model : jina-embeddings-v3 support

convert_hf_to_gguf.py

@@ -3623,44 +3623,93 @@ def _xlmroberta_set_vocab(self) -> None:
         from sentencepiece import sentencepiece_model_pb2 as model
 
         tokenizer_path = self.dir_model / 'sentencepiece.bpe.model'
+
+        tokenizer_json = {}
+        tokenizer_config_json = {}
         if not tokenizer_path.is_file():
-            raise FileNotFoundError(f"File not found: {tokenizer_path}")
+            tokenizer_path = self.dir_model / 'tokenizer.json'
+            tokenizer_config_path = self.dir_model / 'tokenizer_config.json'
 
-        sentencepiece_model = model.ModelProto()  # pyright: ignore[reportAttributeAccessIssue]
-        sentencepiece_model.ParseFromString(open(tokenizer_path, "rb").read())
-        assert sentencepiece_model.trainer_spec.model_type == 1  # UNIGRAM
+            if not tokenizer_path.is_file():
+                raise FileNotFoundError(f"File not found: {tokenizer_path}")
 
-        add_prefix = sentencepiece_model.normalizer_spec.add_dummy_prefix
-        remove_whitespaces = sentencepiece_model.normalizer_spec.remove_extra_whitespaces
-        precompiled_charsmap = sentencepiece_model.normalizer_spec.precompiled_charsmap
+            from base64 import b64decode
+            from transformers import AutoTokenizer
+            tokenizer = AutoTokenizer.from_pretrained(self.dir_model)
 
-        tokenizer = SentencePieceProcessor()
-        tokenizer.LoadFromFile(str(tokenizer_path))
+            with open(tokenizer_path, "r", encoding="utf-8") as fp:
+                tokenizer_json = json.load(fp)
 
-        vocab_size = self.hparams.get('vocab_size', tokenizer.vocab_size())
+            if tokenizer_config_path.is_file():
+                with open(tokenizer_config_path, "r", encoding="utf-8") as fp:
+                    tokenizer_config_json = json.load(fp)
+
+            add_prefix = tokenizer.add_prefix_space
+            remove_whitespaces = tokenizer.clean_up_tokenization_spaces
+            precompiled_charsmap = b64decode(tokenizer_json["normalizer"]["precompiled_charsmap"])
+
+            vocab_size = self.hparams.get("vocab_size", tokenizer.vocab_size)
+        else:
+            sentencepiece_model = model.ModelProto()  # pyright: ignore[reportAttributeAccessIssue]
+            sentencepiece_model.ParseFromString(open(tokenizer_path, "rb").read())
+            assert sentencepiece_model.trainer_spec.model_type == 1  # UNIGRAM
+
+            add_prefix = sentencepiece_model.normalizer_spec.add_dummy_prefix
+            remove_whitespaces = sentencepiece_model.normalizer_spec.remove_extra_whitespaces
+            precompiled_charsmap = sentencepiece_model.normalizer_spec.precompiled_charsmap
+
+            tokenizer = SentencePieceProcessor()
+            tokenizer.LoadFromFile(str(tokenizer_path))
+
+            vocab_size = self.hparams.get('vocab_size', tokenizer.vocab_size())
 
         tokens: list[bytes] = [f"[PAD{i}]".encode("utf-8") for i in range(vocab_size)]
         scores: list[float] = [-10000.0] * vocab_size
         toktypes: list[int] = [SentencePieceTokenTypes.UNUSED] * vocab_size
 
-        for token_id in range(tokenizer.vocab_size()):
-            piece = tokenizer.IdToPiece(token_id)
-            text = piece.encode("utf-8")
-            score = tokenizer.GetScore(token_id)
+        if isinstance(tokenizer, SentencePieceProcessor):
+            for token_id in range(tokenizer.vocab_size()):
+                piece = tokenizer.IdToPiece(token_id)
+                text = piece.encode("utf-8")
+                score = tokenizer.GetScore(token_id)
 
-            toktype = SentencePieceTokenTypes.NORMAL
-            if tokenizer.IsUnknown(token_id):
-                toktype = SentencePieceTokenTypes.UNKNOWN
-            elif tokenizer.IsControl(token_id):
-                toktype = SentencePieceTokenTypes.CONTROL
-            elif tokenizer.IsUnused(token_id):
-                toktype = SentencePieceTokenTypes.UNUSED
-            elif tokenizer.IsByte(token_id):
-                toktype = SentencePieceTokenTypes.BYTE
+                toktype = SentencePieceTokenTypes.NORMAL
+                if tokenizer.IsUnknown(token_id):
+                    toktype = SentencePieceTokenTypes.UNKNOWN
+                elif tokenizer.IsControl(token_id):
+                    toktype = SentencePieceTokenTypes.CONTROL
+                elif tokenizer.IsUnused(token_id):
+                    toktype = SentencePieceTokenTypes.UNUSED
+                elif tokenizer.IsByte(token_id):
+                    toktype = SentencePieceTokenTypes.BYTE
 
