CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-27 08:21:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama : add CodeShell support (#5016)

Browse Source 
* llama: add codeshell support

* llama.cpp: fix codeshell with NeoX rope

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
chiranko
2024-01-19 17:07:27 +08:00
committed by GitHub
parent 993fba8180
commit 2b3b999cac
4 changed files with 268 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

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

@@ -99,6 +99,7 @@ class MODEL_ARCH(IntEnum):
QWEN = auto()
PHI2 = auto()
PLAMO = auto()
CODESHELL = auto()
class MODEL_TENSOR(IntEnum):
@@ -147,6 +148,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.QWEN: "qwen",
MODEL_ARCH.PHI2: "phi2",
MODEL_ARCH.PLAMO: "plamo",
MODEL_ARCH.CODESHELL: "codeshell",
}
TENSOR_NAMES: dict[MODEL_TENSOR, str] = {
@@ -396,6 +398,19 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.FFN_NORM,
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
],
MODEL_ARCH.CODESHELL: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.POS_EMBD,
MODEL_TENSOR.OUTPUT_NORM,
MODEL_TENSOR.OUTPUT,
MODEL_TENSOR.ATTN_NORM,
MODEL_TENSOR.ATTN_QKV,
MODEL_TENSOR.ATTN_OUT,
MODEL_TENSOR.ATTN_ROT_EMBD,
MODEL_TENSOR.FFN_NORM,
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
]
# TODO
}
@@ -417,6 +432,10 @@ MODEL_TENSOR_SKIP: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.ROPE_FREQS,
MODEL_TENSOR.ATTN_ROT_EMBD,
],
MODEL_ARCH.CODESHELL: [
MODEL_TENSOR.ROPE_FREQS,
MODEL_TENSOR.ATTN_ROT_EMBD,
],
}
#