diff options
| author | Galunid <karolek1231456@gmail.com> | 2023-11-14 11:17:12 +0100 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2023-11-14 11:17:12 +0100 |
| commit | 36eed0c42c5b0bf74af81fb9243d262014f9382f (patch) | |
| tree | d951bf9acc7a7ac3f2e3a937ff39edff8516da7e /gguf-py | |
| parent | b46d12f86d56bef3dc8b596dfb3d22f3b08102be (diff) | |
stablelm : StableLM support (#3586)
* Add support for stablelm-3b-4e1t
* Supports GPU offloading of (n-1) layers
Diffstat (limited to 'gguf-py')
| -rw-r--r-- | gguf-py/gguf/constants.py | 17 |
1 files changed, 17 insertions, 0 deletions
diff --git a/gguf-py/gguf/constants.py b/gguf-py/gguf/constants.py index bf1ccf66..7f63361b 100644 --- a/gguf-py/gguf/constants.py +++ b/gguf-py/gguf/constants.py @@ -90,6 +90,7 @@ class MODEL_ARCH(IntEnum): REFACT = auto() BERT = auto() BLOOM = auto() + STABLELM = auto() class MODEL_TENSOR(IntEnum): @@ -129,6 +130,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = { MODEL_ARCH.REFACT: "refact", MODEL_ARCH.BERT: "bert", MODEL_ARCH.BLOOM: "bloom", + MODEL_ARCH.STABLELM: "stablelm", } TENSOR_NAMES: dict[MODEL_TENSOR, str] = { @@ -299,6 +301,21 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = { MODEL_TENSOR.FFN_DOWN, MODEL_TENSOR.FFN_UP, ], + MODEL_ARCH.STABLELM: [ + 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.FFN_NORM, + MODEL_TENSOR.FFN_GATE, + MODEL_TENSOR.FFN_DOWN, + MODEL_TENSOR.FFN_UP, + ], MODEL_ARCH.GPT2: [ # TODO ], |