Bug: train-text-from-scratch

[artiomborovinskii]

What happened?

When I try ./bin/train-text-from-scratch --vocab-model ../models/ggml-vocab-phi-3.gguf --checkpoint-in FN I have normal start but they it write only"----------------------------------------------------------------------------" to terminal

Name and Version

./bin/main --version version: 3044 (d5c0582) built with clang version 18.1.7 for aarch64-unknown-linux-android24

What operating system are you seeing the problem on?

Linux

Relevant log output

No response

[ngxson]

train-text-from-scratch only works with llama architecture. Also, it's unmaintained for quite a long time, so probably it's already broken.

@slaren do you think we should temporary remove it?

[slaren]

I don't think the training examples are useful enough to justify the maintenance cost, but that's really up to @ggerganov.

[ggerganov]

I've also been thinking that we should probably remove them for now and reimplement at a later point when we are able to maintain them

[ngxson]

I'm closing this as we decided to temporary remove training / fine-tuning example