Open NicoDaimler opened 2 months ago
Check available space on the drive. Check your internet connection.
We may need to add detection for when the download speed drops to zero for too long (which probably isn't reflected in the UI) and interrupt/continue. You should be able to trigger this manually by clicking cancel and then resume. I would blame it on a poor internet connection, but I know our CDN is flaky at times (though I haven't had problems lately).
I had this happen to me today for Hermes. I closed the client, restarted, and the button showed RESUME and continued where it left off
Every time I restart the client and resume the download, it automatically pauses after a few seconds. After spending several hours to complete the download, I still can't load the model; the client crashes.
Every time I restart the client and resume the download, it automatically pauses after a few seconds. After spending several hours to complete the download, I still can't load the model; the client crashes.
What is your download rate?
I also experience this issue with the larger models. And my download rate is a pitiful 25Mbps. So I assumed it's an issue with slow Internet and larger model files. ("Larger" model for me means >8GB filesize)
It is not a filesystem full issue. Though that is good to check.
A workaround for now: download the model directly from Huggingface, drop it into the GPT4All folder/dir, and configure the prompt based on the Huggingface model card.
Sometimes the issue is not GPT4All's downloader. I've had Hugginface or my Internet cause direct download hiccups.
It's working fine in my case, check your internet connection and also check if the storage is available in your system or not.
Bug Report
When downloading different language models, it only downloads one piece and then stops and does not continue.
Then you can wait hours and nothing happens.
Your Environment