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

add chatglm3-6b model support huggingface model:

Browse source 
https://hf-mirror.com/THUDM/chatglm3-6b

Signed-off-by: XingXing Qiao <qiaoxx@dingdao.com>
This commit is contained in:
XingXing Qiao 2024-05-29 13:30:07 +08:00
parent b864b50ce5
commit 6630a2da48
6 changed files with 398 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

17
gguf-py/gguf/constants.py
View file

@ -148,6 +148,7 @@ class MODEL_ARCH(IntEnum):
OLMO = auto()
ARCTIC = auto()
DEEPSEEK2 = auto()
CHATGLM = auto()
class MODEL_TENSOR(IntEnum):
@ -236,6 +237,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.OLMO: "olmo",
MODEL_ARCH.ARCTIC: "arctic",
MODEL_ARCH.DEEPSEEK2: "deepseek2",
MODEL_ARCH.CHATGLM: "chatglm",
}
TENSOR_NAMES: dict[MODEL_TENSOR, str] = {
@ -805,6 +807,18 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.FFN_DOWN_SHEXP,
MODEL_TENSOR.FFN_UP_SHEXP,
],
MODEL_ARCH.CHATGLM : [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.ROPE_FREQS,
MODEL_TENSOR.OUTPUT_NORM,
MODEL_TENSOR.OUTPUT,
MODEL_TENSOR.ATTN_NORM,
MODEL_TENSOR.ATTN_QKV,
MODEL_TENSOR.ATTN_OUT,
MODEL_TENSOR.FFN_NORM,
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
],
# TODO
}
@ -842,6 +856,9 @@ MODEL_TENSOR_SKIP: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.ROPE_FREQS,
MODEL_TENSOR.ATTN_ROT_EMBD,
],
MODEL_ARCH.CHATGLM: [
MODEL_TENSOR.ROPE_FREQS,
],
}
#