vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

llama : support MiniCPM3 (#9322)

Browse source 
Co-authored-by: 范睿凯 <fanruikai@modelbest.cn>
This commit is contained in:
CarryFun 2024-09-16 14:45:20 +08:00 • committed by GitHub
parent 441b72b91f
commit 95ca85168b
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 371 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

19
gguf-py/gguf/constants.py
View file

@ -210,6 +210,7 @@ class MODEL_ARCH(IntEnum):
ORION = auto()
INTERNLM2 = auto()
MINICPM = auto()
MINICPM3 = auto()
GEMMA = auto()
GEMMA2 = auto()
STARCODER2 = auto()
@ -364,6 +365,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.ORION: "orion",
MODEL_ARCH.INTERNLM2: "internlm2",
MODEL_ARCH.MINICPM: "minicpm",
MODEL_ARCH.MINICPM3: "minicpm3",
MODEL_ARCH.GEMMA: "gemma",
MODEL_ARCH.GEMMA2: "gemma2",
MODEL_ARCH.STARCODER2: "starcoder2",
@ -867,6 +869,23 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.FFN_DOWN_EXP,
MODEL_TENSOR.FFN_UP_EXP,
],
MODEL_ARCH.MINICPM3: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,
MODEL_TENSOR.OUTPUT,
MODEL_TENSOR.ATTN_NORM,
MODEL_TENSOR.ATTN_Q_A,
MODEL_TENSOR.ATTN_Q_B,
MODEL_TENSOR.ATTN_KV_A_MQA,
MODEL_TENSOR.ATTN_KV_B,
MODEL_TENSOR.ATTN_Q_A_NORM,
MODEL_TENSOR.ATTN_KV_A_NORM,
MODEL_TENSOR.ATTN_OUT,
MODEL_TENSOR.FFN_NORM,
MODEL_TENSOR.FFN_GATE,
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
],
MODEL_ARCH.GEMMA: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,