#include "models.h" // question types of a decision model: choice, score, noul static const uint32_t N_DECISION_TYPES = 3; void llama_model_modern_bert::load_arch_hparams(llama_model_loader & ml) { const bool found_swa = ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa, false); if (found_swa && hparams.n_swa > 0) { hparams.swa_type = LLAMA_SWA_TYPE_SYMMETRIC; ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false); load_swa_pattern(ml, 3, true); } else { hparams.swa_type = LLAMA_SWA_TYPE_NONE; } ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); // Some ModernBert derivatives (e.g. IBM Granite Embedding 97m R2) use // SiLU/SwiGLU in the FFN instead of the default GELU/GeGLU. hparams.llm_ffn_op = LLM_FFN_GEGLU_ERF; std::string hidden_act; if (ml.get_key(LLM_KV_HIDDEN_ACT, hidden_act, false)) { hparams.llm_ffn_op = llm_ffn_op_type_from_string(hidden_act, LLM_FFN_GEGLU_ERF); } // GGUFs without a classifier pooling type use mean (gte-reranker-modernbert-base) if (hparams.pooling_type_cls == LLAMA_POOLING_TYPE_UNSPECIFIED) { hparams.pooling_type_cls = LLAMA_POOLING_TYPE_MEAN; } // GGUFs without a classifier activation use gelu, the transformers default std::string act_cls; if (!ml.get_key(LLM_KV_CLASSIFIER_ACTIVATION, act_cls, false)) { hparams.act_cls = GGML_UNARY_OP_GELU_ERF; } ml.get_key(LLM_KV_DECISION_BLOCK_COUNT, hparams.n_layer_decision, false); if (hparams.n_layer_decision > 0) { if (hparams.n_layer_decision >= hparams.n_layer()) { throw std::runtime_error("invalid number of decision blocks"); } // the head blocks always use full attention for (uint32_t il = hparams.n_layer() - hparams.n_layer_decision; il < hparams.n_layer(); il++) { hparams.is_swa_impl[il] = false; } // the output is one score per question type hparams.n_embd_out_impl = N_DECISION_TYPES; } switch (hparams.n_layer()) { case 12: type = LLM_TYPE_47M; break; // granite-embedding-small case 22: type = LLM_TYPE_149M; break; // modern-bert-base case 28: type = LLM_TYPE_395M; break; // modern-bert-large default: type = LLM_TYPE_UNKNOWN; } } void llama_model_modern_bert::load_arch_tensors(llama_model_loader &) { LLAMA_LOAD_LOCALS; tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0); tok_norm = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight", 0), {n_embd}, 0); output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0); const int n_layer_enc = n_layer - hparams.n_layer_decision; for(int i = 0; i < n_layer_enc; ++i) { auto& layer = layers[i]; if ( i != 0 ) { layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0); } else{ // layer 0 uses identity layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED); } layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, 3 * n_embd }, 0); layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0); layer.ffn_up = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, 2 * n_ff}, 0); layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0); layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0); } 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_norm = create_tensor(tn(LLM_TENSOR_CLS_NORM, "weight"), {n_embd}, TENSOR_NOT_REQUIRED); if (hparams.n_layer_decision == 0) { return; } // decision head: plain pre-norm blocks with biases for (int i = n_layer_enc; i < n_layer; ++i) { auto & layer = layers[i]; const int64_t n_ff_head = hparams.n_ff(i); layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0); layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd}, 0); layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, 3 * n_embd}, 0); layer.wqkv_b = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias", i), {3 * n_embd}, 0); layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0); layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, 0); layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0); layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd}, 0); layer.ffn_up = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff_head}, 0); layer.ffn_up_b = create_tensor(tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff_head}, 0); layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff_head, n_embd}, 0); layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, 0); } if (n_token_types != N_DECISION_TYPES) { throw std::runtime_error("decision model must have one token type per question type"); } type_embd = create_tensor(tn(LLM_TENSOR_TOKEN_TYPES, "weight"), {n_embd, n_token_types}, 0); cls_b = create_tensor(tn(LLM_TENSOR_CLS, "bias"), {n_embd}, 0); cls_norm_b = create_tensor(tn(LLM_TENSOR_CLS_NORM, "bias"), {n_embd}, 0); if (!cls || !cls_norm || !cls_out || !cls_out_b) { throw std::runtime_error("decision model is missing the scorer tensors"); } } std::unique_ptr