UPSTREAM PR #18330: full modern bert support

gguf-py/gguf/constants.py

@@ -614,6 +614,7 @@ class MODEL_TENSOR(IntEnum):
     ENC_OUTPUT_NORM      = auto()
     CLS                  = auto() # classifier
     CLS_OUT              = auto() # classifier output projection
+    CLS_NORM             = auto()
     CONV1D               = auto()
     CONVNEXT_DW          = auto()
     CONVNEXT_NORM        = auto()
@@ -1007,6 +1008,7 @@ class MODEL_TENSOR(IntEnum):
     MODEL_TENSOR.ENC_OUTPUT_NORM:           "enc.output_norm",
     MODEL_TENSOR.CLS:                       "cls",
     MODEL_TENSOR.CLS_OUT:                   "cls.output",
+    MODEL_TENSOR.CLS_NORM:                  "cls.norm",
     MODEL_TENSOR.CONV1D:                    "conv1d",
     MODEL_TENSOR.CONVNEXT_DW:               "convnext.{bid}.dw",
     MODEL_TENSOR.CONVNEXT_NORM:             "convnext.{bid}.norm",
@@ -1382,6 +1384,7 @@ class MODEL_TENSOR(IntEnum):
         MODEL_TENSOR.FFN_NORM,
         MODEL_TENSOR.CLS,
         MODEL_TENSOR.CLS_OUT,
+        MODEL_TENSOR.CLS_NORM,
     ],
     MODEL_ARCH.NOMIC_BERT: [
         MODEL_TENSOR.TOKEN_EMBD,

gguf-py/gguf/tensor_mapping.py

@@ -1138,6 +1138,10 @@ class TensorNameMap:
         MODEL_TENSOR.CLS_OUT: (
             "classifier.out_proj", # roberta
         ),
+
+        MODEL_TENSOR.CLS_NORM: (
+            "head.norm", # modern-bert
+        ),
         #############################################################################
 
         MODEL_TENSOR.CONVNEXT_DW: (

src/llama-arch.cpp

@@ -348,6 +348,7 @@ static const std::map<llm_tensor, const char *> LLM_TENSOR_NAMES = {
     { LLM_TENSOR_TOKEN_TYPES,                            "token_types" },
     { LLM_TENSOR_CLS,                                    "cls" },
     { LLM_TENSOR_CLS_OUT,                                "cls.output" },
+    { LLM_TENSOR_CLS_NORM,                               "cls.norm" },
     { LLM_TENSOR_ENC_OUTPUT_NORM,                        "enc.output_norm" },
     { LLM_TENSOR_FFN_GATE_INP_SHEXP,                     "blk.%d.ffn_gate_inp_shexp" },
     { LLM_TENSOR_SSM_A_NOSCAN,                           "blk.%d.ssm_a" },
@@ -794,6 +795,7 @@ static std::set<llm_tensor> llm_get_tensor_names(llm_arch arch) {
                 LLM_TENSOR_FFN_NORM,
                 LLM_TENSOR_CLS,
                 LLM_TENSOR_CLS_OUT,
+                LLM_TENSOR_CLS_NORM,
             };
         case LLM_ARCH_JINA_BERT_V2:
             return {
@@ -2216,6 +2218,7 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
     {LLM_TENSOR_OUTPUT,                     {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}},
     {LLM_TENSOR_CLS,                        {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}},
     {LLM_TENSOR_CLS_OUT,                    {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}},
+    {LLM_TENSOR_CLS_NORM,                   {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL}},
     {LLM_TENSOR_DENSE_2_OUT,                {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}}, // Dense layer output
     {LLM_TENSOR_DENSE_3_OUT,                {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}}, // Dense layer output
     {LLM_TENSOR_OUTPUT_NORM,                {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL}},

src/llama-arch.h

@@ -468,6 +468,7 @@ enum llm_tensor {
     LLM_TENSOR_ENC_OUTPUT_NORM,
     LLM_TENSOR_CLS,
     LLM_TENSOR_CLS_OUT,
+    LLM_TENSOR_CLS_NORM,
     LLM_TENSOR_CONV1D,
     LLM_TENSOR_CONVNEXT_DW,
     LLM_TENSOR_CONVNEXT_NORM,

src/llama-context.cpp

@@ -2191,6 +2191,7 @@ void llama_context::opt_init(struct llama_model * model, struct llama_opt_params
     llama_set_param(model->cls_b,           param_filter, param_filter_ud);
     llama_set_param(model->cls_out,         param_filter, param_filter_ud);
     llama_set_param(model->cls_out_b,       param_filter, param_filter_ud);
+    llama_set_param(model->cls_norm,        param_filter, param_filter_ud);
 
