8000 model : jina-embeddings-v3 support by CISC · Pull Request #13693 · ggml-org/llama.cpp · GitHub
[go: up one dir, main page]
Skip to content

model : jina-embeddings-v3 support #13693

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
wants to merge 10 commits into
base: master
Choose a base branch
from
Draft

model : jina-embeddings-v3 support #13693

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
6303ea2
initial jina-embeddings-v3 support
CISC May 21, 2025
ba51f89
initial jina-embeddings-v3 support
CISC May 21, 2025
4ac1380
initial jina-embeddings-v3 support
CISC May 21, 2025
1274c8c
fix vocab parsing with only tokenizer.json
CISC May 22, 2025
65a37fa
set mask token lstrip attribute
CISC May 22, 2025
f2d876a
additional unk_token_id fallback just in case [no ci]
CISC May 23, 2025
b17e981
revert vocab_size() change [no ci]
CISC May 26, 2025
4046701
Merge branch 'master' into cisc/jina-embeddings-v3
CISC Jun 1, 2025
f5d0305
merge tensor loading into general bert
CISC Jun 1, 2025
3862d95
rope
CISC Jun 1, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
32 changes: 30 additions & 2 deletions convert_hf_to_gguf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4064,8 +4064,15 @@ def _is_tokenizer_xlmroberta(self) -> bool:
class XLMRobertaModel(BertModel):
model_arch = gguf.MODEL_ARCH.BERT

def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
def __init__(self, dir_model: Path, ftype: gguf.LlamaFileType, fname_out: Path, **kwargs: Any):
hparams = kwargs.pop("hparams", None)
if hparams is None:
hparams = ModelBase.load_hparams(dir_model)

if hparams.get("lora_adaptations"):
self.model_arch = gguf.MODEL_ARCH.JINA_BERT_V3

super().__init__(dir_model, ftype, fname_out, hparams=hparams, **kwargs)
self._xlmroberta_tokenizer_init()

def set_vocab(self):
Expand All @@ -4077,13 +4084,34 @@ def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iter
if name.startswith("roberta."):
name = name[8:]

# jina-embeddings-v3
if ".parametrizations." in name:
name = name.replace(".parametrizations.", ".")
if name.endswith(".original"):
name = name[:-9]

# position embeddings start at pad_token_id + 1, so just chop down the weight tensor
if name == "embeddings.position_embeddings.weight":
if self._position_offset is not None:
data_torch = data_torch[self._position_offset:,:]

if name.endswith(".lora_A"):
# TODO: convert loras
return []

if name.endswith(".lora_B"):
# TODO: convert loras
return []

return super().modify_tensors(data_torch, name, bid)

def set_gguf_parameters(self):
super().set_gguf_parameters()

# jina-embeddings-v3
if rotary_emb_base := self.hparams.get("rotary_emb_base"):
self.gguf_writer.add_rope_freq_base(rotary_emb_base)