llama_model_modern_bert::build_arch_graph(const llm_graph_params & params) const { return std::make_unique(*this, params); } llama_model_modern_bert::graph::graph(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) { const int64_t n_embd_head = hparams.n_embd_head_v(); GGML_ASSERT(n_embd_head == hparams.n_embd_head_k()); ggml_tensor * cur; ggml_tensor * inpL; ggml_tensor * inp_pos = build_inp_pos(); // construct input embeddings (token, type, position) inpL = build_inp_embd(model.tok_embd); cb(inpL, "inp_embd", -1); // embed layer norm inpL = build_norm(inpL, model.tok_norm, nullptr, LLM_NORM, 0); cb(inpL, "inp_norm", 0); ggml_tensor * inp_out_ids = build_inp_out_ids(); auto * inp_attn = build_attn_inp_no_cache(); const int n_layer_dec = hparams.n_layer_decision; const int n_layer_enc = n_layer - n_layer_dec; for (int il = 0; il < n_layer_enc; ++il) { const float freq_base_l = model.get_rope_freq_base(cparams, il); const float freq_scale_l = model.get_rope_freq_scale(cparams, il); cur = inpL; // attention layer norm if (model.layers[il].attn_norm) { cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM, il); cb(cur, "attn_norm", il); } // self attention auto [Qcur, Kcur, Vcur] = build_qkv(model.layers[il], cur, n_embd_head, n_head, n_head_kv, il); // RoPE Qcur = ggml_rope_ext( ctx0, Qcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l, ext_factor, attn_factor, beta_fast, beta_slow ); Kcur = ggml_rope_ext( ctx0, Kcur, inp_pos, nullptr, n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l, ext_factor, attn_factor, beta_fast, beta_slow ); cb(Qcur, "Qcur", il); cb(Kcur, "Kcur", il); cb(Vcur, "Vcur", il); cur = build_attn(inp_attn, model.layers[il].wo, nullptr, model.layers[il].wo_s, Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il); cb(cur, "kqv_out", il); if (il == n_layer - 1 && inp_out_ids) { cur = ggml_get_rows(ctx0, cur, inp_out_ids); inpL = ggml_get_rows(ctx0, inpL, inp_out_ids); } // re-add the layer input ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL); cb(ffn_inp, "ffn_inp", il); // attention layer norm cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM, il); cb(cur, "ffn_norm", il); cur = build_ffn(cur, model.layers[il].ffn_up, NULL, NULL, NULL, NULL, NULL, model.layers[il].ffn_down, NULL, NULL, NULL, hparams.llm_ffn_op, LLM_FFN_SEQ, il); // attentions bypass the intermediate layer cur = ggml_add(ctx0, cur, ffn_inp); // input for next layer inpL = cur; } cur = inpL; cur = build_norm(cur, model.output_norm, NULL, LLM_NORM, -1); cb(cur, "final_norm_out", -1); if (n_layer_dec > 0) { cur = build_decision_head(model, cur, inp_attn, inp_out_ids); } res->t_embd = cur; ggml_build_forward_expand(gf, cur); } // returns one score per question type for each output token, the caller reads them at the option markers ggml_tensor * llama_model_modern_bert::graph::build_decision_head( const llama_model & model, ggml_tensor * inp, llm_graph_input_attn_no_cache * inp_attn, ggml_tensor * inp_out_ids) { const int64_t n_embd_head = hparams.n_embd_head_v(); const int n_layer_enc = n_layer - hparams.n_layer_decision; ggml_tensor * scores = nullptr; // the question type is not a graph input, so the head is evaluated for each of them for (uint32_t it = 0; it < N_DECISION_TYPES; ++it) { ggml_tensor * type_row = ggml_view_1d(ctx0, model.type_embd, n_embd, it * model.type_embd->nb[1]); ggml_tensor * inpL = ggml_add(ctx0, inp, type_row); for (int il = n_layer_enc; il < n_layer; ++il) { const auto & layer = model.layers[il]; ggml_tensor * cur = build_norm(inpL, layer.attn_norm, layer.attn_norm_b, LLM_NORM, il); cb(cur, "attn_norm", il); // no positional encoding in the head auto [Qcur, Kcur, Vcur] = build_qkv(layer, cur, n_embd_head, n_head, n_head_kv, il); cur = build_attn(inp_attn, layer.wo, layer.wo_b, layer.wo_s, Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il); cb(cur, "kqv_out", il); if (il == n_layer - 1 && inp_out_ids) { cur = ggml_get_rows(ctx0, cur, inp_out_ids); inpL = ggml_get_rows(ctx0, inpL, inp_out_ids); } ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL); cb(ffn_inp, "ffn_inp", il); cur = build_norm(ffn_inp, layer.ffn_norm, layer.ffn_norm_b, LLM_NORM, il); cb(cur, "ffn_norm", il); cur = build_ffn(cur, layer.ffn_up, layer.ffn_up_b, NULL, NULL, NULL, NULL, layer.ffn_down, layer.ffn_down_b, NULL, NULL, LLM_FFN_RELU, LLM_FFN_SEQ, il); inpL = ggml_add(ctx0, cur, ffn_inp); } // scorer ggml_tensor * cur = build_norm(inpL, model.cls_norm, model.cls_norm_b, LLM_NORM, -1); cur = ggml_add(ctx0, build_lora_mm(model.cls, cur), model.cls_b); cur = ggml_gelu_erf(ctx0, cur); cur = ggml_add(ctx0, build_lora_mm(model.cls_out, cur), model.cls_out_b); scores = scores ? ggml_concat(ctx0, scores, cur, 0) : cur; } cb(scores, "decision_scores", -1); return scores; }