How to add models into daisi folder

I am unable to download and place the models into their desired folders even if i make python scripts to download and place them into folders. please help. I need to add these models for stable diffusion project.

can you tell us what problems or messages you are seeing when you do this, and provide us with some of the code that you are using to perform the file transfers?

Note that if files are excessively large they may not be able to be uploaded. In those cases they can be placed on another storage that you provide (e.g. cloud stores) and accessed from your Daisi using the new secrets to pass access tokens.

@kali : if you are using HuggingFace transformers or diffusers library, it will download the weights automatically and place them into the cache, similarly to what it would do locally. You can check the code of this Daisi as an example:

Note that in this case, the very first execution will be slow, since it will be downloading the weights. Subsequents executions will be much faster, as it will access directly the weights from the cache

For sure is the app is the github repo it’s perfectly cloned all requirements are installed . But after I open app section it must wait for the first time to setup all files but instead it gives me error like

where the script itself is capable of downloading and placing this file into desired folder.

I got it and have checked this too but there are other models too simultaneously to be downloaded from github to enhance our outputs and in this way we don’t require any further authentication like to get added into a secret key.

Hi @kali : we have found the issue. We are looking at it and will get back to you by tomorrow.

Thank you so much, please rectify the issue and let me know if possible about the issue. You are going to love this project after its resolution literally .

1 Like

Also there is another issue I am facing which I did faced with my previous account is

suppose i want to delete my daisy which is no more useful but I am getting this kind of issue.

Hi @kali

First, please know that we are working on UI features to make it easy to add external files to the container of your Daisi.

The path to store the models to be downloaded was actually a relative path. Changing it to an absolute path solved the issue.

Practically, the attached script contains a couple of functions to automatically download the models and store them in '~/stable_diffusion_models'. It checks if the models are already present, so it won’t re download them at each execution. You need to copy/paste these functions into the main script of your Daisi, and call download_models() in the main function of your script.

There is also a simple utility function in this script that you can call with pydaisi to check that the models are actually present in the container:

import pydaisi as pyd
my_daisi = Daisi("username/daisiname") # here "username is your username and daisiname the name of the Daisi in which the models have been added

my_daisi.exec_cmd(prompt="ls -altrh ~/stable_diffusion_models").value # to list content of ~/stable_diffusion_models for instance

As a side note: given what you want to achieve, I would recommend to actually deploy each one of these model independently as a Daisi, and having one wrapper Daisi calling all of them with pydaisi, with a micro services pattern that Daisi makes easy to achieve.
Check for instance the Stable Diffusion App Daisi which calls this other Daisi. The latter is the deployment of the model, and the former is simply calling it as a service with pydaisi

This will make debugging any possible issue much easier.

Also, to deploy these models on GPUs with torch, make sure to make the GPUs visible in your code:

import os
os.environ["CUDA_VISIBLE_DEVICES"] = "0,1"

and to update the requirements.txt file :

torch==1.12.0+cu116 (1.2 KB)

Ok . Actually I did added the script before deploying the project in the project folder and imported it into the main script but even then I got this message shown while running even I tested this on my pc. Now do we have resolved the issue to get it worked in desired way?

I have been able to deploy these models with the script that I provided… I will check your repo and investigate why it doesn’t work in this context. I will let you know soon

ok I’ll be waiting for you.

One more thing I have observed there is another daisy I made with the name SDIFF which I tried multiple times to get deleted but it’s not getting. Can you see the issue with it?