@ModelBase.register("GemmaForCausalLM")
class GemmaModel(TextModel):
Expand Down
14 changes: 14 additions & 0 deletions gguf-py/gguf/constants.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -292,6 +292,7 @@ class MODEL_ARCH(IntEnum):
NOMIC_BERT = auto()
NOMIC_BERT_MOE = auto()
JINA_BERT_V2 = auto()
JINA_BERT_V3 = auto()
BLOOM = auto()
STABLELM = auto()
QWEN = auto()
Expand Down Expand Up @@ -572,6 +573,7 @@ class MODEL_TENSOR(IntEnum):
MODEL_ARCH.NOMIC_BERT: "nomic-bert",
MODEL_ARCH.NOMIC_BERT_MOE: "nomic-bert-moe",
MODEL_ARCH.JINA_BERT_V2: "jina-bert-v2",
MODEL_ARCH.JINA_BERT_V3: "jina-bert-v3",
MODEL_ARCH.BLOOM: "bloom",
MODEL_ARCH.STABLELM: "stablelm",
MODEL_ARCH.QWEN: "qwen",
Expand Down Expand Up @@ -1095,6 +1097,18 @@ class MODEL_TENSOR(IntEnum):
MODEL_TENSOR.LAYER_OUT_NORM,
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_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,
Expand Down
15 changes: 15 additions & 0 deletions src/llama-arch.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
{ LLM_ARCH_NOMIC_BERT, "nomic-bert" },
{ LLM_ARCH_NOMIC_BERT_MOE, "nomic-bert-moe" },
{ LLM_ARCH_JINA_BERT_V2, "jina-bert-v2" },
{ LLM_ARCH_JINA_BERT_V3, "jina-bert-v3" },
{ LLM_ARCH_BLOOM, "bloom" },
{ LLM_ARCH_STABLELM, "stablelm" },
{ LLM_ARCH_QWEN, "qwen" },
Expand Down Expand Up @@ -516,6 +517,20 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
{ LLM_TENSOR_CLS, "cls" },
},
},
{
LLM_ARCH_JINA_BERT_V3,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_TOKEN_EMBD_NORM, "token_embd_norm" },
{ LLM_TENSOR_TOKEN_TYPES, "token_types" },
{ LLM_TENSOR_ATTN_OUT_NORM, "blk.%d.attn_output_norm" },
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
{ LLM_TENSOR_LAYER_OUT_NORM, "blk.%d.layer_output_norm" },
},
},
{
LLM_ARCH_BLOOM,
{
Expand Down
1 change: 1 addition & 0 deletions src/llama-arch.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,7 @@ enum llm_arch {
LLM_ARCH_NOMIC_BERT,
LLM_ARCH_NOMIC_BERT_MOE,
LLM_ARCH_JINA_BERT_V2,
LLM_ARCH_JINA_BERT_V3,
LLM_ARCH_BLOOM,
LLM_ARCH_STABLELM,
LLM_ARCH_QWEN,
Expand Down
37 changes: 26 additions & 11 deletions src/llama-model.cpp
View file
Open in desktop
F438
Original file line number Diff line number Diff line change
Expand Up @@ -44,6 +44,7 @@ const char * llm_type_name(llm_type type) {
case LLM_TYPE_410M: return "410M";
case LLM_TYPE_450M: return "450M";
case LLM_TYPE_475M: return "475M";
case LLM_TYPE_558M: return "558M";
case LLM_TYPE_770M: return "770M";
case LLM_TYPE_780M: return "780M";
case LLM_TYPE_0_5B: return "0.5B";
Expand Down Expand Up @@ -717,6 +718,18 @@ void llama_model::load_hparams(llama_model_loader & ml) {
default: type = LLM_TYPE_UNKNOWN;
}
} break;
case LLM_ARCH_JINA_BERT_V3:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
ml.get_key(LLM_KV_ATTENTION_CAUSAL, hparams.causal_attn);
ml.get_key(LLM_KV_POOLING_TYPE, hparams.pooling_type, false);

switch (hparams.n_layer) {
case 24:
type = LLM_TYPE_558M; break;
default: type = LLM_TYPE_UNKNOWN;
}
} break;
case LLM_ARCH_NOMIC_BERT:
case LLM_ARCH_NOMIC_BERT_MOE:
{
Expand Down Expand Up @@ -2115,6 +2128,7 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
case LLM_ARCH_BERT:
case LLM_ARCH_NOMIC_BERT:
case LLM_ARCH_NOMIC_BERT_MOE:
case LLM_ARCH_JINA_BERT_V3:
{
tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
type_embd = create_tensor(tn(LLM_TENSOR_TOKEN_TYPES, "weight"), {n_embd, n_token_types}, TENSOR_NOT_REQUIRED);
Expand Down Expand Up @@ -2150,24 +2164,22 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
}

layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
layer.bo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);

layer.attn_out_norm = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd}, 0);
layer.attn_out_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "bias", i), {n_embd}, 0);

if (hparams.moe_every_n_layers > 0 && i % hparams.moe_every_n_layers == 1) {
layer.bo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, 0);
layer.ffn_up_exps = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i), { n_embd, n_ff, n_expert}, 0);
layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), { n_ff, n_embd, n_expert}, 0);
layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
} else {
layer.ffn_up = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}, 0);
layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);

