CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-11 10:36:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'master' into compilade/refactor-kv-cache

Browse Source
This commit is contained in:
Francis Couture-Harpin
2025-07-08 15:09:02 -04:00
parent f71635824b 6efcd65945
commit b0b280ea28
32 changed files with 999 additions and 219 deletions
Download Patch File Download Diff File Expand all files Collapse all files

41
gguf-py/gguf/constants.py
View File

@@ -358,6 +358,8 @@ class MODEL_ARCH(IntEnum):
DOTS1 = auto()
ARCEE = auto()
ERNIE4_5 = auto()
HUNYUAN_MOE = auto()
SMOLLM3 = auto()
class VISION_PROJECTOR_TYPE(IntEnum):
@@ -665,6 +667,8 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.DOTS1: "dots1",
MODEL_ARCH.ARCEE: "arcee",
MODEL_ARCH.ERNIE4_5: "ernie4_5",
MODEL_ARCH.HUNYUAN_MOE: "hunyuan-moe",
MODEL_ARCH.SMOLLM3: "smollm3",
}
VISION_PROJECTOR_TYPE_NAMES: dict[VISION_PROJECTOR_TYPE, str] = {
@@ -2247,6 +2251,43 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
],
MODEL_ARCH.HUNYUAN_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_GATE_INP,
MODEL_TENSOR.FFN_NORM,
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_ARCH.SMOLLM3: [
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_K,
MODEL_TENSOR.ATTN_V,
MODEL_TENSOR.ATTN_OUT,
MODEL_TENSOR.ATTN_ROT_EMBD,
MODEL_TENSOR.FFN_NORM,
MODEL_TENSOR.FFN_GATE,
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
],
# TODO
}

6
gguf-py/gguf/tensor_mapping.py
View File

@@ -306,6 +306,7 @@ class TensorNameMap:
"model.layers.{bid}.block_sparse_moe.router.layer", # granitemoe
"model.layers.{bid}.feed_forward.router", # llama4
"encoder.layers.{bid}.mlp.router.layer", # nomic-bert-moe
"model.layers.{bid}.mlp.gate.wg", # hunyuan
),
MODEL_TENSOR.FFN_GATE_INP_SHEXP: (
@@ -366,6 +367,7 @@ class TensorNameMap:
"model.layers.{bid}.mlp.shared_expert.up_proj", # qwen2moe
"model.layers.{bid}.mlp.shared_experts.up_proj", # deepseek deepseek2
"model.layers.{bid}.feed_forward.shared_expert.up_proj", # llama4
"model.layers.{bid}.mlp.shared_mlp.up_proj", # hunyuan
),
# AWQ-activation gate
@@ -403,6 +405,7 @@ class TensorNameMap:
"model.layers.{bid}.mlp.shared_expert.gate_proj", # qwen2moe
"model.layers.{bid}.mlp.shared_experts.gate_proj", # deepseek deepseek2
"model.layers.{bid}.feed_forward.shared_expert.gate_proj", # llama4
"model.layers.{bid}.mlp.shared_mlp.gate_proj", # hunyuan
),
# Feed-forward down
@@ -453,11 +456,13 @@ class TensorNameMap:
"model.layers.{bid}.mlp.shared_experts.down_proj", # deepseek deepseek2
"model.layers.{bid}.feed_forward.shared_expert.down_proj", # llama4
"model.layers.{bid}.shared_mlp.output_linear", # granitemoe
"model.layers.{bid}.mlp.shared_mlp.down_proj", # hunyuan
),
MODEL_TENSOR.ATTN_Q_NORM: (
"language_model.encoder.layers.{bid}.self_attention.q_layernorm",
"model.layers.{bid}.self_attn.q_layernorm", # persimmon
"model.layers.{bid}.self_attn.query_layernorm", # hunyuan
"model.layers.{bid}.self_attn.q_norm", # cohere olmoe chameleon olmo2
"transformer.blocks.{bid}.attn.q_ln", # sea-lion
"encoder.layer.{bid}.attention.self.layer_norm_q", # jina-bert-v2
@@ -467,6 +472,7 @@ class TensorNameMap:
MODEL_TENSOR.ATTN_K_NORM: (
"language_model.encoder.layers.{bid}.self_attention.k_layernorm",
"model.layers.{bid}.self_attn.k_layernorm", # persimmon
"model.layers.{bid}.self_attn.key_layernorm", # hunyuan
"model.layers.{bid}.self_attn.k_norm", # cohere olmoe chameleon olmo2
"transformer.blocks.{bid}.attn.k_ln", # sea-lion
"encoder.layer.{bid}.attention.self.layer_norm_k", # jina-bert-v2