tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-26 19:55:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama : Support llama 4 text-only (#12791)

Browse Source 
* llama4 conversion

* initial support, no chat template

* clean up a bit

* fix tokenizer conversion

* correct hparams

* try this

* fix shexp

* ffn_inp_normed

* chat template

* clean up model conversion

* add_bos

* add scale_before_ffn

* fix order

* weight_before_ffn

* llm_graph_input_attn_temp

* add chunk attn mask

* build_inp_attn_scale()

* add comment about ggml_repeat

* clarify comments

* fix build
This commit is contained in:
Xuan-Son Nguyen
2025-04-07 23:06:44 +02:00
committed by GitHub
parent 82974011f3
commit 1466621e73
17 changed files with 532 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
gguf-py/gguf/constants.py
View File

@ -116,6 +116,7 @@ class Keys:
RESIDUAL_SCALE = "{arch}.residual_scale"
EMBEDDING_SCALE = "{arch}.embedding_scale"
TOKEN_SHIFT_COUNT = "{arch}.token_shift_count"
INTERLEAVE_MOE_LAYER_STEP = "{arch}.interleave_moe_layer_step"
class Attention:
HEAD_COUNT = "{arch}.attention.head_count"
@ -227,6 +228,7 @@ class GGUFType:
class MODEL_ARCH(IntEnum):
LLAMA = auto()
LLAMA4 = auto()
DECI = auto()
FALCON = auto()
BAICHUAN = auto()
@ -431,6 +433,7 @@ class MODEL_TENSOR(IntEnum):
MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.LLAMA: "llama",
MODEL_ARCH.LLAMA4: "llama4",
MODEL_ARCH.DECI: "deci",
MODEL_ARCH.FALCON: "falcon",
MODEL_ARCH.BAICHUAN: "baichuan",
@ -654,6 +657,29 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.FFN_DOWN_EXP,
MODEL_TENSOR.FFN_UP_EXP,
],
MODEL_ARCH.LLAMA4: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,
MODEL_TENSOR.OUTPUT,
MODEL_TENSOR.ROPE_FREQS,
MODEL_TENSOR.ATTN_NORM,
MODEL_TENSOR.ATTN_Q,
MODEL_TENSOR.ATTN_K,
MODEL_TENSOR.ATTN_V,
MODEL_TENSOR.ATTN_OUT,
MODEL_TENSOR.ATTN_ROT_EMBD,
MODEL_TENSOR.FFN_GATE_INP,
MODEL_TENSOR.FFN_NORM,
MODEL_TENSOR.FFN_GATE,
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
MODEL_TENSOR.FFN_GATE_EXP,
MODEL_TENSOR.FFN_DOWN_EXP,
MODEL_TENSOR.FFN_UP_EXP,
MODEL_TENSOR.FFN_GATE_SHEXP,
MODEL_TENSOR.FFN_DOWN_SHEXP,
MODEL_TENSOR.FFN_UP_SHEXP,
],
MODEL_ARCH.DECI: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,

3
gguf-py/gguf/gguf_writer.py
View File

@ -746,6 +746,9 @@ class GGUFWriter:
def add_token_shift_count(self, count: int) -> None:
self.add_uint32(Keys.LLM.TOKEN_SHIFT_COUNT.format(arch=self.arch), count)
def add_interleave_moe_layer_step(self, value: int) -> None:
self.add_uint32(Keys.LLM.INTERLEAVE_MOE_LAYER_STEP.format(arch=self.arch), value)
def add_layer_norm_eps(self, value: float) -> None:
self.add_float32(Keys.Attention.LAYERNORM_EPS.format(arch=self.arch), value)