diff options
| author | Georgi Gerganov <ggerganov@gmail.com> | 2024-02-22 20:13:25 +0200 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-02-22 20:13:25 +0200 |
| commit | 5a9e2f60ba3d8362ba17c77ac3092906d49b813f (patch) | |
| tree | 20792c6cfcd1efcb7a47b8609851c62371da80c6 | |
| parent | 373ee3fbbabc4c1508eed4f5c3795b23a20939a3 (diff) | |
py : minor fixes (#5668)
| -rwxr-xr-x | convert-hf-to-gguf.py | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/convert-hf-to-gguf.py b/convert-hf-to-gguf.py index 9771fccf..8630bbf2 100755 --- a/convert-hf-to-gguf.py +++ b/convert-hf-to-gguf.py @@ -655,6 +655,8 @@ class OrionModel(Model): self.gguf_writer.add_feed_forward_length(self.hparams["intermediate_size"]) self.gguf_writer.add_head_count(head_count) self.gguf_writer.add_head_count_kv(head_count_kv) + # note: config provides rms norm but it is actually layer norm + # ref: https://huggingface.co/OrionStarAI/Orion-14B-Chat/blob/276a17221ce42beb45f66fac657a41540e71f4f5/modeling_orion.py#L570-L571 self.gguf_writer.add_layer_norm_eps(self.hparams["rms_norm_eps"]) def write_tensors(self): @@ -1031,7 +1033,6 @@ class PersimmonModel(Model): self.gguf_writer.add_head_count_kv(head_count_kv) self.gguf_writer.add_rope_freq_base(self.hparams["rope_theta"]) self.gguf_writer.add_layer_norm_eps(self.hparams["layer_norm_eps"]) - self.gguf_writer.add_layer_norm_rms_eps(self.hparams["rms_norm_eps"]) def set_vocab(self): self._set_vocab_sentencepiece() |