ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-04-30 11:21:56 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add support for bitnet2b_2501 model (#337)

Browse Source 
* add support for bitnet2b_2501 model

* Fixes

* Support both model names

---------

Co-authored-by: potassiummmm <zhou.hansong@outlook.com>
This commit is contained in:
saood06
2025-04-22 01:34:13 -05:00
committed by GitHub
parent 16f945d9bb
commit e6c85a5b95
4 changed files with 330 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
gguf-py/gguf/constants.py
View File

@@ -219,6 +219,7 @@ class MODEL_ARCH(IntEnum):
DEEPSEEK2 = auto()
CHATGLM = auto()
BITNET = auto()
BITNET_25 = auto()
T5 = auto()
T5ENCODER = auto()
JAIS = auto()
@@ -351,6 +352,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.DEEPSEEK2: "deepseek2",
MODEL_ARCH.CHATGLM: "chatglm",
MODEL_ARCH.BITNET: "bitnet",
MODEL_ARCH.BITNET_25: "bitnet-25",
MODEL_ARCH.T5: "t5",
MODEL_ARCH.T5ENCODER: "t5encoder",
MODEL_ARCH.JAIS: "jais",
@@ -1019,6 +1021,28 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.ATTN_SUB_NORM,
MODEL_TENSOR.FFN_SUB_NORM,
],
MODEL_ARCH.BITNET_25: [
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_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,
MODEL_TENSOR.ATTN_SUB_NORM,
MODEL_TENSOR.FFN_SUB_NORM,
],
MODEL_ARCH.T5: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT,

5
gguf-py/gguf/tensor_mapping.py
View File

@@ -131,6 +131,7 @@ class TensorNameMap:
"model.layers.{bid}.self_attn.qkv_proj", # phi3
"encoder.layers.{bid}.self_attention.query_key_value", # chatglm
"transformer.layers.{bid}.attn.qkv_proj", # openelm
"layers.{bid}.attention.wqkv",
),
# Attention query
@@ -464,10 +465,14 @@ class TensorNameMap:
MODEL_TENSOR.ATTN_SUB_NORM: (
"model.layers.{bid}.self_attn.inner_attn_ln", # bitnet
"layers.{bid}.attention.attn_sub_norm", # bitnet
"model.layers.{bid}.self_attn.attn_sub_norm",
),
MODEL_TENSOR.FFN_SUB_NORM: (
"model.layers.{bid}.mlp.ffn_layernorm", # bitnet
"layers.{bid}.feed_forward.ffn_sub_norm", # bitnet
"model.layers.{bid}.mlp.ffn_sub_norm",
),
MODEL_TENSOR.DEC_ATTN_NORM: (