Sigbjørn Skjæret

They've removed the alternate `(?i:'s|'t|'re|'ve|'m|'ll|'d)` regex from the Llama pre-tokenizer, meaning you (or someone else) will have to update and run `convert_hf_to_gguf_update.py` to generate the `chkhsh` entry and submit a...

> Hi. Pleias here: we plan to release a new version of the model for the model and can bundle a modified tokenizer to ease conversion. It would be preferable...

@pockers21 What's up?

@pockers21 You need to address the tensor mappings, as pointed out by @ngxson, use `tensor_mapping.py` where possible.

Hmmm, there's a major issue with conversion, the `text_config` is normally applied on top of the remote `jina-embeddings-v3` `config.json` by `transformers`, however `convert_hf_to_gguf.py` has no concept of this when reading...

TBH, I'm not sure this is acceptable, I would expect to be able to convert the original model, granted it's a little tricky due to the way it's constructed, but...

See #1503, although it was done mainly for built-in chat templates it should also work for the `chatml-function-calling` format, albeit currently only when selecting the tool specifically, not with `auto`,...

> It is better to keep all special tags visible for testing/debugging purposes. Keeping the tags would be hard, I don't think it's much of an issue as long as...

The guard against stripped reasoning is very nice, prevents crashes with several templates! However something is not quite right, f.ex. with `Qwen3-4B-Thinking-2507` the following happens on the second prompt (after...

> Interesting. Did you type an empty message as second input? Any chance you would provide the full command history/conversation transcript? I will work on debugging that model. I merely...