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

Add OLMo November 2024 constants

Browse source
This commit is contained in:
Shane A 2024-11-18 10:34:50 -08:00
parent 09ecbcb596
commit 8c4a9137f3
1 changed files with 18 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

18
gguf-py/gguf/constants.py
View file

@ -243,6 +243,7 @@ class MODEL_ARCH(IntEnum):
COMMAND_R = auto() COMMAND_R = auto()
DBRX = auto() DBRX = auto()
OLMO = auto() OLMO = auto()
OLMO_1124 = auto()
OLMOE = auto() OLMOE = auto()
OPENELM = auto() OPENELM = auto()
ARCTIC = auto() ARCTIC = auto()
@ -404,6 +405,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.COMMAND_R: "command-r", MODEL_ARCH.COMMAND_R: "command-r",
MODEL_ARCH.DBRX: "dbrx", MODEL_ARCH.DBRX: "dbrx",
MODEL_ARCH.OLMO: "olmo", MODEL_ARCH.OLMO: "olmo",
MODEL_ARCH.OLMO_1124: "olmo_1124",
MODEL_ARCH.OLMOE: "olmoe", MODEL_ARCH.OLMOE: "olmoe",
MODEL_ARCH.OPENELM: "openelm", MODEL_ARCH.OPENELM: "openelm",
MODEL_ARCH.ARCTIC: "arctic", MODEL_ARCH.ARCTIC: "arctic",
@ -1069,6 +1071,22 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.FFN_DOWN, MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP, MODEL_TENSOR.FFN_UP,
], ],
MODEL_ARCH.OLMO_1124: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,
MODEL_TENSOR.OUTPUT,
MODEL_TENSOR.ATTN_Q,
MODEL_TENSOR.ATTN_K,
MODEL_TENSOR.ATTN_V,
MODEL_TENSOR.ATTN_OUT,
MODEL_TENSOR.ATTN_POST_NORM,
MODEL_TENSOR.ATTN_Q_NORM,
MODEL_TENSOR.ATTN_K_NORM,
MODEL_TENSOR.FFN_POST_NORM,
MODEL_TENSOR.FFN_GATE,
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
],
MODEL_ARCH.OLMOE: [ MODEL_ARCH.OLMOE: [
MODEL_TENSOR.TOKEN_EMBD, MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM, MODEL_TENSOR.OUTPUT_NORM,