EngineX-Ascend/enginex-ascend-910-llama.cpp
10
0
Fork 0
Code Issues Pull Requests Actions 4 Projects Releases Wiki Activity

model : gemma3n text-only (#14400)

Browse Source 
* gemma3n

* add llm_graph_input_one
This commit is contained in:
Xuan-Son Nguyen
2025-06-26 19:34:02 +02:00
committed by GitHub
parent a01047b041
commit 8846aace49
13 changed files with 960 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
gguf-py/gguf/gguf_writer.py
View File

@@ -672,6 +672,18 @@ class GGUFWriter:
def add_decoder_start_token_id(self, id: int) -> None:
self.add_uint32(Keys.LLM.DECODER_START_TOKEN_ID.format(arch=self.arch), id)
def add_embedding_length_per_layer_input(self, value: int) -> None:
self.add_uint32(Keys.LLM.EMBD_LENGTH_PER_LAYER_INP.format(arch=self.arch), value)
def add_altup_active_idx(self, val: int) -> None:
self.add_uint32(Keys.LLM.ALTUP_ACTIVE_IDX.format(arch=self.arch), val)
def add_altup_num_inputs(self, val: int) -> None:
self.add_uint32(Keys.LLM.ALTUP_NUM_INPUTS.format(arch=self.arch), val)
def add_activation_sparsity_scale(self, values: Sequence[float]) -> None:
self.add_array(Keys.LLM.ACTIVATION_SPARSITY_SCALE.format(arch=self.arch), values)
def add_head_count(self, count: int | Sequence[int]) -> None:
if isinstance(count, int):
self.add_uint32(Keys.Attention.HEAD_COUNT.format(arch=self.arch), count)
@@ -702,6 +714,12 @@ class GGUFWriter:
def add_clamp_kqv(self, value: float) -> None:
self.add_float32(Keys.Attention.CLAMP_KQV.format(arch=self.arch), value)
def add_shared_kv_layers(self, value: float) -> None:
self.add_float32(Keys.Attention.SHARED_KV_LAYERS.format(arch=self.arch), value)
def add_sliding_window_pattern(self, value: Sequence[bool]) -> None:
self.add_array(Keys.Attention.SLIDING_WINDOW_PATTERN.format(arch=self.arch), value)
def add_logit_scale(self, value: float) -> None:
self.add_float32(Keys.LLM.LOGIT_SCALE.format(arch=self.arch), value)