marfrit/rk-llama.cpp
1
0
Fork 0
Code Issues Pull Requests Actions 30 Packages Projects Releases Wiki Activity

model : add EXAONE MoE (#18543)

Browse Source 
* Add EXAONE MoE implementations

Co-authored-by: Junwon Hwang <nuclear1221@gmail.com>

* Address PR feedback

* Address PR feedback

* [WIP] Add MTP for EXAONE-MoE

* Address PR feedback

* Address PR feedback

* Address PR feedback

* Address PR feedback

* Address PR feedback

* Address PR feedback

* Address PR feedback

---------

Co-authored-by: LG-AI-EXAONE <exaonemodels@lgresearch.ai>
This commit is contained in:
Junwon Hwang
2026-01-14 07:28:38 +09:00
committed by GitHub
parent e4832e3ae4
commit 60591f01d4
17 changed files with 656 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
gguf-py/gguf/constants.py
+34
View File
@@ -424,6 +424,7 @@ class MODEL_ARCH(IntEnum):
NEMOTRON_H_MOE = auto()
EXAONE = auto()
EXAONE4 = auto()
EXAONE_MOE = auto()
GRANITE = auto()
GRANITE_MOE = auto()
GRANITE_HYBRID = auto()
@@ -843,6 +844,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.NEMOTRON_H_MOE: "nemotron_h_moe",
MODEL_ARCH.EXAONE: "exaone",
MODEL_ARCH.EXAONE4: "exaone4",
MODEL_ARCH.EXAONE_MOE: "exaone-moe",
MODEL_ARCH.GRANITE: "granite",
MODEL_ARCH.GRANITE_MOE: "granitemoe",
MODEL_ARCH.GRANITE_HYBRID: "granitehybrid",
@@ -2754,6 +2756,38 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.FFN_UP,
MODEL_TENSOR.FFN_POST_NORM,
],
MODEL_ARCH.EXAONE_MOE: [
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_Q_NORM,
MODEL_TENSOR.ATTN_K,
MODEL_TENSOR.ATTN_K_NORM,
MODEL_TENSOR.ATTN_V,
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_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.FFN_EXP_PROBS_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_HEAD,
MODEL_TENSOR.NEXTN_SHARED_HEAD_NORM,
],
MODEL_ARCH.GRANITE: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,
gguf-py/gguf/tensor_mapping.py
+3 -2
View File
@@ -436,7 +436,8 @@ class TensorNameMap:
"model.layers.{bid}.mlp.expert_bias", # afmoe
"model.layers.{bid}.feed_forward.expert_bias", # lfm2moe
"model.layers.{bid}.block_sparse_moe.e_score_correction", # minimax-m2
"backbone.layers.{bid}.mixer.gate.e_score_correction" # nemotron-h-moe
"backbone.layers.{bid}.mixer.gate.e_score_correction", # nemotron-h-moe
"model.layers.{bid}.mlp.e_score_correction", # exaone-moe
),
# Feed-forward up
@@ -1794,7 +1795,7 @@ class TensorNameMap:
"model.embed_audio.soft_embedding_norm", # gemma3n
),
# NextN/MTP tensors for GLM4_MOE
# NextN/MTP tensors
MODEL_TENSOR.NEXTN_EH_PROJ: (
"model.layers.{bid}.eh_proj",
),