Impossible to unload OpenAIWhisperParserLocal model

Checked other resources

[X] I added a very descriptive title to this issue.
[X] I searched the LangChain documentation with the integrated search.
[X] I used the GitHub search to find a similar question and didn't find it.
[X] I am sure that this is a bug in LangChain rather than my code.
[X] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).

Example Code

from langchain_community.document_loaders.parsers.audio import OpenAIWhisperParserLocal 

whisper = OpenAIWhisperParserLocal(
    # device='cuda',
    lang_model="openai/whisper-medium.en",
    # batch_size=8,
    # chunk_length=30,
)

# model is now in VRAM

#Try to unload
del whisper

import torch
torch.cuda.empty_cache()
import gc
gc.collect()

# model is still in VRAM

Error Message and Stack Trace (if applicable)

No response

Description

The protocol for unloading the whisper model from memory is detail here: https://github.com/openai/whisper/discussions/1313#discussioncomment-5813140

However, the python lang chain wrapper for whisper doesn't release the model upon deletion of the object, so these steps don't work when using lang chain, making it impossible to unload the model until the python script has shut down. this makes it impossible to build apps since the model cannot be unloaded while the app is still running in order to free up VRAM for other models or other processes. This is a seriously fatal flaw and makes the library a nonstarter for many of its intended use cases.

Is there any mechanism exposed that could unload this model from memory?

System Info

'0.2.12' windows

Hi, @openSourcerer9000. I'm Dosu, and I'm helping the LangChain team manage their backlog. I'm marking this issue as stale.

Issue Summary:

You reported an issue with the OpenAIWhisperParserLocal model not unloading from VRAM in the LangChain library.
Attempts to delete the object and clear the cache have not resolved the issue.
This limitation affects applications requiring dynamic model loading and unloading by not freeing up resources.

Next Steps:

Could you confirm if this issue is still relevant with the latest version of the LangChain repository? If so, please comment to keep the discussion open.
If there is no further activity, this issue will be automatically closed in 7 days.

Thank you for your understanding and contribution!

langchain-ai / langchain