accelerator-config=ethos-u55-256 --optimise Performance --config model\tf_fp32\vela.ini --system-config=RA8P1 --memory-mode=Sram_Only --output-dir model\tf_fp32\out_xxx_int8_sram_mram
    Info: Changing const_mem_area from Sram to OnChipFlash. This will use the same characteristics as Sram.
    Memory '' not found in configured memories.
    Traceback (most recent call last):
    File "<frozen runpy>", line 198, in _run_module_as_main
    File "<frozen runpy>", line 88, in _run_code
    File "C:\Users\abc\AppData\Local\miniforge3\envs\py_312\Scripts\vela.exe\__main__.py", line 6, in <module>
    File "C:\Users\abc\AppData\Local\miniforge3\envs\py_312\Lib\site-packages\ethosu\vela\vela.py", line 1249, in main
    process_regor(
    File "C:\Users\abc\AppData\Local\miniforge3\envs\py_312\Lib\site-packages\ethosu\vela\vela.py", line 135, in process_regor
    compiled_model = regor.compile(accelerator, network, fmt, system_config, options=options, verbose=True)
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    RuntimeError: Compilation failed: Staging memory not configured

Reply
  • Hi Ben,

    Thanks for your reply.

    I am currently using Vela 4.2.0 with Ethos‑U55, and the inference processing works well with correct results.

    Recently, I am trying to upgrade my inference workflow to the latest Vela 5.0.0. During this process, I noticed that Vela 5.0.0 introduces new parameters in the vela.ini configuration file compared with Vela 4.2.0.

    When I reuse my existing vela.ini file from Vela 4.2.0 directly with Vela 5.0.0, I encounter the following error (see below). This seems to indicate that my current configuration file is no longer fully compatible and needs to be updated.

    My questions are:

    • What are the required changes when migrating a vela.ini file from Vela 4.2.0 to 5.0.0?
    • Is there any guideline or example vela.ini for Vela 5.0.0 with Ethos‑U55 that can be referenced?

    Thanks in advance for your help.

    Vela version: 5.0.0
    Target NPU: Ethos-U55
    Previous Vela version: 4.2.0
    Error message:
    (py_312) C:\xxxx>vela model\tf_fp32\xxx_int8.tflite --accelerator-config=ethos-u55-256 --optimise Performance --config model\tf_fp32\vela.ini --system-config=RA8P1 --memory-mode=Sram_Only --output-dir model\tf_fp32\out_xxx_int8_sram_mram
    Info: Changing const_mem_area from Sram to OnChipFlash. This will use the same characteristics as Sram.
    Memory '' not found in configured memories.
    Traceback (most recent call last):
    File "<frozen runpy>", line 198, in _run_module_as_main
    File "<frozen runpy>", line 88, in _run_code
    File "C:\Users\abc\AppData\Local\miniforge3\envs\py_312\Scripts\vela.exe\__main__.py", line 6, in <module>
    File "C:\Users\abc\AppData\Local\miniforge3\envs\py_312\Lib\site-packages\ethosu\vela\vela.py", line 1249, in main
    process_regor(
    File "C:\Users\abc\AppData\Local\miniforge3\envs\py_312\Lib\site-packages\ethosu\vela\vela.py", line 135, in process_regor
    compiled_model = regor.compile(accelerator, network, fmt, system_config, options=options, verbose=True)
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    RuntimeError: Compilation failed: Staging memory not configured

Children
No data