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

llama : add StarCoder2 support (#5795)

Browse source 
* Add support for starcoder2

* handle rope type

* skip rope freq and rotary embeddings from being serialized

* resolve comments

* Update llama.cpp

* remove redundant changes

* handle `rope-theta`

* llama : change starcoder2 rope type

* address comment

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Sourab Mangrulkar 2024-03-02 01:00:46 +05:30 • committed by GitHub
parent 38d16b1426
commit c29af7e225
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
4 changed files with 229 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

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

@ -112,6 +112,7 @@ class MODEL_ARCH(IntEnum):
INTERNLM2 = auto()
MINICPM = auto()
GEMMA = auto()
STARCODER2 = auto()
class MODEL_TENSOR(IntEnum):
@ -169,6 +170,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.INTERNLM2: "internlm2",
MODEL_ARCH.MINICPM: "minicpm",
MODEL_ARCH.GEMMA: "gemma",
MODEL_ARCH.STARCODER2: "starcoder2",
}
TENSOR_NAMES: dict[MODEL_TENSOR, str] = {
@ -526,6 +528,21 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.FFN_UP,
MODEL_TENSOR.FFN_NORM,
],
MODEL_ARCH.STARCODER2: [
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_NORM,
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
],
# TODO
}
@ -554,6 +571,10 @@ MODEL_TENSOR_SKIP: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.ROPE_FREQS,
MODEL_TENSOR.ATTN_ROT_EMBD,
],
MODEL_ARCH.STARCODER2: [
MODEL_TENSOR.ROPE_FREQS,
MODEL_TENSOR.ATTN_ROT_EMBD,
],
}
#