mikhail/llama.cpp-mtp-turboquant
1
0
Fork 0
Code Issues Pull Requests Actions 29 Packages Projects Releases Wiki Activity

revert : "[Model] Qwen3.5 dense and MoE support (no vision) (#19435)" (#19453)

Browse Source 
This reverts commit 39bf692af1.
This commit is contained in:
Georgi Gerganov
2026-02-09 14:57:51 +02:00
committed by GitHub
parent f5e7734ff2
commit 972f323e73
14 changed files with 399 additions and 1532 deletions
Download Patch File Download Diff File Expand all files Collapse all files
gguf-py/gguf/constants.py
-59
View File
@@ -382,8 +382,6 @@ class MODEL_ARCH(IntEnum):
QWEN3 = auto()
QWEN3MOE = auto()
QWEN3NEXT = auto()
QWEN3_5 = auto()
QWEN3_5_MOE = auto()
QWEN3VL = auto()
QWEN3VLMOE = auto()
PHI2 = auto()
@@ -814,8 +812,6 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.QWEN3: "qwen3",
MODEL_ARCH.QWEN3MOE: "qwen3moe",
MODEL_ARCH.QWEN3NEXT: "qwen3next",
MODEL_ARCH.QWEN3_5: "qwen3_5",
MODEL_ARCH.QWEN3_5_MOE: "qwen3_5moe",
MODEL_ARCH.QWEN3VL: "qwen3vl",
MODEL_ARCH.QWEN3VLMOE: "qwen3vlmoe",
MODEL_ARCH.PHI2: "phi2",
@@ -1788,61 +1784,6 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.SSM_BETA_ALPHA,
MODEL_TENSOR.SSM_OUT
],
MODEL_ARCH.QWEN3_5: [
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.ATTN_GATE,
MODEL_TENSOR.ATTN_QKV,
MODEL_TENSOR.FFN_GATE,
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
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.QWEN3_5_MOE: [
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.ATTN_GATE,
MODEL_TENSOR.ATTN_QKV,
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.QWEN3VL: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,