vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-15 11:29:43 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

llama : fix kq_scale for the attention layers of PLaMo2 (#14892)

Browse source 
* Fix dimensions for expand

* Change dimensions to copy states to cache

* Fix the default value for plamo2 conversion

* Fix scale given to build_attn

* Update src/llama-model.cpp

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

* Update src/llama-model.cpp

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

* Update src/llama-model.cpp

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

---------

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
This commit is contained in:
Shunta Saito 2025-07-27 16:38:44 +09:00 • committed by GitHub
parent 446595b9b3
commit 1dc9614e06
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 10 additions and 9 deletions
Download patch file Download diff file Expand all files Collapse all files

4
convert_hf_to_gguf.py
View file

@ -3791,7 +3791,7 @@ class Plamo2Model(TextModel):
self.gguf_writer.add_block_count(block_count)
self.gguf_writer.add_head_count(hparams.get("num_attention_heads", 32))
self.gguf_writer.add_layer_norm_rms_eps(hparams.get("rms_norm_eps", 1e-06))
self.gguf_writer.add_rope_freq_base(hparams.get("rope_theta", 1000000.0))
self.gguf_writer.add_rope_freq_base(hparams.get("rope_theta", 10000))
# Mamba parameters
self.gguf_writer.add_ssm_state_size(hparams.get("mamba_d_state", 64))
@ -3802,7 +3802,7 @@ class Plamo2Model(TextModel):
self.gguf_writer.add_ssm_group_count(0)
# MLP feed forward parameters (for attention layers)
self.gguf_writer.add_feed_forward_length(hparams.get("intermediate_size", 16384))
self.gguf_writer.add_feed_forward_length(hparams.get("intermediate_size", 13312))
self.gguf_writer.add_file_type(self.ftype)
def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]: