nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 18 Packages Projects Releases Wiki Activity

convert : more consistent handling of rope_parameters (#24833)

Browse Source
This commit is contained in:
Sigbjørn Skjæret
2026-06-20 13:42:36 +03:00
committed by GitHub
parent f449e05537
commit f4043fec01
15 changed files with 40 additions and 39 deletions
Download Patch File Download Diff File Expand all files Collapse all files
conversion/exaone.py
+3 -3
View File
@@ -24,7 +24,7 @@ class ExaoneModel(TextModel):
assert (hparams["activation_function"] == "silu")
rotary_factor = self.find_hparam(["partial_rotary_factor", "rope_pct"], optional=True)
rotary_factor = self.rope_parameters.get("partial_rotary_factor")
rotary_factor = rotary_factor if rotary_factor is not None else 1.0
self.gguf_writer.add_rope_dimension_count(int(rotary_factor * (hparams["hidden_size"] // hparams["num_attention_heads"])))
@@ -39,7 +39,7 @@ class ExaoneModel(TextModel):
factor = rope_params.get("factor", 8.0)
low_freq_factor = rope_params.get("low_freq_factor", 1.0)
high_freq_factor = rope_params.get("high_freq_factor", 4.0)
old_context_len = self.hparams.get("original_max_position_embeddings", 8192)
old_context_len = rope_params.get("original_max_position_embeddings", 8192)
low_freq_wavelen = old_context_len / low_freq_factor
high_freq_wavelen = old_context_len / high_freq_factor
@@ -104,7 +104,7 @@ class Exaone4Model(TextModel):
factor = rope_params.get("factor", 16.0)
low_freq_factor = rope_params.get("low_freq_factor", 1.0)
high_freq_factor = rope_params.get("high_freq_factor", 4.0)
old_context_len = self.hparams.get("original_max_position_embeddings", 8192)
old_context_len = rope_params.get("original_max_position_embeddings", 8192)
low_freq_wavelen = old_context_len / low_freq_factor
high_freq_wavelen = old_context_len / high_freq_factor