ruGPT3XL_generation.ipynb not working

[NikitaCartes]

ruGPT3XL_generation doesn't work in Google Colab:

In cell from src.xl_wrapper import RuGPT3XL

ModuleNotFoundError                       Traceback (most recent call last)
[<ipython-input-24-c08b8e4ed74a>](https://localhost:8080/#) in <module>
----> 1 from src.xl_wrapper import RuGPT3XL

ModuleNotFoundError: No module named 'src'```

[horror]

!DS_BUILD_CPU_ADAM=1 DS_BUILD_SPARSE_ATTN=1 pip install deepspeed

Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/ Collecting deepspeed Downloading deepspeed-0.8.0.tar.gz (749 kB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 749.9/749.9 KB 29.4 MB/s eta 0:00:00 error: subprocess-exited-with-error

× python setup.py egg_info did not run successfully. │ exit code: 1 ╰─> See above for output.

note: This error originates from a subprocess, and is likely not a problem with pip. Preparing metadata (setup.py) ... error error: metadata-generation-failed

× Encountered error while generating package metadata. ╰─> See above for output.

note: This is an issue with the package mentioned above, not pip. hint: See above for details.

Проверьте в гугл калабе не работает

[romanzoniit]

!DS_BUILD_CPU_ADAM=1 DS_BUILD_SPARSE_ATTN=1 pip install

deepspeed

Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/ Collecting deepspeed Downloading deepspeed-0.8.0.tar.gz (749 kB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 749.9/749.9 KB 29.4 MB/s eta 0:00:00 error: subprocess-exited-with-error

× python setup.py egg_info did not run successfully. │ exit code: 1 ╰─> See above for output.

note: This error originates from a subprocess, and is likely not a problem with pip. Preparing metadata (setup.py) ... error error: metadata-generation-failed

× Encountered error while generating package metadata. ╰─> See above for output.

note: This is an issue with the package mentioned above, not pip. hint: See above for details.

Проверьте в гугл калабе не работает

Установите версию triton pip install triton==1.0.0 После изменения у меня заработало. @king-menin исправьте в google colab upd: далее вылезла эта ошибка

IndexError        Traceback (most recent call last)
[<ipython-input-38-6c27b7064ddf>](https://localhost:8080/#) in <module>
----> 1 logits = gpt("Кто был президентом США в 2020? ").logits

18 frames
[/usr/local/lib/python3.8/dist-packages/triton/code_gen.py](https://localhost:8080/#) in _compile(self, device, attributes, constants, num_warps, num_stages, force_nc_cache, *wargs, **meta)
    548         tt_device = _triton.driver.cu_device(device.index, False)
    549         # Compile to machine code
--> 550         mod, ker, shared_mem, ir_asm = _triton.code_gen.add_passes_to_emit_bin(generator.module, tt_device, num_warps, num_stages, force_nc_cache)
    551         if shared_mem > tt_device.max_shared_memory():
    552             raise  OutOfResources(shared_mem, tt_device.max_shared_memory(), "shared memory")
IndexError: map::at

[LEv145]

Если использовать конкретные версии из ruGPT3XL_finetune_example.ipynb, то ошибка та же:

[/usr/local/lib/python3.8/dist-packages/triton/code_gen.py](https://localhost:8080/#) in _compile(self, device, attributes, constants, num_warps, num_stages, force_nc_cache, *wargs, **meta)
    548         tt_device = _triton.driver.cu_device(device.index, False)
    549         # Compile to machine code
--> 550         mod, ker, shared_mem, ir_asm = _triton.code_gen.add_passes_to_emit_bin(generator.module, tt_device, num_warps, num_stages, force_nc_cache)
    551         if shared_mem > tt_device.max_shared_memory():
    552             raise  OutOfResources(shared_mem, tt_device.max_shared_memory(), "shared memory")

IndexError: map::at

Но случайно после опытов у меня получилось заставить это запускаться: https://github.com/ai-forever/ru-gpts/issues/105#issue-1588278808