Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

spec: add eagle3-v3 support for gpt-oss model (#25794)

Browse Source
This commit is contained in:
Ruixiang Wang
2026-07-28 09:58:16 +03:00
committed by GitHub
parent c6292cfb8e
commit f87067841b
9 changed files with 59 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
gguf-py/gguf/constants.py
+2
View File
@@ -161,6 +161,7 @@ class Keys:
TARGET_HIDDEN_SIZE = "{arch}.target_hidden_size"
BLOCK_SIZE = "{arch}.block_size"
NORM_BEFORE_RESIDUAL = "{arch}.norm_before_residual"
NORM_BEFORE_FC = "{arch}.norm_before_fc"
class Attention:
HEAD_COUNT = "{arch}.attention.head_count"
@@ -4343,6 +4344,7 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
MODEL_TENSOR.FC,
MODEL_TENSOR.ENC_OUTPUT_NORM,
MODEL_TENSOR.D2T,
],
MODEL_ARCH.DFLASH: [
gguf-py/gguf/gguf_writer.py
+3
View File
@@ -971,6 +971,9 @@ class GGUFWriter:
def add_norm_before_residual(self, value: bool) -> None:
self.add_bool(Keys.LLM.NORM_BEFORE_RESIDUAL.format(arch=self.arch), value)
def add_norm_before_fc(self, value: bool) -> None:
self.add_bool(Keys.LLM.NORM_BEFORE_FC.format(arch=self.arch), value)
def add_attention_output_group_count(self, count: int) -> None:
self.add_uint32(Keys.Attention.OUTPUT_GROUP_COUNT.format(arch=self.arch), count)