Add support for BERT embedding models (ggerganov#5423)

* BERT model graph construction (build_bert)
* WordPiece tokenizer (llm_tokenize_wpm)
* Add flag for non-causal attention models
* Allow for models that only output embeddings
* Support conversion of BERT models to GGUF
* Based on prior work by @xyzhang626 and @skeskinen

---------

Co-authored-by: Jared Van Bortel <jared@nomic.ai>
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

.flake8

@@ -1,2 +1,3 @@
 [flake8]
 max-line-length = 125
+ignore = W503

convert-hf-to-gguf.py

@@ -209,6 +209,8 @@ def from_model_architecture(model_architecture):
             return InternLM2Model
         if model_architecture == "MiniCPMForCausalLM":
             return MiniCPMModel
+        if model_architecture == "BertModel":
+            return BertModel
         return Model
 
     def _is_model_safetensors(self) -> bool:
@@ -264,6 +266,8 @@ def _get_model_architecture(self) -> gguf.MODEL_ARCH:
             return gguf.MODEL_ARCH.INTERNLM2
         if arch == "MiniCPMForCausalLM":
             return gguf.MODEL_ARCH.MINICPM
+        if arch == "BertModel":
+            return gguf.MODEL_ARCH.BERT
 
         raise NotImplementedError(f'Architecture "{arch}" not supported!')
 
@@ -1629,6 +1633,96 @@ def write_tensors(self):
                 self.post_write_tensors(tensor_map, name, data_torch)
 
 
+class BertModel(Model):
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.block_count = self.hparams["num_hidden_layers"]
+
+    def set_gguf_parameters(self):
+        # TODO(cebtenzzre): merge with parent class
+        self.gguf_writer.add_name(self.dir_model.name)
+        self.gguf_writer.add_context_length(self.hparams["max_position_embeddings"])
+        self.gguf_writer.add_embedding_length(self.hparams["hidden_size"])
+        self.gguf_writer.add_feed_forward_length(self.hparams["intermediate_size"])
+        self.gguf_writer.add_block_count(self.block_count)
+        self.gguf_writer.add_head_count(self.hparams["num_attention_heads"])
+        self.gguf_writer.add_layer_norm_eps(self.hparams["layer_norm_eps"])
+        self.gguf_writer.add_causal_attention(False)
+        self.gguf_writer.add_file_type(self.ftype)
+
+    def set_vocab(self):
+        path = self.dir_model
+        added_tokens_path = self.dir_model if self.dir_model.exists() else None
+
+        # use huggingface vocab to get all tokens
+        vocab = HfVocab(path, added_tokens_path)
+        tokens, scores, toktypes = zip(*vocab.all_tokens())
+        assert len(tokens) == vocab.vocab_size
+
+        # we need this to validate the size of the token_type embeddings
+        # though currently we are passing all zeros to the token_type embeddings
+        n_token_types = len(set(toktypes))
+        self.gguf_writer.add_token_type_count(n_token_types)
+
+        # convert to phantom space vocab
+        def phantom(tok, typ):
+            if tok.startswith(b"[") and tok.endswith(b"]"):
+                return tok
+            if tok.startswith(b"##"):
+                return tok[2:]
+            return b"\xe2\x96\x81" + tok
+        tokens = [phantom(t, y) for t, y in zip(tokens, toktypes)]
+
+        # set up bos and eos tokens (cls and sep)
+        self.gguf_writer.add_bos_token_id(vocab.tokenizer.cls_token_id)
+        self.gguf_writer.add_eos_token_id(vocab.tokenizer.sep_token_id)
+
+        # add vocab to gguf
+        self.gguf_writer.add_tokenizer_model("bert")
+        self.gguf_writer.add_token_list(tokens)
+        self.gguf_writer.add_token_scores(scores)
+        self.gguf_writer.add_token_types(toktypes)
+
+        # handle special tokens
+        special_vocab = gguf.SpecialVocab(self.dir_model, n_vocab=len(tokens))
+        special_vocab.add_to_gguf(self.gguf_writer)
+
+    def write_tensors(self):
+        tensor_map = gguf.get_tensor_name_map(self.model_arch, self.block_count)
+        tensors = dict(self.get_tensors())
+        for name, data_torch in tensors.items():
+            # we are only using BERT for embeddings so we don't need the pooling layer
+            if name in ("embeddings.position_ids", "pooler.dense.weight", "pooler.dense.bias"):
+                continue  # we don't need these
+
+            # map tensor names
+            new_name = tensor_map.get_name(name, try_suffixes=(".weight", ".bias"))
+            if new_name is None:
+                print(f"Can not map tensor {name!r}")
+                sys.exit()
+
+            data = data_torch.squeeze().numpy()
+            n_dims = len(data.shape)
+            new_dtype: type[np.floating[Any]]
+
+            if (
+                self.ftype == 1 and name.endswith(".weight") and n_dims == 2
+                and name != "embeddings.token_type_embeddings.weight"  # not used with get_rows, must be F32
+            ):
+                # if f16 desired, convert any float32 2-dim weight tensors to float16
+                new_dtype = np.float16
+            else:
+                # if f32 desired, convert any float16 to float32
+                new_dtype = np.float32
+
+            print(f"{new_name}, n_dims = {n_dims}, {data_torch.dtype} --> {new_dtype}")
+
+            if data.dtype != new_dtype:
+                data = data.astype(new_dtype)
+
+            self.gguf_writer.add_tensor(new_name, data)
+
+
 ###### CONVERSION LOGIC ######
 
