Cannot install on Windows

Question

Cannot install on Windows

SoftologyPro opened this issue 2 years ago · 17 comments

git clone https://github.com/declare-lab/tango/
cd tango
pip install -r requirements.txt

Ends with

Collecting accelerate==0.18.0
  Using cached accelerate-0.18.0-py3-none-any.whl (215 kB)
Collecting bitsandbytes==0.38.1
  Using cached bitsandbytes-0.38.1-py3-none-any.whl (104.3 MB)
Collecting black==22.3.0
  Using cached black-22.3.0-cp310-cp310-win_amd64.whl (1.1 MB)
Collecting compel==1.1.3
  Using cached compel-1.1.3-py3-none-any.whl (24 kB)
Collecting d4rl==1.1
  Using cached d4rl-1.1-py3-none-any.whl (26.4 MB)
Collecting data_generator==1.0.1
  Using cached Data_Generator-1.0.1-py3-none-any.whl (11 kB)
Collecting deepdiff==6.3.0
  Using cached deepdiff-6.3.0-py3-none-any.whl (69 kB)
Collecting diffusion==6.9.1
  Using cached diffusion-6.9.1-1-py3-none-any.whl (179 kB)
Collecting einops==0.6.1
  Using cached einops-0.6.1-py3-none-any.whl (42 kB)
Collecting flash_attn==1.0.3.post0
  Using cached flash_attn-1.0.3.post0.tar.gz (2.0 MB)
  Preparing metadata (setup.py) ... error
  error: subprocess-exited-with-error

  × python setup.py egg_info did not run successfully.
  │ exit code: 1
  ╰─> [6 lines of output]
      Traceback (most recent call last):
        File "<string>", line 2, in <module>
        File "<pip-setuptools-caller>", line 34, in <module>
        File "C:\Users\Jason\AppData\Local\Temp\pip-install-ftl4rk2a\flash-attn_362c7a8bea504f679c59d4a480161f3e\setup.py", line 6, in <module>
          from packaging.version import parse, Version
      ModuleNotFoundError: No module named 'packaging'
      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
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.

Answer 1 · 2023-04-25T02:57:25.000Z

I have updated the requirements.txt file. Can you try a clean install again?

Answer 2 · 2023-04-25T03:10:09.000Z

Collecting importlib_metadata==6.3.0
  Using cached importlib_metadata-6.3.0-py3-none-any.whl (22 kB)
ERROR: Ignored the following versions that require a different python version: 6.6.0 Requires-Python >=3.7,<3.10; 6.6.0rc2 Requires-Python >=3.7,<3.10; 6.6.1 Requires-Python >=3.7,<3.10; 6.6.2 Requires-Python >=3.7,<3.10; 6.6.3 Requires-Python >=3.7,<3.10; 6.6.4 Requires-Python >=3.7,<3.10; 6.6.5 Requires-Python >=3.7,<3.10; 6.6.6 Requires-Python >=3.7,<3.10; 6.6.7 Requires-Python >=3.7,<3.10; 6.6.8 Requires-Python >=3.7,<3.10; 6.6rc1 Requires-Python >=3.7,<3.9
ERROR: Could not find a version that satisfies the requirement intel_extension_for_pytorch==2.0.0 (from versions: none)
ERROR: No matching distribution found for intel_extension_for_pytorch==2.0.0

To be sure I have a clean setup I am making a new python venv before each test.
ie

python -m venv venv
venv\scitps\activate

Answer 3 · 2023-04-25T03:20:49.000Z

I do not encounter this on Linux. Can you try installing from the latest requirements file? Otherwise, you can try installing it separately following the original instructions.

Answer 4 · 2023-04-25T03:32:58.000Z

Neither v 2.0.0 or 1.113.0 install on Windows (CPU or GPU versions).
If I delete the line for intel_extension_for_pytorch then it next fails at
ERROR: No matching distribution found for jaxlib>=0.1.37
Are all these packages really needed just for tango?

Answer 5 · 2023-04-25T03:39:22.000Z

OK, trying to build intel_extension_for_pytorch gives the error

File "D:\intel-extension-for-pytorch\setup.py", line 726, in make_relative_rpath
    raise "Windows support is in the plan. Intel Extension for PyTorch supports Linux now."

So, no go for Windows. Hopefully a future release of tango will consider Windows support.

Answer 6 · 2023-04-25T03:44:20.000Z

I created the requirements file using pipreqs. The dependencies from audioldm and audioldm_eval were also included here. But I suspect many of the dependencies are not required. I will push a minimal requirements file in some time, and you can give it a try :)

Answer 7 · 2023-04-25T05:11:37.000Z

Hey, I have updated the file with a minimal set of requirements. Let me know if it works now