     for (struct llama_layer & layer : model->layers) {
         for (size_t i = 0; i < sizeof(layer)/sizeof(struct ggml_tensor *); ++i) {

src/llama-graph.cpp

@@ -2006,7 +2006,8 @@ void llm_graph_context::build_pooling(
         ggml_tensor * cls,
         ggml_tensor * cls_b,
         ggml_tensor * cls_out,
-        ggml_tensor * cls_out_b) const {
+        ggml_tensor * cls_out_b,
+        ggml_tensor * cls_norm) const {
     if (!cparams.embeddings) {
         return;
     }
@@ -2056,6 +2057,10 @@ void llm_graph_context::build_pooling(
                         cur = ggml_add(ctx0, cur, cls_b);
                     }
                     cur = ggml_tanh(ctx0, cur);
+                    if (cls_norm) {
+                        // head norm
+                        cur = build_norm(cur, cls_norm, NULL, LLM_NORM, -1);
+                    }
                 }
 
                 // some models don't have `cls_out`, for example: https://huggingface.co/jinaai/jina-reranker-v1-tiny-en

src/llama-graph.h

@@ -830,7 +830,8 @@ struct llm_graph_context {
             ggml_tensor * cls,
             ggml_tensor * cls_b,
             ggml_tensor * cls_out,
-            ggml_tensor * cls_out_b) const;
+            ggml_tensor * cls_out_b,
+            ggml_tensor * cls_norm) const;
 
     //
     // dense (out)

src/llama-model-saver.cpp

@@ -268,6 +268,7 @@ void llama_model_saver::add_tensors_from_model() {
     add_tensor(model.cls_b);
     add_tensor(model.cls_out);
     add_tensor(model.cls_out_b);
+    add_tensor(model.cls_norm);
 
     for (const struct llama_layer & layer : model.layers) {
         for (size_t i = 0; i < sizeof(layer)/sizeof(struct ggml_tensor *); ++i) {

src/llama-model.cpp

@@ -3212,9 +3212,10 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
                         layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
                     }
 
-                    cls       = create_tensor(tn(LLM_TENSOR_CLS,     "weight"), {n_embd, n_embd}, TENSOR_NOT_REQUIRED);
-                    cls_out   = create_tensor(tn(LLM_TENSOR_CLS_OUT, "weight"), {n_embd, hparams.n_cls_out}, TENSOR_NOT_REQUIRED);
-                    cls_out_b = create_tensor(tn(LLM_TENSOR_CLS_OUT, "bias"),   {hparams.n_cls_out},         TENSOR_NOT_REQUIRED);
+                    cls       = create_tensor(tn(LLM_TENSOR_CLS,      "weight"), {n_embd, n_embd},            TENSOR_NOT_REQUIRED);
+                    cls_out   = create_tensor(tn(LLM_TENSOR_CLS_OUT,  "weight"), {n_embd, hparams.n_cls_out}, TENSOR_NOT_REQUIRED);
+                    cls_out_b = create_tensor(tn(LLM_TENSOR_CLS_OUT,  "bias"),   {hparams.n_cls_out},         TENSOR_NOT_REQUIRED);
+                    cls_norm  = create_tensor(tn(LLM_TENSOR_CLS_NORM, "weight"), {n_embd},                    TENSOR_NOT_REQUIRED);
 
                 } break;
             case LLM_ARCH_NEO_BERT:
@@ -7709,7 +7710,7 @@ ggml_cgraph * llama_model::build_graph(const llm_graph_params & params) const {
     }
 
     // add on pooling layer
-    llm->build_pooling(cls, cls_b, cls_out, cls_out_b);
+    llm->build_pooling(cls, cls_b, cls_out, cls_out_b, cls_norm);
 
     // if the gguf model was converted with --sentence-transformers-dense-modules
     // there will be two additional dense projection layers

src/llama-model.h

@@ -446,6 +446,7 @@ struct llama_model {
     struct ggml_tensor * cls_b     = nullptr;
     struct ggml_tensor * cls_out   = nullptr;
     struct ggml_tensor * cls_out_b = nullptr;
+    struct ggml_tensor * cls_norm  = nullptr;
 
     struct ggml_tensor * conv1d   = nullptr;
     struct ggml_tensor * conv1d_b = nullptr;