Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

model: add dots3-note (#27060)

Browse Source 
* text: conversion

* init impl

* address review comments

* fix rope

* move to a new llama_kv_cache_dsa_iswa
This commit is contained in:
Xuan-Son Nguyen
2026-08-21 19:52:34 +02:00
committed by GitHub
parent 873e5d8e39
commit 5a32f7b66e
20 changed files with 1412 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files
gguf-py/gguf/constants.py
+43
View File
@@ -205,6 +205,9 @@ class Keys:
VALUE_LENGTH_MLA = "{arch}.attention.value_length_mla"
KEY_LENGTH_SWA = "{arch}.attention.key_length_swa"
VALUE_LENGTH_SWA = "{arch}.attention.value_length_swa"
KEY_LENGTH_MLA_SWA = "{arch}.attention.key_length_mla_swa"
VALUE_LENGTH_MLA_SWA = "{arch}.attention.value_length_mla_swa"
KV_LORA_RANK_SWA = "{arch}.attention.kv_lora_rank_swa"
SHARED_KV_LAYERS = "{arch}.attention.shared_kv_layers"
SLIDING_WINDOW_PATTERN = "{arch}.attention.sliding_window_pattern"
TEMPERATURE_SCALE = "{arch}.attention.temperature_scale"
@@ -558,6 +561,7 @@ class MODEL_ARCH(IntEnum):
BAILINGMOE2 = auto()
BAILINGMOE3 = auto()
DOTS1 = auto()
DOTS3NOTE = auto()
ARCEE = auto()
AFMOE = auto()
LAGUNA = auto()
@@ -1275,6 +1279,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.BAILINGMOE2: "bailingmoe2",
MODEL_ARCH.BAILINGMOE3: "bailingmoe3",
MODEL_ARCH.DOTS1: "dots1",
MODEL_ARCH.DOTS3NOTE: "dots3note",
MODEL_ARCH.ARCEE: "arcee",
MODEL_ARCH.AFMOE: "afmoe",
MODEL_ARCH.LAGUNA: "laguna",
@@ -4334,6 +4339,44 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.FFN_UP_EXP,
MODEL_TENSOR.FFN_UP_SHEXP,
],
MODEL_ARCH.DOTS3NOTE: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,
MODEL_TENSOR.OUTPUT,
MODEL_TENSOR.ATTN_NORM,
MODEL_TENSOR.ATTN_Q_A,
MODEL_TENSOR.ATTN_Q_B,
MODEL_TENSOR.ATTN_KV_A_MQA,
MODEL_TENSOR.ATTN_K_B,
MODEL_TENSOR.ATTN_V_B,
MODEL_TENSOR.ATTN_Q_A_NORM,
MODEL_TENSOR.ATTN_KV_A_NORM,
MODEL_TENSOR.ATTN_K_NORM,
MODEL_TENSOR.ATTN_GATE,
MODEL_TENSOR.ATTN_OUT,
MODEL_TENSOR.FFN_NORM,
MODEL_TENSOR.FFN_GATE,
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
MODEL_TENSOR.FFN_GATE_INP,
MODEL_TENSOR.FFN_EXP_PROBS_B,
MODEL_TENSOR.FFN_GATE_EXP,
MODEL_TENSOR.FFN_DOWN_EXP,
MODEL_TENSOR.FFN_UP_EXP,
MODEL_TENSOR.FFN_GATE_SHEXP,
MODEL_TENSOR.FFN_DOWN_SHEXP,
MODEL_TENSOR.FFN_UP_SHEXP,
MODEL_TENSOR.INDEXER_K_NORM,
MODEL_TENSOR.INDEXER_PROJ,
MODEL_TENSOR.INDEXER_ATTN_K,
MODEL_TENSOR.INDEXER_ATTN_Q_B,
# NextN/MTP tensors - preserved but unused
MODEL_TENSOR.NEXTN_EH_PROJ,
MODEL_TENSOR.NEXTN_EMBED_TOKENS,
MODEL_TENSOR.NEXTN_ENORM,
MODEL_TENSOR.NEXTN_HNORM,
MODEL_TENSOR.NEXTN_SHARED_HEAD_NORM,
],
MODEL_ARCH.ARCEE: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,