 

examples/embedding/embedding.cpp

@@ -87,7 +87,17 @@ int main(int argc, char ** argv) {
     }
 
     const int n_embd = llama_n_embd(model);
-    const auto * embeddings = llama_get_embeddings(ctx);
+    auto * embeddings = llama_get_embeddings(ctx);
+
+    // l2-normalize embeddings
+    float norm = 0;
+    for (int i = 0; i < n_embd; i++) {
+        norm += embeddings[i] * embeddings[i];
+    }
+    norm = sqrt(norm);
+    for (int i = 0; i < n_embd; i++) {
+        embeddings[i] /= norm;
+    }
 
     for (int i = 0; i < n_embd; i++) {
         printf("%f ", embeddings[i]);

gguf-py/gguf/constants.py

@@ -50,6 +50,7 @@ class Attention:
         VALUE_LENGTH      = "{arch}.attention.value_length"
         LAYERNORM_EPS     = "{arch}.attention.layer_norm_epsilon"
         LAYERNORM_RMS_EPS = "{arch}.attention.layer_norm_rms_epsilon"
+        CAUSAL            = "{arch}.attention.causal"
 
     class Rope:
         DIMENSION_COUNT      = "{arch}.rope.dimension_count"
@@ -60,22 +61,23 @@ class Rope:
         SCALING_FINETUNED    = "{arch}.rope.scaling.finetuned"
 
     class Tokenizer:
-        MODEL         = "tokenizer.ggml.model"
-        LIST          = "tokenizer.ggml.tokens"
-        TOKEN_TYPE    = "tokenizer.ggml.token_type"
-        SCORES        = "tokenizer.ggml.scores"
-        MERGES        = "tokenizer.ggml.merges"
-        BOS_ID        = "tokenizer.ggml.bos_token_id"
-        EOS_ID        = "tokenizer.ggml.eos_token_id"
-        UNK_ID        = "tokenizer.ggml.unknown_token_id"
-        SEP_ID        = "tokenizer.ggml.seperator_token_id"
-        PAD_ID        = "tokenizer.ggml.padding_token_id"
-        ADD_BOS       = "tokenizer.ggml.add_bos_token"
-        ADD_EOS       = "tokenizer.ggml.add_eos_token"
-        ADD_PREFIX    = "tokenizer.ggml.add_space_prefix"
-        HF_JSON       = "tokenizer.huggingface.json"
-        RWKV          = "tokenizer.rwkv.world"
-        CHAT_TEMPLATE = "tokenizer.chat_template"
+        MODEL            = "tokenizer.ggml.model"
+        LIST             = "tokenizer.ggml.tokens"
+        TOKEN_TYPE       = "tokenizer.ggml.token_type"
+        TOKEN_TYPE_COUNT = "tokenizer.ggml.token_type_count"  # for BERT-style token types
+        SCORES           = "tokenizer.ggml.scores"
+        MERGES           = "tokenizer.ggml.merges"
+        BOS_ID           = "tokenizer.ggml.bos_token_id"
+        EOS_ID           = "tokenizer.ggml.eos_token_id"
+        UNK_ID           = "tokenizer.ggml.unknown_token_id"
+        SEP_ID           = "tokenizer.ggml.seperator_token_id"
+        PAD_ID           = "tokenizer.ggml.padding_token_id"
+        ADD_BOS          = "tokenizer.ggml.add_bos_token"
+        ADD_EOS          = "tokenizer.ggml.add_eos_token"
+        ADD_PREFIX       = "tokenizer.ggml.add_space_prefix"
+        HF_JSON          = "tokenizer.huggingface.json"
+        RWKV             = "tokenizer.rwkv.world"
+        CHAT_TEMPLATE    = "tokenizer.chat_template"
 
