Embdedings & Model files Location in local system.

[SelvakumarTS]

Hi.. Thanks for the model. It works good in my personal computer. In my office system, i have an issue connecting to huggingface.co via the code. It gives some SSL certificate error. Tried to fix it with some options available, but no success. I am just wondering, can you let me know here the models are saved while i run it in my personal systems , so that i can transfer those embedding and model files to my office laptop and use it directly instead of trying to connect to huggingface. Basically i need to load the embeddings from a local repo.

Instead of below code, is there any alternate to load the Embedings directly from local system
embeddings = HuggingFaceInstructEmbeddings(model_name=EMBEDDING_MODEL_NAME, model_kwargs={"device": device_type})

and any other code changes in run_localGPT.py

[SaiAkhil066]

I hope your having a proper internet connection, beacuse when you run (python run_locagpt.py), at first, embedding model (hkunlp/instructor-large) which is selected as default model, will be started downloading, after that you, the LLM will be downloaded, so after this process, you dont need any of the internet connections.
I hope this helps you, Reply if any problem occurs.

[lavericklavericklaverick]

...after that you, the LLM will be downloaded, so after this process, you dont need any of the internet connections. I hope this helps you, Reply if any problem occurs.

i think this is incorrect somehow... i have downloaded the model and run it fine using an open network. however when i reconnect to the secure network (which doesnt like pip downloads etc) im getting SSL certificate issues. i dont know why this would happen unless its trying to access the internet.

[SelvakumarTS]

The SSL certificate issue got fixed for me by installing the pip install python-certifi-win32. Now i am able to run it and download.

Before finding the above solution, i used to run the code in Google Colab. This will download the files and store it in an certain folder format for both embeddings and Model. You can download the files to your local repository in the same folder format and update the code where it loads the embedding from your local path and also the code where it loads the model with your local path to the folder. This way it worked for me. In Constant.py file you can update the embedding and model path like shown below

EMBEDDING_MODEL_NAME = "C:/Users/Tokenizer/Embedding" # Inside the Embedding folder you will have downloaded embedding files

MODELS_PATH = "./models" # Your local path where the models are downloaded and stored. The folder structure of model will contain subfolders like ( blobs, ref, etc.)

[lavericklavericklaverick]

thanks! that worked perfectly!