🀍 Zirel-3 82B-A12B

#2836
by redaihf - opened

https://huggingface.co/Daemontatox/Zirel-3

This is a large model so please queue responsibly.

We tried this model last October and it failed with the following error. It seems highly unlikely that anything changed since then:

Zirel-3 Traceback (most recent call last):
Zirel-3   File "/llmjob/llama.cpp/convert_hf_to_gguf.py", line 9554, in <module>
Zirel-3     main()
Zirel-3   File "/llmjob/llama.cpp/convert_hf_to_gguf.py", line 9548, in main
Zirel-3     model_instance.write()
Zirel-3   File "/llmjob/llama.cpp/convert_hf_to_gguf.py", line 442, in write
Zirel-3     self.prepare_tensors()
Zirel-3   File "/llmjob/llama.cpp/convert_hf_to_gguf.py", line 7341, in prepare_tensors
Zirel-3     super().prepare_tensors()
Zirel-3   File "/llmjob/llama.cpp/convert_hf_to_gguf.py", line 303, in prepare_tensors
Zirel-3     for new_name, data_torch in (self.modify_tensors(data_torch, name, bid)):
Zirel-3   File "/llmjob/llama.cpp/convert_hf_to_gguf.py", line 7320, in modify_tensors
Zirel-3     datas.append(self._experts[bid][ename])
Zirel-3 KeyError: 'model.layers.46.mlp.experts.8.down_proj.weight'
Zirel-3 yes: standard output: Broken pipe
Zirel-3 job finished, status 1
Zirel-3 job-done<0 Zirel-3 noquant 1>
Zirel-3
Zirel-3 NAME: Zirel-3
Zirel-3 TIME: Wed Oct 22 05:03:12 2025

Thanks for checking!

redaihf changed discussion status to closed

Sign up or log in to comment