 
 #
@@ -122,6 +124,7 @@ class MODEL_TENSOR(IntEnum):
     ATTN_OUT        = auto()
     ATTN_NORM       = auto()
     ATTN_NORM_2     = auto()
+    ATTN_OUT_NORM   = auto()
     ATTN_ROT_EMBD   = auto()
     FFN_GATE_INP    = auto()
     FFN_NORM        = auto()
@@ -134,6 +137,7 @@ class MODEL_TENSOR(IntEnum):
     FFN_UP_EXP      = auto()
     ATTN_Q_NORM     = auto()
     ATTN_K_NORM     = auto()
+    LAYER_OUT_NORM  = auto()
 
 
 MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
@@ -178,6 +182,7 @@ class MODEL_TENSOR(IntEnum):
     MODEL_TENSOR.ATTN_ROT_EMBD:   "blk.{bid}.attn_rot_embd",
     MODEL_TENSOR.ATTN_Q_NORM:     "blk.{bid}.attn_q_norm",
     MODEL_TENSOR.ATTN_K_NORM:     "blk.{bid}.attn_k_norm",
+    MODEL_TENSOR.ATTN_OUT_NORM:   "blk.{bid}.attn_output_norm",
     MODEL_TENSOR.FFN_GATE_INP:    "blk.{bid}.ffn_gate_inp",
     MODEL_TENSOR.FFN_NORM:        "blk.{bid}.ffn_norm",
     MODEL_TENSOR.FFN_GATE:        "blk.{bid}.ffn_gate",
@@ -187,6 +192,7 @@ class MODEL_TENSOR(IntEnum):
     MODEL_TENSOR.FFN_GATE_EXP:    "blk.{bid}.ffn_gate.{xid}",
     MODEL_TENSOR.FFN_DOWN_EXP:    "blk.{bid}.ffn_down.{xid}",
     MODEL_TENSOR.FFN_UP_EXP:      "blk.{bid}.ffn_up.{xid}",
+    MODEL_TENSOR.LAYER_OUT_NORM:  "blk.{bid}.layer_output_norm",
 }
 
 MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
@@ -262,17 +268,18 @@ class MODEL_TENSOR(IntEnum):
     ],
     MODEL_ARCH.BERT: [
         MODEL_TENSOR.TOKEN_EMBD,
+        MODEL_TENSOR.TOKEN_EMBD_NORM,
         MODEL_TENSOR.TOKEN_TYPES,
         MODEL_TENSOR.POS_EMBD,
         MODEL_TENSOR.OUTPUT_NORM,
-        MODEL_TENSOR.ATTN_NORM,
+        MODEL_TENSOR.ATTN_OUT_NORM,
         MODEL_TENSOR.ATTN_Q,
         MODEL_TENSOR.ATTN_K,
         MODEL_TENSOR.ATTN_V,
         MODEL_TENSOR.ATTN_OUT,
-        MODEL_TENSOR.FFN_NORM,
         MODEL_TENSOR.FFN_DOWN,
         MODEL_TENSOR.FFN_UP,
+        MODEL_TENSOR.LAYER_OUT_NORM,
     ],
     MODEL_ARCH.MPT: [
         MODEL_TENSOR.TOKEN_EMBD,

gguf-py/gguf/gguf_writer.py

@@ -357,6 +357,9 @@ def add_layer_norm_eps(self, value: float) -> None:
     def add_layer_norm_rms_eps(self, value: float) -> None:
         self.add_float32(Keys.Attention.LAYERNORM_RMS_EPS.format(arch=self.arch), value)
 
