transformers
fix(llama): allow explicit head_dim when hidden_size not divisible by num_attention_heads
#46211
Open
Open
fix(llama): allow explicit head_dim when hidden_size not divisible by num_attention_heads #46211
Sriniketh24
wants to merge
2 commits into
huggingface:main
from
Sriniketh24:fix/llama-config-explicit-head-dim-validation
fix(llama): allow explicit head_dim when hidden_size is not divisible…
9bac047c
chore: regenerate modular configs that inherit LlamaConfig.__post_init__
b83db827
Rocketknight1
commented
on 2026-05-26
Login to write a write a comment.
Login via GitHub