CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-28 08:31:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

First draft

Browse Source
This commit is contained in:
Piotr Wilkin
2025-09-18 00:21:17 +02:00
parent cd08fc3ecc
commit 344331c2b6
11 changed files with 856 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

32
gguf-py/gguf/constants.py
View File

@@ -335,6 +335,7 @@ class MODEL_ARCH(IntEnum):
QWEN2VL = auto()
QWEN3 = auto()
QWEN3MOE = auto()
QWEN3NEXT = auto()
PHI2 = auto()
PHI3 = auto()
PHIMOE = auto()
@@ -481,6 +482,7 @@ class MODEL_TENSOR(IntEnum):
SSM_D = auto()
SSM_NORM = auto()
SSM_OUT = auto()
SSM_BETA_ALPHA = auto()
TIME_MIX_W0 = auto()
TIME_MIX_W1 = auto()
TIME_MIX_W2 = auto()
@@ -671,6 +673,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.QWEN2VL: "qwen2vl",
MODEL_ARCH.QWEN3: "qwen3",
MODEL_ARCH.QWEN3MOE: "qwen3moe",
MODEL_ARCH.QWEN3NEXT: "qwen3next",
MODEL_ARCH.PHI2: "phi2",
MODEL_ARCH.PHI3: "phi3",
MODEL_ARCH.PHIMOE: "phimoe",
@@ -818,6 +821,7 @@ TENSOR_NAMES: dict[MODEL_TENSOR, str] = {
MODEL_TENSOR.SSM_D: "blk.{bid}.ssm_d",
MODEL_TENSOR.SSM_NORM: "blk.{bid}.ssm_norm",
MODEL_TENSOR.SSM_OUT: "blk.{bid}.ssm_out",
MODEL_TENSOR.SSM_BETA_ALPHA: "blk.{bid}.ssm_ba",
MODEL_TENSOR.TIME_MIX_W0: "blk.{bid}.time_mix_w0",
MODEL_TENSOR.TIME_MIX_W1: "blk.{bid}.time_mix_w1",
MODEL_TENSOR.TIME_MIX_W2: "blk.{bid}.time_mix_w2",
@@ -1462,6 +1466,34 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.FFN_DOWN_EXP,
MODEL_TENSOR.FFN_UP_EXP,
],
MODEL_ARCH.QWEN3NEXT: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,
MODEL_TENSOR.OUTPUT,
MODEL_TENSOR.ATTN_NORM,
MODEL_TENSOR.ATTN_Q,
MODEL_TENSOR.ATTN_Q_NORM,
MODEL_TENSOR.ATTN_K,
MODEL_TENSOR.ATTN_K_NORM,
MODEL_TENSOR.ATTN_V,
MODEL_TENSOR.ATTN_OUT,
MODEL_TENSOR.ATTN_POST_NORM,
MODEL_TENSOR.FFN_GATE_INP,
MODEL_TENSOR.FFN_GATE_INP_SHEXP,
MODEL_TENSOR.FFN_UP_SHEXP,
MODEL_TENSOR.FFN_DOWN_SHEXP,
MODEL_TENSOR.FFN_GATE_SHEXP,
MODEL_TENSOR.FFN_DOWN_EXP,
MODEL_TENSOR.FFN_UP_EXP,
MODEL_TENSOR.FFN_GATE_EXP,
MODEL_TENSOR.SSM_A,
MODEL_TENSOR.SSM_CONV1D,
MODEL_TENSOR.SSM_DT,
MODEL_TENSOR.SSM_NORM,
MODEL_TENSOR.SSM_IN,
MODEL_TENSOR.SSM_BETA_ALPHA,
MODEL_TENSOR.SSM_OUT
],
MODEL_ARCH.PLAMO: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,

22
gguf-py/gguf/tensor_mapping.py
View File

@@ -628,10 +628,11 @@ class TensorNameMap:
),
MODEL_TENSOR.SSM_IN: (
"model.layers.{bid}.in_proj", # mamba-hf
"backbone.layers.{bid}.mixer.in_proj", # mamba
"model.layers.{bid}.mamba.in_proj", # jamba falcon-h1 granite-hybrid
"model.layers.layers.{bid}.mixer.in_proj", # plamo2
"model.layers.{bid}.in_proj", # mamba-hf
"backbone.layers.{bid}.mixer.in_proj", # mamba
"model.layers.{bid}.mamba.in_proj", # jamba falcon-h1 granite-hybrid
"model.layers.layers.{bid}.mixer.in_proj", # plamo2
"model.layers.{bid}.linear_attn.in_proj_qkvz", # qwen3next
),
MODEL_TENSOR.SSM_CONV1D: (
@@ -639,6 +640,7 @@ class TensorNameMap:
"backbone.layers.{bid}.mixer.conv1d", # mamba
"model.layers.{bid}.mamba.conv1d", # jamba falcon-h1 granite-hybrid
"model.layers.layers.{bid}.mixer.conv1d", # plamo2
"model.layers.{bid}.linear_attn.conv1d", # qwen3next
),
MODEL_TENSOR.SSM_X: (
@@ -653,6 +655,7 @@ class TensorNameMap:
"backbone.layers.{bid}.mixer.dt_proj", # mamba
"model.layers.{bid}.mamba.dt_proj", # jamba falcon-h1 granite-hybrid
"model.layers.layers.{bid}.mixer.dt_proj", # plamo2
"model.layers.{bid}.linear_attn.dt_proj", # qwen3next
),
MODEL_TENSOR.SSM_DT_NORM: (
@@ -665,6 +668,7 @@ class TensorNameMap:
"backbone.layers.{bid}.mixer.A_log", # mamba
"model.layers.{bid}.mamba.A_log", # jamba falcon-h1 granite-hybrid
"model.layers.layers.{bid}.mixer.A_log", # plamo2
"model.layers.{bid}.linear_attn.A_log", # qwen3next
),
MODEL_TENSOR.SSM_B_NORM: (
@@ -687,17 +691,23 @@ class TensorNameMap:
),
MODEL_TENSOR.SSM_NORM: (
"model.layers.{bid}.mamba.norm", # falcon-h1 granite-hybrid
"backbone.layers.{bid}.mixer.norm", # mamba2
"model.layers.{bid}.mamba.norm", # falcon-h1 granite-hybrid
"model.layers.{bid}.linear_attn.norm", # qwen3next
"backbone.layers.{bid}.mixer.norm", # mamba2
),
MODEL_TENSOR.SSM_OUT: (
"model.layers.{bid}.out_proj", # mamba-hf
"backbone.layers.{bid}.mixer.out_proj", # mamba
"model.layers.{bid}.mamba.out_proj", # jamba falcon-h1 granite-hybrid
"model.layers.{bid}.linear_attn.out_proj", # qwen3next
"model.layers.layers.{bid}.mixer.out_proj", # plamo2
),
MODEL_TENSOR.SSM_BETA_ALPHA: (
"model.layers.{bid}.linear_attn.in_proj_ba", # qwen3next
),
MODEL_TENSOR.TIME_MIX_W0: (
"model.layers.{bid}.attention.w0", # rwkv7
),