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

Add support for ARWKV7 Hybrid models

Browse source 
Signed-off-by: Molly Sophia <mollysophia379@gmail.com>
This commit is contained in:
Molly Sophia 2025-01-27 17:45:43 +08:00
parent e9ba411d3e
commit f6be4dc661
7 changed files with 420 additions and 72 deletions
Download patch file Download diff file Expand all files Collapse all files

33
gguf-py/gguf/constants.py
View file

@ -261,6 +261,7 @@ class MODEL_ARCH(IntEnum):
RWKV6 = auto()
RWKV6QWEN2 = auto()
RWKV7 = auto()
ARWKV7 = auto()
MAMBA = auto()
XVERSE = auto()
COMMAND_R = auto()
@ -461,6 +462,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.RWKV6: "rwkv6",
MODEL_ARCH.RWKV6QWEN2: "rwkv6qwen2",
MODEL_ARCH.RWKV7: "rwkv7",
MODEL_ARCH.ARWKV7: "arwkv7",
MODEL_ARCH.MAMBA: "mamba",
MODEL_ARCH.XVERSE: "xverse",
MODEL_ARCH.COMMAND_R: "command-r",
@ -1214,6 +1216,37 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.CHANNEL_MIX_KEY,
MODEL_TENSOR.CHANNEL_MIX_VALUE,
],
MODEL_ARCH.ARWKV7: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.TOKEN_EMBD_NORM,
MODEL_TENSOR.OUTPUT_NORM,
MODEL_TENSOR.OUTPUT,
MODEL_TENSOR.ATTN_NORM,
MODEL_TENSOR.TIME_MIX_LERP_FUSED,
MODEL_TENSOR.TIME_MIX_W0,
MODEL_TENSOR.TIME_MIX_W1,
MODEL_TENSOR.TIME_MIX_W2,
MODEL_TENSOR.TIME_MIX_A0,
MODEL_TENSOR.TIME_MIX_A1,
MODEL_TENSOR.TIME_MIX_A2,
MODEL_TENSOR.TIME_MIX_V0,
MODEL_TENSOR.TIME_MIX_V1,
MODEL_TENSOR.TIME_MIX_V2,
MODEL_TENSOR.TIME_MIX_G1,
MODEL_TENSOR.TIME_MIX_G2,
MODEL_TENSOR.TIME_MIX_K_K,
MODEL_TENSOR.TIME_MIX_K_A,
MODEL_TENSOR.TIME_MIX_R_K,
MODEL_TENSOR.TIME_MIX_KEY,
MODEL_TENSOR.TIME_MIX_VALUE,
MODEL_TENSOR.TIME_MIX_RECEPTANCE,
MODEL_TENSOR.TIME_MIX_LN,
MODEL_TENSOR.TIME_MIX_OUTPUT,
MODEL_TENSOR.FFN_NORM,
MODEL_TENSOR.FFN_GATE,
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
],
MODEL_ARCH.MAMBA: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,