vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

fix typo

Browse source
This commit is contained in:
Jared Van Bortel 2024-02-24 16:31:36 -05:00
parent 19891864e2
commit a0c57b39a5
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
convert-hf-to-gguf.py
View file

@ -1074,7 +1074,7 @@ class StableLMModel(Model):
self.gguf_writer.add_embedding_length(hparams["hidden_size"])
self.gguf_writer.add_block_count(block_count)
self.gguf_writer.add_feed_forward_length(hparams["intermediate_size"])
rotary_factor = self.find_hparam(["partial_rotary_factor", "rotary_pct"])
rotary_factor = self.find_hparam(["partial_rotary_factor", "rope_pct"])
self.gguf_writer.add_rope_dimension_count(int(rotary_factor * (hparams["hidden_size"] // hparams["num_attention_heads"])))
self.gguf_writer.add_head_count(hparams["num_attention_heads"])
self.gguf_writer.add_parallel_residual(hparams["use_parallel_residual"] if "use_parallel_residual" in hparams else True)