Issue: <Invalid format for ggml-gpt4all-j-v13-groovy.bin>

[Vik1922]

Issue you'd like to raise.

After updating gpt4all from ver 2.4.12 to 2.4.17, was not able to load the "ggml-gpt4all-j-v13-groovy.bin" model.

got the error: Could not load model due to invalid format for ggml-gpt4all-j-v13-groovybin Need a solution or workaround please!

also:

1. are there any other models apart from groovy.bin that can access & chat interactively with local docs?
2. any models dedicated to science/medical data

thanks a lot in adv!

Suggestion:

No response

[WhatsUpInTheSky37]

Same here all of my .bin files are not working. I am on a m2 with 64 gb ram

[cosmic-snow]

Note that model files in an older GGML format are no longer supported with the recent updates. You need them to be in GGMLv3 format. The models in the download list are all in that format now. They might have similar names but are not the same.

Additionally, for Vulkan GPU support you need enough VRAM for the whole model + some more for the generated text. (This does not apply to Mac M1/M2 as it is using Metal, not Vulkan.)

With all that in mind, can you confirm that you have a GGMLv3 model, preferably with q4_0 quantisation, which does (or doesn't) work?

[Vik1922]

Note that model files in an older GGML format are no longer supported with the recent updates. You need them to be in GGMLv3 format. The models in the download list are all in that format now. They might have similar names but are not the same.

Additionally, for Vulkan GPU support you need enough VRAM for the whole model + some more for the generated text. (This does not apply to Mac M1/M2 as it is using Metal, not Vulkan.)

With all that in mind, can you confirm that you have a GGMLv3 model, preferably with q4_0 quantisation, which does (or doesn't) work?

reg the dload model: xxd not working in my win11 os. tried to install it but it wouldnt. any other way to find out the version.

---

i dloaded the model file in aug23. the version in the pic above says that it is 4 months old-dont know if you are referring to this one or another file at another site. is the above file the correct version that you mentioned. if not pl share the link of the correct version (ggml-gpt4all-j-v13-groovy.bin) reg the gpu support- mine is nvidia-rtx- no issue with gpu or vram. also could you pl suggest: are there any other models apart from groovy.bin that can access & chat interactively with local docs? any models dedicated to science/medical data

thanks again cos snow!

[cosmic-snow]

reg the dload model: xxd not working in my win11 os. tried to install it but it wouldnt. any other way to find out the version.

---

i dloaded the model file in aug23. the version in the pic above says that it is 4 months old-dont know if you are referring to this one or another file at another site. is the above file the correct version that you mentioned. if not pl share the link of the correct version (ggml-gpt4all-j-v13-groovy.bin)

I think that groovy model simply doesn't have a newer version. Recommended right now are:

• Falcon from the download list if you have trouble with a bug and your GPU, because it doesn't yet support GPU. So using that as default should help against bugs.
• Orca Mini (Small) to test GPU support because with 3B it's the smallest model available.

reg the gpu support- mine is nvidia-rtx- no issue with gpu or vram.

Then just use the models that are still supported (GGMLv3).

also could you pl suggest: are there any other models apart from groovy.bin that can access & chat interactively with local docs? any models dedicated to science/medical data

thanks again cos snow!

Any of the models should be able to be used with LocalDocs. What it does is basically it looks for keywords in your prompt and if it has matches, it inserts text chunks from the indexed documents as text to send to the model. It's all just text, no model specifics required.

[Vik1922]

yes i dloaded the most updated version of groovy yet it doesnt work. but snoozy with 13b works well enough. no complaints there! thanks again!

[Vik1922]

hi, the gpt4all did an update today and after that i am unable to install any of the models- it completes the dload, checks md5 then it stops and shows the screen prior to the dload. so what is happening guys? any solutions? thanks in advance!

[cosmic-snow]

@Vik1922 Best I can tell you there is to try again, it looks like a connection/download problem.

If your connection is problematic, you might want to try again with your browser or a separate download program. They might be better equipped to handle that. See https://gpt4all.io/ and scroll to Model Explorer or download from a site like Hugging Face.

Also, you should probably not reply to an existing, closed issue with a separate problem. There's no guarantee someone will look at these.

[cebtenzzre]

hi, the gpt4all did an update today and after that i am unable to install any of the models- it completes the dload, checks md5 then it stops and shows the screen prior to the dload. so what is happening guys? any solutions? thanks in advance!

See #1377 and #1413

[Vik1922]

sorry i should have asked in a new thread-will do next time!🙏 i have dloaded the models from H-face directly and copied them to the gpt4all main directory. still not resolved. uninstalled it twice and reinstalled it with its default settings-still not working! earlier it worked well with many models! once the update was done it got screwed!

On Wed, Oct 25, 2023 at 2:52 AM cosmic-snow @.***> wrote:

@Vik1922 https://github.com/Vik1922 Best I can tell you there is to try again, it looks like a connection/download problem.

If your connection is problematic, you might want to try again with your browser or a separate download program. They might be better equipped to handle that. See https://gpt4all.io/ and scroll to Model Explorer or download from a site like Hugging Face https://huggingface.co/.

Also, you should probably not reply to an existing, closed issue with a separate problem. There's no guarantee someone will look at these.

— Reply to this email directly, view it on GitHub https://github.com/nomic-ai/gpt4all/issues/1428#issuecomment-1778060589, or unsubscribe https://github.com/notifications/unsubscribe-auth/A3E5RZI2323G4RASM7COZM3YBAWQPAVCNFSM6AAAAAA4ZPQFLWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONZYGA3DANJYHE . You are receiving this because you were mentioned.Message ID: @.***>

[Zibri]

I still can't run it

[cebtenzzre]

I still can't run it

Please open a new issue with more information. This issue is not relevant to current versions of GPT4All, which use a different file format.