-            tokens[token_id] = text
-            scores[token_id] = score
-            toktypes[token_id] = toktype
+                tokens[token_id] = text
+                scores[token_id] = score
+                toktypes[token_id] = toktype
+        else:
+            added_vocab = tokenizer.get_added_vocab()
+            unk_token = tokenizer_config_json.get("unk_token")
+            unk_token_id = added_vocab.get(unk_token, tokenizer_json["model"].get("unk_id", 3))
+
+            for token_id in range(vocab_size):
+                piece = tokenizer._convert_id_to_token(token_id)
+                text = piece.encode("utf-8")
+                score = tokenizer_json["model"]["vocab"][token_id][1]
+
+                toktype = SentencePieceTokenTypes.NORMAL
+                if token_id == unk_token_id:
+                    toktype = SentencePieceTokenTypes.UNKNOWN
+                elif token_id in tokenizer.all_special_ids:
+                    toktype = SentencePieceTokenTypes.CONTROL
+                elif token_id in added_vocab.values():
+                    toktype = SentencePieceTokenTypes.USER_DEFINED
+                # No reliable way to detect this, but jina-embeddings-v3 doesn't have any
+                # elif tokenizer.IsByte(token_id):
+                #    toktype = SentencePieceTokenTypes.BYTE
+
+                tokens[token_id] = text
+                scores[token_id] = score
+                toktypes[token_id] = toktype
 
         if vocab_size > len(tokens):
             pad_count = vocab_size - len(tokens)
@@ -3670,15 +3719,16 @@ def _xlmroberta_set_vocab(self) -> None:
                 scores.append(-1000.0)
                 toktypes.append(SentencePieceTokenTypes.UNUSED)
 
-        # realign tokens (see HF tokenizer code)
-        tokens = [b'<s>', b'<pad>', b'</s>', b'<unk>'] + tokens[3:-1]
-        scores = [0.0, 0.0, 0.0, 0.0] + scores[3:-1]
-        toktypes = [
-            SentencePieceTokenTypes.CONTROL,
-            SentencePieceTokenTypes.CONTROL,
-            SentencePieceTokenTypes.CONTROL,
-            SentencePieceTokenTypes.UNKNOWN,
-        ] + toktypes[3:-1]
+        if isinstance(tokenizer, SentencePieceProcessor):
+            # realign tokens (see HF tokenizer code)
+            tokens = [b'<s>', b'<pad>', b'</s>', b'<unk>'] + tokens[3:-1]
+            scores = [0.0, 0.0, 0.0, 0.0] + scores[3:-1]
+            toktypes = [
+                SentencePieceTokenTypes.CONTROL,
+                SentencePieceTokenTypes.CONTROL,
+                SentencePieceTokenTypes.CONTROL,
+                SentencePieceTokenTypes.UNKNOWN,
+            ] + toktypes[3:-1]
 
         self.gguf_writer.add_tokenizer_model("t5")
         self.gguf_writer.add_tokenizer_pre("default")
@@ -3829,8 +3879,15 @@ 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):
@@ -3842,13 +3899,34 @@ 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):

gguf-py/gguf/constants.py

@@ -272,6 +272,7 @@ 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()
@@ -534,6 +535,7 @@ 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",
@@ -1020,6 +1022,18 @@ 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,

gguf-py/gguf/tensor_mapping.py

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

src/llama-arch.cpp

@@ -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"             },
@@ -513,6 +514,20 @@ 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,
         {

src/llama-arch.h

@@ -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,

src/llama-model.cpp

@@ -41,6 +41,7 @@ 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";
@@ -710,6 +711,18 @@ 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:
             {
@@ -2215,6 +2228,36 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
                         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}, 0);
 
+                        layer.layer_out_norm   = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd}, 0);
+                        layer.layer_out_norm_b = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "bias",   i), {n_embd}, 0);
+                    }
+                } break;
+            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}, 0);
+
+                    tok_norm   = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight"), {n_embd}, 0);
+                    tok_norm_b = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "bias"),   {n_embd}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+                        layer.bqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias",   i), {n_embd + 2*n_embd_gqa}, 0);
+
+                        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}, 0);
+
+                        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);
+
+                        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}, 0);
+
+                        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}, 0);
+
                         layer.layer_out_norm   = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd}, 0);
                         layer.layer_out_norm_b = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "bias",   i), {n_embd}, 0);
                     }
@@ -5931,7 +5974,7 @@ struct llm_build_bert : public llm_graph_context {
                 cur = build_lora_mm(model.layers[il].wqkv, cur);
                 cb(cur, "wqkv", il);
 
-                if (model.arch == LLM_ARCH_NOMIC_BERT_MOE) {
+                if (model.arch == LLM_ARCH_NOMIC_BERT_MOE || model.arch == LLM_ARCH_JINA_BERT_V3) {
                     cur = ggml_add(ctx0, cur, model.layers[il].bqkv);
                     cb(cur, "bqkv", il);
                 }
@@ -6003,7 +6046,7 @@ 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,
@@ -13187,6 +13230,7 @@ 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:
             {
@@ -13292,6 +13336,7 @@ 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:
             {
@@ -13658,6 +13703,7 @@ 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: