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

llama : add MiniCPM support (#5346)

Browse source 
* support minicpm arch.

* fix tab/space typo.

* convert minicpm model via convert-hf-gguf.py

* try to make tokenizer work

* fix bug for quantize minicpm

* fix for flake8 lint

* remove convert-minicpm.py

* fix for editorconfig

* correct minicpm model type (size)

* constants expanded for minicpm

* Minor change of the constant names for minicpm
This commit is contained in:
runfuture 2024-02-07 14:15:56 +08:00 • committed by GitHub
parent f3e2b4fa3f
commit 316c7faf77
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 259 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

21
gguf-py/gguf/constants.py
View file

@ -104,6 +104,7 @@ class MODEL_ARCH(IntEnum):
CODESHELL = auto()
ORION = auto()
INTERNLM2 = auto()
MINICPM = auto()
class MODEL_TENSOR(IntEnum):
@ -156,6 +157,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.CODESHELL: "codeshell",
MODEL_ARCH.ORION: "orion",
MODEL_ARCH.INTERNLM2: "internlm2",
MODEL_ARCH.MINICPM: "minicpm",
}
TENSOR_NAMES: dict[MODEL_TENSOR, str] = {
@ -464,6 +466,25 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
],
MODEL_ARCH.MINICPM: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,
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,
],
# TODO
}