No module named 'nemo.collections'

[Dannynis]

when running "Introduction to End-To-End Automatic Speech Recognition" on google colab I run into:

import nemo
     
import nemo.collections.asr as nemo_asr

ModuleNotFoundError: No module named `'nemo.collections'`

I did all the steps sequentially, and they all succeeded until this block.

[Dannynis]

!pip install git+https://github.com/NVIDIA/NeMo.git
installing using git solved it

[josiewilliams13]

I successfully installed the git by running
!pip install git+https://github.com/NVIDIA/NeMo.git

but it's still not working. any other suggestions?

[kode-git]

I successfully installed the git by running !pip install git+https://github.com/NVIDIA/NeMo.git

but it's still not working. any other suggestions?

Did you installed correctly dependencies?

[titu1994]

That's an incorrect install - look at readme how to install from source. https://github.com/NVIDIA/NeMo#pip-from-source

[muizzhaque]

I have followed the steps but each time i'm installing nemo_toolkit['all'] two packages pynini and mecab-python3 is failing to build wheels, also tried nemo_toolkit['asr'] which was successful but as soon as i run the code, i got the same module not found error
btw, I'am using windows 10

[computersrmyfriends]

That's an incorrect install - look at readme how to install from source. https://github.com/NVIDIA/NeMo#pip-from-source

Despite following the readme, I see this:

Traceback (most recent call last):
File "nvidianemo2/listmodels.py", line 1, in
import nemo.collections.tts as nemo_tts
File "nvidianemo2/nemo.py", line 7, in
from nemo.collections.asr.models import EncDecCTCModel
ModuleNotFoundError: No module named 'nemo.collections'; 'nemo' is not a package

tried without source and with source, exactly following the documentation. it was infact working for me . All of a sudden, it stopped working and does not work on a new environment also. same issue. Not sure if something changed.

[titu1994]

Can you paste your install command ?

[computersrmyfriends]

@titu1994 Thanks for your response.

First, I tried using the pip version etc and finally resorted to this:

conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia
apt-get update && apt-get install -y libsndfile1 ffmpeg
pip install Cython
pip install nemo_toolkit['all']
git clone https://github.com/NVIDIA/NeMo
cd NeMo
./reinstall.sh

Traceback (most recent call last):
  File "projects/nvidianemo2/listmodels.py", line 1, in <module>
    import nemo.collections.tts as nemo_tts
  File "nvidianemo2/nemo.py", line 7, in <module>
    from nemo.collections.asr.models import EncDecCTCModel
ModuleNotFoundError: No module named 'nemo.collections'; 'nemo' is not a package

[computersrmyfriends]

Here's the code inside the listmodels.py:

mport nemo.collections.tts as nemo_tts

# Function to print models with classification
def print_models(model_list, model_type, classification):
    print(f"\nAvailable {model_type} Models ({classification}):")
    for model in model_list:
        print(model.pretrained_model_name)

# TTS Models
print("Listing available TTS and Vocoder models using NVIDIA's NeMo toolkit:")
try:
    print_models(nemo_tts.models.Tacotron2Model.list_available_models(), "Tacotron 2", "TTS")
    print_models(nemo_tts.models.FastPitchModel.list_available_models(), "FastPitch", "TTS")
except Exception as e:
    print(f"Error listing {model_type} models: {e}")

# Vocoder Models
try:
    print_models(nemo_tts.models.WaveGlowModel.list_available_models(), "WaveGlow", "Vocoder")
    print_models(nemo_tts.models.HifiGanModel.list_available_models(), "HiFi-GAN", "Vocoder")
except Exception as e:
    print(f"Error listing {model_type} models: {e}")

# Speaker IDs and Names for FastPitch Multispeaker Model
speaker_ids_and_names = {
    92: "Cori Samuel",
    6097: "Phil Benson",
    9017: "John Van Stan",
    6670: "Mike Pelton",
    6671: "Tony Oliva",
    8051: "Maria Kasper",
    9136: "Helen Taylor",
    11614: "Sylviamb",
    11697: "Celine Major",
    12787: "LikeManyWaters"
}

print("\nSpeaker IDs and Names for FastPitch Multispeaker Model:")
for id, name in speaker_ids_and_names.items():
    print(f"Speaker ID: {id}, Name: {name}")

[titu1994]

The issue is your file name - dont name it Nemo.py, python is trying to resolve Nemo library from your file path (cause your file path gets prepended to Python path automatically in most ides like pycharm or code editors like vs code.

Just renaming the file should work

[computersrmyfriends]

wow, good one. Didn't open my eyes to the stack trace. That did the trick. I didn't realize that!!

Thank you!!!!

[titu1994]

Having faced this issue more times than I care to admit, glad to see the solution is that simple 😄

[duarteocarmo]

The issue is your file name - dont name it Nemo.py, python is trying to resolve Nemo library from your file path (cause your file path gets prepended to Python path automatically in most ides like pycharm or code editors like vs code.

Just renaming the file should work

Gets me everytime.