Failing to download models

[DMTarmey]

Hi i just installed the windows installation application and trying to download a model, but it just doesn't seem to finish any download.

[subversion23]

same here. got the error:

ERROR: Download error MD5SUM did not match: "65023f2b4ad4f10d75a0c1c019cdea6f" != "6cb4ee297537c9133bddab9692879de0" for "ggml-stable-vicuna-13B.q4_2.bin"

tried other model. same.

[Korayem]

Same on mac v2.3.0

I wait for download to finish then poof! nothing is in the folder. I just wasted tons of bandwidth with my ISP.

[Woisek]

Is there a list of download links to all the models, so we can manually download?

[ansonhorse]

@Woisek

Manually download:

1. It is recommended to install JSON Formatter for a better viewing experience. This step is optional.

2. Open https://gpt4all.io/models/models.json. Once opened, you will see all the models.

3. Taking the model mpt-7b-chat as an example, you can find its corresponding filename on the page you just opened. The content of filename is ggml-mpt-7b-chat.bin. Copy this filename.

4. Concatenate the download link. The format is https://gpt4all.io/models/ + filename. Using the example model above, the resulting link would be https://gpt4all.io/models/ggml-mpt-7b-chat.bin.

5. Use an appropriate download tool (a browser can also be used) to download the obtained link.

6. Move the downloaded ggml-mpt-7b-chat.bin file to the model storage directory of GPT4All. You can find the directory path in the model download dialog of GPT4All. For example, on macOS, it would be /Users/someone/Library/Application Support/nomic.ai/GTP4All.

7. Quit and reopen GPT4All.

---

Additionally, it is recommended to verify whether the file is downloaded completely. Use any tool capable of calculating the MD5 checksum of a file to calculate the MD5 checksum of the ggml-mpt-7b-chat.bin file. Compare this checksum with the md5sum listed on the models.json page. If they do not match, it indicates that the file is incomplete, which may result in the model not functioning properly.

[Woisek]

Perfect! Works like a charm! Thanks! 😊

[sunil-dhaka]

Works!

[StillMelo]

thank you so much

[johntsunami]

I concatenated the URL to https://gpt4all.io/models/ggml-model-gpt4all-falcon-q4_0.bin and i tried other models like this one https://gpt4all.io/models/nous-hermes-13b.ggmlv3.q4_0.bin , but i'm still getting this error. Do i need to download all of the models individually or do i just use the link you made "https://gpt4all.io/models/ggml-mpt-7b-chat.bin" to use all of them?

Error 404 This object could not be viewed You are not authorized to view this object

This object does not exist or is not publicly accessible at this URL. Check the URL of the object that you're looking for or contact the owner to enable Public access.

Is this your bucket?

Learn how to enable Public Access

Also the model i downloaded of "https://gpt4all.io/models/ggml-mpt-7b-chat.bin" responses are cut off and incomplete from the chatgpt responses.. so i wouldn't recommend anyone using this one.

FYI chatgpt from openai is down at the moment.. maybe thats causing my current issue?

[johntsunami]

Found the fix.. Just download the bin file from the url from the Json file. Like this one. "url": "https://huggingface.co/TheBloke/Nous-Hermes-13B-GGML/resolve/main/nous-hermes-13b.ggmlv3.q4_0.bin",

[3204853406]

Great, thank you very much for your reply.

---Original--- From: "John @.> Date: Tue, Jul 11, 2023 23:07 PM To: @.>; Cc: @.@.>; Subject: Re: [nomic-ai/gpt4all] Failing to download models (Issue #477)

Found the fix.. Just download the bin file from the url from the Json file. Like this one. "url": "https://huggingface.co/TheBloke/Nous-Hermes-13B-GGML/resolve/main/nous-hermes-13b.ggmlv3.q4_0.bin",

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.***>

[jancborchardt]

@johntsunami I had that issue with the Falcon model as well – here is another source for downloading it: https://huggingface.co/nomic-ai/gpt4all-falcon-ggml/tree/main From there you need to follow step 6 and 7 as described by @ansonhorse above at https://github.com/nomic-ai/gpt4all/issues/477#issuecomment-1553945164 :)

EDIT: Nevermind my comment – had an outdated version of this issue open for some reason. Good that you found a fix too!

[nav9]

If it helps anyone, as mentioned here, I could solve it by using Turbo Download Manager Firefox extension.

[filippopossenti]

I've been experiencing the same problem. I downloaded Gpt4All today, tried to use its interface to download several models. They all failed at the very end. Sometimes they mentioned errors in the hash, sometimes they didn't.

Seems to me there's some problem either in Gpt4All or in the API that provides the models. I would suggest adding an override to avoid evaluating the checksum, at least until the underlying issue is solved.

[WGrav01]

After putting the downloaded .bin file for GPT4All Falcon into /Users/namehere/Library/Application Support/nomic.ai/GPT4All/ (on macOS 12.6.6) and then relaunching GPT4All, it fails to detect the model there and simply doesn't work. Any ideas?

[cebtenzzre]

After putting the downloaded .bin file

Assuming you are using GPT4All v2.5.0+, you need to download a .gguf file. GGML (.bin) files are no longer supported.

[WGrav01]

After putting the downloaded .bin file

Assuming you are using GPT4All v2.5.0+, you need to download a .gguf file. GGML (.bin) files are no longer supported.

That makes sense, (I am using v3.5.2) but the json file above doesn't have any .gguf files. Where do I get those?

[cebtenzzre]

the json file above doesn't have any .gguf files. Where do I get those?

http://gpt4all.io/models/models2.json