David-AU-github
David-AU-github
> Thanks for sharing your results here! > > DARE-TIES does have a randomized element, yeah - it's part of the algorithm by design. If you want more reproducible merges...
thank you !!!
Quick question: Same format with DARE TIES? other merge types?
> it appears they slightly changed the names of some things there, not an expert but will try to see if i can find any differences Seems the issue may...
@Jiahao-Yuan @cg123 Built a Llama 3.2 - 8X3B (18.4B parameters) this way: Found if you patched mergekit\_data\architectures\mistral.json as follows; it works: ``` "post_weights": [ { "name": "model.norm.weight", "input_space": "h_${num_layers}" },...
Added issue: seems even when using "mergekit" work around ; that merge kit is not creating "tokenizer.model" for Gemma models. (previously it did). RESULT: Can't quant models from source without...
Might be in "pedantic" python package (?) ; I had to revert to an older Mergekit version.
Confirming exact same error ; mergekit can not find the "base_model" ; including if the path is local (absolute) on windows. Funny thing is some mergekits work fine - no...
@cg123 Thank you so much.;
> @ubergarm, just finished uploading [Qwen3-30B-A3B-GGUF](https://huggingface.co/eaddario/Qwen3-30B-A3B-GGUF). Summary of scores in the model card, and actual results in the _scores_ folder. > > A few things to consider: > > *...