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/minicpm.py
+6 -10
View File
@@ -32,11 +32,9 @@ class MiniCPMModel(TextModel):
def generate_extra_tensors(self) -> Iterable[tuple[str, Tensor]]:
rope_dims = self.hparams["hidden_size"] // self.hparams["num_attention_heads"]
rope_scaling = self.find_hparam(['rope_scaling'], True)
if rope_scaling is not None:
long_factors = rope_scaling.get('long_factor', None)
short_factors = rope_scaling.get('short_factor', None)
long_factors = self.rope_parameters.get('long_factor')
short_factors = self.rope_parameters.get('short_factor')
if long_factors or short_factors:
if long_factors is None or short_factors is None:
raise KeyError('Missing the required key rope_scaling.long_factor or rope_scaling_short_factor')
@@ -85,13 +83,11 @@ class MiniCPM3Model(TextModel):
self.gguf_writer.add_rope_dimension_count(hparams["qk_rope_head_dim"])
def generate_extra_tensors(self) -> Iterable[tuple[str, Tensor]]:
rope_scaling = self.find_hparam(['rope_scaling'], True)
if rope_scaling is not None:
long_factors = self.rope_parameters.get('long_factor')
short_factors = self.rope_parameters.get('short_factor')
if long_factors or short_factors:
rope_dims = self.hparams["qk_rope_head_dim"]
long_factors = rope_scaling.get('long_factor', None)
short_factors = rope_scaling.get('short_factor', None)
if long_factors is None or short_factors is None:
raise KeyError('Missing the required key rope_scaling.long_factor or rope_scaling_short_factor')