Failed Model Downloads

[zenkodr]

Out of the Blue today every model download fails. I have checked network connections, firewall and antivirus settings nothing showing that it is blocking downloads. the Only way I can get new models is to copy the model address in the model downloads and manually download through browser.

[yagil]

@zenkodr in case you are on Windows, you can try this or look into something related to this: https://answers.microsoft.com/en-us/windows/forum/all/random-connection-reset-error/6ca8afce-f190-4b3d-955f-2dcdda6a7e79?page=3

[wladmirfb]

@yagil Thank you. Happend the same to me on Windows 11 Dell PC just a matter of a killer process stopped on Task Manager.

[Kaleemullahqasim]

Any update on this issue I can't download or search any model at all. It says, "Download failed. Time-out. Please try to resume".

[yagil]

@Kaleemullahqasim did it work and then stop working, or did it never work for you?

[Kaleemullahqasim]

@yagil, it never worked for me; I also tried using another laptop. The model search function also never worked. ollama is running fine

[Kaleemullahqasim]

Capacity is not an issue running on an m3 max chip with 64GB. It is more like a network connection. Search functionality should work!

[yagil]

@Kaleemullahqasim Are you on our Discord? I can share with you a quick beta build for macOS that might solve this. Please tag me in the help channel. Discord link: https://discord.gg/aPQfnNkxGC

[angleunion]

Happend the same to me on Windows 11

Happend the same to me on Windows 11

[odoob2t]

If you are on a Dell computer, try to disable completely the Killer Intelligent Center. It worked for me.

[Sojou]

Any tips on fixing failed downloads in the Linux version? Thanks.

[AwesomeDong]

i have the same issue in linux, thanks

[Sojou]

i have the same issue in linux, thanks

I found out how to do a workaround:

1. Download the gguf file(s) for the particular model you want.
2. With the file manager of your choice, go to /home/whateverYourUSERnameIs/.cache/lm-studio/models (you may have enable 'view hidden files' in your file manager to see the .cache directory) .
3. In the models directory, create a folder called "Publisher" .
4. Open the Pubilsher directory. This is where you do the work from.
5. Copy or download all your downloaded gguf models into this Publisher directory.
6. In the publisher directory you now rightclick on a gguf file and look at 'properties' and copy the name of it (unless you are good at typing....i'm not). then close properties pop-up.
7. Rightclick in the empty space of your file manager and create a new folder and paste the name of the gguf file you copied from the 'properties' popup BUT add a letter or a number to the end of the name and save it. If you try to make a folder with the exact name you copied, it will tell you there is already something there with that name. Linux treats names of folders and names of files with equality so it will tell you there is already something in the folder with that name so you have to alter the folder name temporarily.
8. Once the folder is made, drag that particular gguf file over and put it in that folder.
9. Now, rightclick on that folder and you can now change the name and take away the extra number or digit you had to add to the name and rename it to the very exact name of the gguf file that is now in that folder.
10. Rinse and repeat with all your downloaded gguf files. Your path to each gguf file will now be /home/whateverYourUSERnameIs/.cache/lm-studio/models/Publisher/TheFolderWithTheNameOfTheGGUFfile/TheNameOfTheGGUFfile.gguf When you fire up LM-Studio, each model will now be selectable.
11. Now file up LM-Studio and click the AI-chat icon on the side bar under the search icon. It looks like a cloud with three dots.
12. There is a purple bar at the top now which you can select which model to load. All your models will be there.
13. Have fun. LLMs are sooo cool! It not quick to set up but it works just fine.

[Spark-Plug]

same here, win11 pc, I can't download or search any model at all.

[MACSjRichter]

Same here ... What worked for me was disconnect from current Network and connecting to my Mobile Hotspot.

[nagarjunr]

I did encouter the same issue. I tried to run lm studio in Macbook M3 Pro. Is it possible to expose the network settings in the UI, so that we can set it manually including proxy configuration to resolve this issue?

[egfthomas]

For Linux, just disable IPv6 on your network adapter and voila

[abussl]

On my Acer Predator / Win 11, i disabled the Priorization Engine and the Auto Bandwith in the "Killer Intelligence Center", and then it worked.

[pierrekhouryy]

On my Acer Predator / Win 11, i disabled the Priorization Engine and the Auto Bandwith in the "Killer Intelligence Center", and then it worked.

this also fixed it for me, acer nitor/win11

[Sojou]

" For Linux, just disable IPv6 on your network adapter and voila" I found this working once I did that AND rebooted. Thank you. At first, I tried disabling the IPV6 on the VPN. That didn't work. Then I disabled it on the adapter and it didn't work....at first. Once I rebooted I'm now getting downloads. Thank you.

[Valentinouri]

On my Acer Predator / Win 11, i disabled the Priorization Engine and the Auto Bandwith in the "Killer Intelligence Center", and then it worked.

this also fixed it for me, acer nitor/win11

Same here, acer nitro 5

[pengjiang80]

Any suggestion for MAC OS(M1 Pro)? I tried disabling IPv6 and restart but doesn't help. Model search doesn't work either.

Btw: Ollama works and I also tried downloading with huggingface-cli without problem. So network itself shouldn't be a problem.