tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-06 01:05:03 -04:00
Code Issues Packages Projects Releases Wiki Activity

initial jina-embeddings-v3 support

Browse Source
This commit is contained in:
Sigbjørn Skjæret
2025-05-21 20:17:35 +02:00
committed by GitHub
parent 8e186ef0e7
commit 6303ea2640
2 changed files with 16 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
gguf-py/gguf/constants.py
View File

@@ -272,6 +272,7 @@ class MODEL_ARCH(IntEnum):
NOMIC_BERT = auto() NOMIC_BERT = auto()
NOMIC_BERT_MOE = auto() NOMIC_BERT_MOE = auto()
JINA_BERT_V2 = auto() JINA_BERT_V2 = auto()
JINA_BERT_V3 = auto()
BLOOM = auto() BLOOM = auto()
STABLELM = auto() STABLELM = auto()
QWEN = auto() QWEN = auto()
@@ -534,6 +535,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.NOMIC_BERT: "nomic-bert", MODEL_ARCH.NOMIC_BERT: "nomic-bert",
MODEL_ARCH.NOMIC_BERT_MOE: "nomic-bert-moe", MODEL_ARCH.NOMIC_BERT_MOE: "nomic-bert-moe",
MODEL_ARCH.JINA_BERT_V2: "jina-bert-v2", MODEL_ARCH.JINA_BERT_V2: "jina-bert-v2",
MODEL_ARCH.JINA_BERT_V3: "jina-bert-v3",
MODEL_ARCH.BLOOM: "bloom", MODEL_ARCH.BLOOM: "bloom",
MODEL_ARCH.STABLELM: "stablelm", MODEL_ARCH.STABLELM: "stablelm",
MODEL_ARCH.QWEN: "qwen", MODEL_ARCH.QWEN: "qwen",
@@ -1020,6 +1022,18 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.LAYER_OUT_NORM, MODEL_TENSOR.LAYER_OUT_NORM,
MODEL_TENSOR.CLS, MODEL_TENSOR.CLS,
], ],
MODEL_ARCH.JINA_BERT_V3: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.TOKEN_EMBD_NORM,
MODEL_TENSOR.TOKEN_TYPES,
MODEL_TENSOR.OUTPUT_NORM,
MODEL_TENSOR.ATTN_OUT_NORM,
MODEL_TENSOR.ATTN_QKV,
MODEL_TENSOR.ATTN_OUT,
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
MODEL_TENSOR.LAYER_OUT_NORM,
],
MODEL_ARCH.MPT: [ MODEL_ARCH.MPT: [
MODEL_TENSOR.TOKEN_EMBD, MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM, MODEL_TENSOR.OUTPUT_NORM,

2
gguf-py/gguf/tensor_mapping.py
View File

@@ -157,6 +157,7 @@ class TensorNameMap:
"h.{bid}.attn.c_attn", # gpt2 "h.{bid}.attn.c_attn", # gpt2
"transformer.h.{bid}.mixer.Wqkv", # phi2 "transformer.h.{bid}.mixer.Wqkv", # phi2
"encoder.layers.{bid}.attn.Wqkv", # nomic-bert "encoder.layers.{bid}.attn.Wqkv", # nomic-bert
"encoder.layers.{bid}.mixer.Wqkv", # jina-bert-v3
"model.layers.{bid}.self_attn.qkv_proj", # phi3 "model.layers.{bid}.self_attn.qkv_proj", # phi3
"encoder.layers.{bid}.self_attention.query_key_value", # chatglm "encoder.layers.{bid}.self_attention.query_key_value", # chatglm
"transformer.layers.{bid}.attn.qkv_proj", # openelm "transformer.layers.{bid}.attn.qkv_proj", # openelm
@@ -224,6 +225,7 @@ class TensorNameMap:
"model.layers.layers.{bid}.self_attn.o_proj", # plamo "model.layers.layers.{bid}.self_attn.o_proj", # plamo
"model.layers.{bid}.attention.wo", # internlm2 "model.layers.{bid}.attention.wo", # internlm2
"encoder.layers.{bid}.attn.out_proj", # nomic-bert "encoder.layers.{bid}.attn.out_proj", # nomic-bert
"encoder.layers.{bid}.mixer.out_proj", # jina-bert-v3
"transformer.decoder_layer.{bid}.multi_head_attention.linear", # Grok "transformer.decoder_layer.{bid}.multi_head_attention.linear", # Grok
"transformer.blocks.{bid}.norm_attn_norm.attn.out_proj", # dbrx "transformer.blocks.{bid}.norm_attn_norm.attn.out_proj", # dbrx
"encoder.layers.{bid}.self_attention.dense", # chatglm "encoder.layers.{bid}.self_attention.dense", # chatglm