if (arch == LLM_ARCH_BERT || arch == LLM_ARCH_NOMIC_BERT_MOE) {
layer.bo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, 0);
layer.ffn_up_b = create_tensor(tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff}, 0);
layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, 0);
} else {
layer.ffn_up = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}, 0);
layer.ffn_up_b = create_tensor(tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);

if (arch == LLM_ARCH_NOMIC_BERT) {
layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
}
}
Expand Down Expand Up @@ -5949,7 +5961,7 @@ struct llm_build_bert : public llm_graph_context {
Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head, n_head_kv, n_tokens);

// RoPE
if (model.arch == LLM_ARCH_NOMIC_BERT || model.arch == LLM_ARCH_NOMIC_BERT_MOE) {
if (model.arch == LLM_ARCH_NOMIC_BERT || model.arch == LLM_ARCH_NOMIC_BERT_MOE || model.arch == LLM_ARCH_JINA_BERT_V3) {
Qcur = ggml_rope_ext(
ctx0, Qcur, inp_pos, nullptr,
n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
Expand Down Expand Up @@ -6009,7 +6021,7 @@ struct llm_build_bert : public llm_graph_context {
0.0f,
LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX, il);
cb(cur, "ffn_moe_out", il);
} else if (model.arch == LLM_ARCH_BERT || model.arch == LLM_ARCH_NOMIC_BERT_MOE) {
} else if (model.arch == LLM_ARCH_BERT || model.arch == LLM_ARCH_NOMIC_BERT_MOE || model.arch == LLM_ARCH_JINA_BERT_V3) {
cur = build_ffn(cur,
model.layers[il].ffn_up, model.layers[il].ffn_up_b, NULL,
NULL, NULL, NULL,
Expand Down Expand Up @@ -13193,6 +13205,7 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
switch (arch) {
case LLM_ARCH_BERT:
case LLM_ARCH_JINA_BERT_V2:
case LLM_ARCH_JINA_BERT_V3:
case LLM_ARCH_NOMIC_BERT:
case LLM_ARCH_NOMIC_BERT_MOE:
case LLM_ARCH_WAVTOKENIZER_DEC:
Expand Down Expand Up @@ -13298,6 +13311,7 @@ llm_graph_result_ptr llama_model::build_graph(
} break;
case LLM_ARCH_BERT:
case LLM_ARCH_JINA_BERT_V2:
case LLM_ARCH_JINA_BERT_V3:
case LLM_ARCH_NOMIC_BERT:
case LLM_ARCH_NOMIC_BERT_MOE:
{
Expand Down Expand Up @@ -13669,6 +13683,7 @@ llama_rope_type llama_model_rope_type(const llama_model * model) {
case LLM_ARCH_GROK:
case LLM_ARCH_DBRX:
case LLM_ARCH_BERT:
case LLM_ARCH_JINA_BERT_V3:
case LLM_ARCH_NOMIC_BERT:
case LLM_ARCH_NOMIC_BERT_MOE:
case LLM_ARCH_STABLELM:
Expand Down
1 change: 1 addition & 0 deletions src/llama-model.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,6 +37,7 @@ enum llm_type {
LLM_TYPE_410M,
LLM_TYPE_450M,
LLM_TYPE_475M,
LLM_TYPE_558M,
LLM_TYPE_770M,
LLM_TYPE_780M,
LLM_TYPE_0_5B,
Expand Down
2 changes: 1 addition & 1 deletion src/llama-vocab.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2096,7 +2096,7 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
// set attributes by model/tokenizer/architecture name
if (false
|| _contains_any(tokenizer_pre, {"jina-v2-de", "jina-v2-es", "jina-v2-code"})
|| _contains_any(general_arch, {"nomic-bert-moe"})
|| _contains_any(general_arch, {"nomic-bert-moe", "jina-bert-v3"})
) {
_set_token_attr("<mask>", LLAMA_TOKEN_ATTR_LSTRIP, true);
} else if (_contains_any(model_name, {"phi-3", "phi3"})) {
Expand Down
Loading
0