Error: Model filename not in model list

[dv0zn3r]

Unexpected behaviour: For models ggml-mpt-7b-base, ggml-mpt-7b-instruct, ggml-gpt4all-j-v1, after the first prompt the model is not downloaded, instead an error occurs Error: Model filename not in model list: <model name>.

[ryanfb]

With Python gpt4all 1.0.8, I currently get the following models working out of the box with llm + llm-gpt4all:

• GPT4All-13B-snoozy
• nous-hermes-13b
• wizardLM-13B-Uncensored
• orca-mini-3b
• orca-mini-7b
• ggml-model-gpt4all-falcon-q4_0
• ggml-replit-code-v1-3b
• orca-mini-13b

The following models give me the "Model filename not in model list" error out of the box:

• ggml-all-MiniLM-L6-v2-f16
• ggml-gpt4all-j-v1
• ggml-vicuna-7b-1
• ggml-wizardLM-7B
• ggml-mpt-7b-base
• ggml-mpt-7b-instruct
• ggml-mpt-7b-chat
• wizardlm-13b-v1
• starcoderbase-3b-ggml
• ggml-vicuna-13b-1
• ggml-nous-gpt4-vicuna-13b
• ggml-stable-vicuna-13B

However, I've found a workaround that works for some (all?). If you drop into a Python shell and run e.g.:

from gpt4all import GPT4All
model = GPT4All("ggml-vicuna-7b-1.1-q4_2.bin")

Using the filename from the exception, subsequent runs of llm seem be able to use the cached model.

The Python source for the gpt4all package is available here: https://github.com/nomic-ai/gpt4all/tree/main/gpt4all-bindings/python