Answer 8 · 2023-04-25T05:53:05.000Z

OK, getting closer.
The pip install -r requirements.txt finishes without error.
But then testing with this script

import soundfile as sf
from tango import Tango
tango = Tango("declare-lab/tango")
prompt = "An audience cheering and clapping"
audio = tango.generate(prompt, steps=200)
sf.write(f"{prompt}.wav", audio, samplerate=16000)

gives this error

Traceback (most recent call last):
  File "D:\tango\tango_generate.py", line 3, in <module>
    from tango import Tango
  File "D:\tango\tango.py", line 5, in <module>
    from models import AudioDiffusion, DDPMScheduler
  File "D:\tango\models.py", line 14, in <module>
    from audioldm.audio.stft import TacotronSTFT
  File "D:\tango\audioldm\__init__.py", line 3, in <module>
    from .pipeline import *
  File "D:\tango\audioldm\pipeline.py", line 11, in <module>
    from audioldm.audio import wav_to_fbank, TacotronSTFT, read_wav_file
  File "D:\tango\audioldm\audio\__init__.py", line 2, in <module>
    from .stft import TacotronSTFT
  File "D:\tango\audioldm\audio\stft.py", line 5, in <module>
    from librosa.util import pad_center, tiny
  File "D:\tango\voc_tango\lib\site-packages\librosa\__init__.py", line 211, in <module>
    from . import decompose
  File "D:\tango\voc_tango\lib\site-packages\librosa\decompose.py", line 19, in <module>
    import sklearn.decomposition
  File "D:\tango\voc_tango\lib\site-packages\sklearn\__init__.py", line 82, in <module>
    from .base import clone
  File "D:\tango\voc_tango\lib\site-packages\sklearn\base.py", line 17, in <module>
    from .utils import _IS_32BIT
  File "D:\tango\voc_tango\lib\site-packages\sklearn\utils\__init__.py", line 19, in <module>
    from .murmurhash import murmurhash3_32
  File "sklearn\utils\murmurhash.pyx", line 1, in init sklearn.utils.murmurhash
ValueError: numpy.ufunc size changed, may indicate binary incompatibility. Expected 232 from C header, got 216 from PyObject

Answer 9 · 2023-04-25T06:04:51.000Z

Hmm, I think the numpy version needs to be either upgraded or downgraded to 1.22.0 or 1.20.0

Answer 10 · 2023-04-25T06:17:14.000Z

OK, v 1.22,0 numpy gets past that error.
Now...

  File "D:\tango\tango_generate.py", line 2, in <module>
    from tango import Tango
  File "D:\tango\tango.py", line 5, in <module>
    from models import AudioDiffusion, DDPMScheduler
  File "D:\tango\models.py", line 22, in <module>
    from diffusers.utils import randn_tensor
ImportError: cannot import name 'randn_tensor' from 'diffusers.utils' (unknown location)

Answer 11 · 2023-04-25T06:27:21.000Z

Did you install diffusers from the directory provided in this repo as specified in the ReadMe?

Otherwise, you might need to add tango/diffusers/src/ to your path. randn_tensor is defined in tango/diffusers/src/diffusers/utils

Answer 12 · 2023-04-25T06:54:43.000Z

OK,

cd diffusers
pip install -e .

got past that but then when running the text script it complains

    raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled

Answer 13 · 2023-04-25T06:59:37.000Z

You can use tango = Tango("declare-lab/tango", "cpu")

Answer 14 · 2023-04-25T07:06:37.000Z

One step closer, then...

  fft_window = pad_center(fft_window, filter_length)
Traceback (most recent call last):
  File "D:\tango\tango_generate.py", line 3, in <module>
    tango = Tango("declare-lab/tango", "cpu")
  File "D:\tango\tango.py", line 19, in __init__
    self.stft = TacotronSTFT(**stft_config).to(device)
  File "D:\tango\audioldm\audio\stft.py", line 150, in __init__
    self.stft_fn = STFT(filter_length, hop_length, win_length)
  File "D:\tango\audioldm\audio\stft.py", line 43, in __init__
    fft_window = torch.from_numpy(fft_window).float()
RuntimeError: Numpy is not available

pip list shows numpy v1.22.0

Answer 15 · 2023-04-25T07:15:57.000Z

Can you try with numpy 1.23.0? Also you might want to use less number of steps in tango.generate() as it will take a long time to generate on CPU. The default number of steps is 100.

Answer 16 · 2023-04-26T01:50:57.000Z

numpy 1.23.0 and
tango = Tango("declare-lab/tango", "cpu")
gives this

  File "D:\tango\voc_tango\lib\site-packages\torch\serialization.py", line 166, in validate_cuda_device
    raise RuntimeError('Attempting to deserialize object on a CUDA '
RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU.

Removing the cpu directive gets it back to

  File "D:\tango\voc_tango\lib\site-packages\torch\cuda\__init__.py", line 221, in _lazy_init
    raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled

Getting torch back to a GPU version...

pip uninstall -y torch
pip install --no-cache-dir --ignore-installed --force-reinstall --no-warn-conflicts torch==1.13.1+cu116 torchvision==0.14.1+cu116 torchaudio==0.13.1+cu116 --extra-index-url https://download.pytorch.org/whl/cu116

then gets

    from numba.np.ufunc import _internal
SystemError: initialization of _internal failed without raising an exception

getting numpy back to v1.23.0

pip uninstall -y numpy
pip install --no-cache-dir --ignore-installed --force-reinstall --no-warn-conflicts numpy==1.23.0

and IT WORKS!!! So, yes, numpy v 1.23.0 in requirements will help others.
Not sure if you want to specify the GPU versions of torch?

Answer 17 · 2023-04-26T02:27:57.000Z

We used pytorch version 1.13.1+cu117. I have updated the ReadMe to include this.

Glad that the issue got resolved! Will close this now.