mradermacher

/

model_requests

like 100

We tried that model in the past and it failed due to its BPE pre-tokenizer not being supported by llama.cpp. Its pre-tokenizer hash is b0f33aec525001c9de427a8f9958d1c8a3956f476bec64403680521281c032e2. If you are certain a pre-tokenizer supported by llama.cpp is fully compatible with this model we could quant it using a different pre-tokenizer than the one specified by the model author.