+    def add_causal_attention(self, value: bool) -> None:
+        self.add_bool(Keys.Attention.CAUSAL.format(arch=self.arch), value)
+
     def add_rope_dimension_count(self, count: int) -> None:
         self.add_uint32(Keys.Rope.DIMENSION_COUNT.format(arch=self.arch), count)
 
@@ -387,6 +390,9 @@ def add_token_merges(self, merges: Sequence[str] | Sequence[bytes] | Sequence[by
     def add_token_types(self, types: Sequence[TokenType] | Sequence[int]) -> None:
         self.add_array(Keys.Tokenizer.TOKEN_TYPE, types)
 
+    def add_token_type_count(self, value: int) -> None:
+        self.add_uint32(Keys.Tokenizer.TOKEN_TYPE_COUNT, value)
+
     def add_token_scores(self, scores: Sequence[float]) -> None:
         self.add_array(Keys.Tokenizer.SCORES, scores)
 

gguf-py/gguf/tensor_mapping.py

@@ -30,6 +30,7 @@ class TensorNameMap:
         # Normalization of token embeddings
         MODEL_TENSOR.TOKEN_EMBD_NORM: (
             "word_embeddings_layernorm",  # bloom
+            "embeddings.LayerNorm",       # bert
         ),
 
         # Position embeddings
@@ -54,7 +55,6 @@ class TensorNameMap:
             "transformer.ln_f",                        # gpt2 gpt-j falcon
             "model.norm",                              # llama-hf baichuan internlm2
             "norm",                                    # llama-pth
-            "embeddings.LayerNorm",                    # bert
             "transformer.norm_f",                      # mpt
             "ln_f",                                    # refact bloom qwen gpt2
             "language_model.encoder.final_layernorm",  # persimmon
@@ -79,7 +79,6 @@ class TensorNameMap:
             "transformer.h.{bid}.ln_mlp",                           # falcon40b
             "model.layers.{bid}.input_layernorm",                   # llama-hf
             "layers.{bid}.attention_norm",                          # llama-pth
-            "encoder.layer.{bid}.attention.output.LayerNorm",       # bert
             "language_model.encoder.layers.{bid}.input_layernorm",  # persimmon
             "model.layers.{bid}.ln1",                               # yi
             "h.{bid}.ln_1",                                         # gpt2
@@ -155,6 +154,11 @@ class TensorNameMap:
             "model.layers.{bid}.attention.wo",                           # internlm2
         ),
 
+        # Attention output norm
+        MODEL_TENSOR.ATTN_OUT_NORM: (
+            "encoder.layer.{bid}.attention.output.LayerNorm",  # bert
+        ),
+
         # Rotary embeddings
         MODEL_TENSOR.ATTN_ROT_EMBD: (
             "model.layers.{bid}.self_attn.rotary_emb.inv_freq",        # llama-hf
@@ -171,7 +175,6 @@ class TensorNameMap:
             "transformer.blocks.{bid}.norm_2",                               # mpt
             "model.layers.{bid}.post_attention_layernorm",                   # llama-hf
             "layers.{bid}.ffn_norm",                                         # llama-pth
-            "encoder.layer.{bid}.output.LayerNorm",                          # bert
             "language_model.encoder.layers.{bid}.post_attention_layernorm",  # persimmon
             "model.layers.{bid}.ln2",                                        # yi
             "h.{bid}.ln_2",                                                  # gpt2
@@ -266,6 +269,10 @@ class TensorNameMap:
         MODEL_TENSOR.ROPE_FREQS: (
             "language_model.encoder.layers.{bid}.self_attention.rotary_emb.inv_freq",  # persimmon
         ),
+
+        MODEL_TENSOR.LAYER_OUT_NORM: (
+            "encoder.layer.{bid}.output.LayerNorm",  # bert
+        )
     }
 
     mapping: dict[str, tuple[MODEL_TENSOR, str]]