NameError: name 'torch' is not defined

[sdugoten]

I am trying to convert using the following command

ct2-transformers-converter --model g:\whisper-large-v2-cantonese\pytorch_model.bin --output_dir fast_canton --copy_files tokenizer_config.json --quantization float16

from this model https://huggingface.co/simonl0909/whisper-large-v2-cantonese

I got error like this

PS G:> ct2-transformers-converter --model g:\whisper-large-v2-cantonese\pytorch_model.bin --output_dir fast_canton --copy_files tokenizer_config.json --quantization float16
Traceback (most recent call last):
File "D:\Program Files\Python\Phthon3.10.6\lib\runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "D:\Program Files\Python\Phthon3.10.6\lib\runpy.py", line 86, in run_code
exec(code, run_globals)
File "D:\Program Files\Python\Phthon3.10.6\Scripts\ct2-transformers-converter.exe_main.py", line 7, in
File "D:\Program Files\Python\Phthon3.10.6\lib\site-packages\ctranslate2\converters\transformers.py", line 1771, in main
converter.convert_from_args(args)
File "D:\Program Files\Python\Phthon3.10.6\lib\site-packages\ctranslate2\converters\converter.py", line 50, in convert_from_args
return self.convert(
File "D:\Program Files\Python\Phthon3.10.6\lib\site-packages\ctranslate2\converters\converter.py", line 89, in convert
model_spec = self._load()
File "D:\Program Files\Python\Phthon3.10.6\lib\site-packages\ctranslate2\converters\transformers.py", line 97, in _load
with torch.no_grad():
NameError: name 'torch' is not defined

[winstxnhdw]

Have you tried

ct2-transformers-converter --model simonl0909/whisper-large-v2-cantonese --output_dir fast_canton --copy_files tokenizer_config.json --quantization float16

[sdugoten]

Have you tried

ct2-transformers-converter --model simonl0909/whisper-large-v2-cantonese --output_dir fast_canton --copy_files tokenizer_config.json --quantization float16

G:>ct2-transformers-converter --model simonl0909/whisper-large-v2-cantonese --output_dir fast_canton --copy_files tokenizer_config.json --quantization float16

Traceback (most recent call last):
File "D:\Program Files\Python\Phthon3.10.6\lib\runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "D:\Program Files\Python\Phthon3.10.6\lib\runpy.py", line 86, in run_code
exec(code, run_globals)
File "D:\Program Files\Python\Phthon3.10.6\Scripts\ct2-transformers-converter.exe_main.py", line 7, in
File "D:\Program Files\Python\Phthon3.10.6\lib\site-packages\ctranslate2\converters\transformers.py", line 1771, in main
converter.convert_from_args(args)
File "D:\Program Files\Python\Phthon3.10.6\lib\site-packages\ctranslate2\converters\converter.py", line 50, in convert_from_args
return self.convert(
File "D:\Program Files\Python\Phthon3.10.6\lib\site-packages\ctranslate2\converters\converter.py", line 89, in convert
model_spec = self._load()
File "D:\Program Files\Python\Phthon3.10.6\lib\site-packages\ctranslate2\converters\transformers.py", line 97, in _load
with torch.no_grad():
NameError: name 'torch' is not defined

[winstxnhdw]

Can you do

pip install torch

[sdugoten]

Can you do

pip install torch

It's already installed, the version that works with CUDA

[sdugoten]

Can you do

pip install torch

Just for your information, I try it one more time on my Mac instead of the Windows that you saw above, same result

temp3 % ct2-transformers-converter --model simonl0909/whisper-large-v2-cantonese --output_dir fast_canton --copy_files tokenizer_config.json --quantization float16

Traceback (most recent call last):
File "/opt/homebrew/bin//ct2-transformers-converter", line 8, in
sys.exit(main())
File "/opt/homebrew/lib/python3.10/site-packages/ctranslate2/converters/transformers.py", line 1771, in main
converter.convert_from_args(args)
File "/opt/homebrew/lib/python3.10/site-packages/ctranslate2/converters/converter.py", line 50, in convert_from_args
return self.convert(
File "/opt/homebrew/lib/python3.10/site-packages/ctranslate2/converters/converter.py", line 89, in convert
model_spec = self._load()
File "/opt/homebrew/lib/python3.10/site-packages/ctranslate2/converters/transformers.py", line 97, in _load
with torch.no_grad():
NameError: name 'torch' is not defined

goten@Gotens-Mac-Studio-116 temp3 %
goten@Gotens-Mac-Studio-116 temp3 %
goten@Gotens-Mac-Studio-116 temp3 % pip3 list
Package Version

---

aiohttp 3.8.4
aiosignal 1.3.1
async-timeout 4.0.2
attrs 23.1.0
bbook-maker 0.3.0
beautifulsoup4 4.12.2
bs4 0.0.1
certifi 2022.12.7
charset-normalizer 3.1.0
ctranslate2 3.20.0
EbookLib 0.18
filelock 3.12.4
frozenlist 1.3.3
idna 3.4
Jinja2 3.1.2
lxml 4.9.2
markdown-it-py 2.2.0
MarkupSafe 2.1.3
mdurl 0.1.2
mpmath 1.3.0
multidict 6.0.4
networkx 3.1
numpy 1.26.0
openai 0.27.5
pip 23.2.1
protobuf 4.21.12
Pygments 2.15.1
PyYAML 6.0.1
regex 2023.3.23
requests 2.29.0
rich 13.3.5
setuptools 67.2.0
six 1.16.0
soupsieve 2.4.1
sympy 1.12
tiktoken 0.3.3
torch 2.0.1
tqdm 4.65.0
typing_extensions 4.8.0
urllib3 1.26.15
wheel 0.38.4
yarl 1.9.2

[phineas-pta]

pip install transformers huggingface-hub

[winstxnhdw]

Please use a proper Python virtual environment like mamba or venv next time to avoid this unnecessary pain.

[sdugoten]

pip install transformers huggingface-hub

thanks for help, I eventually use my Mac to convert the file into .bin that can be used by faster-whisper. I wonder if something can be converted that is used by Whisper OpenAI that takes .py file?

Sorry keep asking silly question as I am no programmer.

[winstxnhdw]

No, it's not possible. .bin and .py files are not close to being equivalents.

[Anthonyg5005]

a bit late but make sure you use the correct version of pytorch and that transformers and huggingface-hub are installed. if it still doesn't work add import torch at the beginning of transformers.py
there's a chance you might also need to import transformers and huggingface-hub as well. I had to do this once on a linux vm but not on my windows computer.
after checking again all I did was make sure those were all installed correctly with correct versions from the pytorch website then replaced this:
try:
import huggingface_hub
import torch
import transformers
except ImportError:
pass

with this:
import huggingface_hub
import torch
import transformers