diff --git a/llama.cpp/llama.cpp b/llama.cpp/llama.cpp index 21b582f003..aa898c6581 100644 --- a/llama.cpp/llama.cpp +++ b/llama.cpp/llama.cpp @@ -178,6 +178,8 @@ enum llm_arch { LLM_ARCH_JAIS, LLM_ARCH_NEMOTRON, LLM_ARCH_EXAONE, + LLM_ARCH_GRANITE, + LLM_ARCH_GRANITE_MOE, LLM_ARCH_UNKNOWN, }; @@ -225,6 +227,8 @@ static const std::map LLM_ARCH_NAMES = { { LLM_ARCH_JAIS, "jais" }, { LLM_ARCH_NEMOTRON, "nemotron" }, { LLM_ARCH_EXAONE, "exaone" }, + { LLM_ARCH_GRANITE, "granite" }, + { LLM_ARCH_GRANITE_MOE, "granitemoe" }, { LLM_ARCH_UNKNOWN, "(unknown)" }, }; @@ -261,6 +265,8 @@ enum llm_kv { LLM_KV_DECODER_START_TOKEN_ID, LLM_KV_ATTN_LOGIT_SOFTCAPPING, LLM_KV_FINAL_LOGIT_SOFTCAPPING, + LLM_KV_RESIDUAL_SCALE, + LLM_KV_EMBEDDING_SCALE, LLM_KV_ATTENTION_HEAD_COUNT, LLM_KV_ATTENTION_HEAD_COUNT_KV, @@ -275,6 +281,7 @@ enum llm_kv { LLM_KV_ATTENTION_KV_LORA_RANK, LLM_KV_ATTENTION_RELATIVE_BUCKETS_COUNT, LLM_KV_ATTENTION_SLIDING_WINDOW, + LLM_KV_ATTENTION_SCALE, LLM_KV_ROPE_DIMENSION_COUNT, LLM_KV_ROPE_FREQ_BASE, @@ -359,6 +366,8 @@ static const std::map LLM_KV_NAMES = { { LLM_KV_DECODER_START_TOKEN_ID, "%s.decoder_start_token_id" }, { LLM_KV_ATTN_LOGIT_SOFTCAPPING, "%s.attn_logit_softcapping" }, { LLM_KV_FINAL_LOGIT_SOFTCAPPING, "%s.final_logit_softcapping" }, + { LLM_KV_RESIDUAL_SCALE, "%s.residual_scale" }, + { LLM_KV_EMBEDDING_SCALE, "%s.embedding_scale" }, { LLM_KV_ATTENTION_HEAD_COUNT, "%s.attention.head_count" }, { LLM_KV_ATTENTION_HEAD_COUNT_KV, "%s.attention.head_count_kv" }, @@ -373,6 +382,7 @@ static const std::map LLM_KV_NAMES = { { LLM_KV_ATTENTION_KV_LORA_RANK, "%s.attention.kv_lora_rank" }, { LLM_KV_ATTENTION_RELATIVE_BUCKETS_COUNT, "%s.attention.relative_buckets_count" }, { LLM_KV_ATTENTION_SLIDING_WINDOW, "%s.attention.sliding_window" }, + { LLM_KV_ATTENTION_SCALE, "%s.attention.scale" }, { LLM_KV_ROPE_DIMENSION_COUNT, "%s.rope.dimension_count" }, { LLM_KV_ROPE_FREQ_BASE, "%s.rope.freq_base" }, @@ -1303,6 +1313,41 @@ static const std::map> LLM_TENSOR_NA { LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" }, }, }, + { + LLM_ARCH_GRANITE, + { + { LLM_TENSOR_TOKEN_EMBD, "token_embd" }, + { LLM_TENSOR_OUTPUT_NORM, "output_norm" }, + { LLM_TENSOR_OUTPUT, "output" }, + { LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" }, + { LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" }, + { LLM_TENSOR_ATTN_K, "blk.%d.attn_k" }, + { LLM_TENSOR_ATTN_V, "blk.%d.attn_v" }, + { LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" }, + { LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" }, + { LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" }, + { LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" }, + { LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" }, + }, + }, + { + LLM_ARCH_GRANITE_MOE, + { + { LLM_TENSOR_TOKEN_EMBD, "token_embd" }, + { LLM_TENSOR_OUTPUT_NORM, "output_norm" }, + { LLM_TENSOR_OUTPUT, "output" }, + { LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" }, + { LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" }, + { LLM_TENSOR_ATTN_K, "blk.%d.attn_k" }, + { LLM_TENSOR_ATTN_V, "blk.%d.attn_v" }, + { LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" }, + { LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" }, + { LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" }, + { LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" }, + { LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" }, + { LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" }, + }, + }, { LLM_ARCH_UNKNOWN, { @@ -1931,6 +1976,11 @@ struct llama_hparams { float f_max_alibi_bias = 0.0f; float f_logit_scale = 0.0f; + // Additional scale factors (Granite/Granite MoE) + float f_residual_scale = 0.0f; + float f_embedding_scale = 0.0f; + float f_attention_scale = 0.0f; + bool causal_attn = true; bool use_alibi = false; bool attn_soft_cap = false; @@ -1987,6 +2037,9 @@ struct llama_hparams { if (!is_float_close(this->rope_freq_scale_train, other.rope_freq_scale_train, EPSILON)) return true; if (!is_float_close(this->expert_weights_scale, other.expert_weights_scale, EPSILON)) return true; if (!is_float_close(this->rope_yarn_log_mul, other.rope_yarn_log_mul, EPSILON)) return true; + if (!is_float_close(this->f_residual_scale, other.f_residual_scale, EPSILON)) return true; + if (!is_float_close(this->f_embedding_scale, other.f_embedding_scale, EPSILON)) return true; + if (!is_float_close(this->f_attention_scale, other.f_attention_scale, EPSILON)) return true; return false; } @@ -4998,6 +5051,22 @@ static void llm_load_hparams( default: model.type = e_model::MODEL_UNKNOWN; } } break; + case LLM_ARCH_GRANITE: + case LLM_ARCH_GRANITE_MOE: + { + ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); + ml.get_key(LLM_KV_LOGIT_SCALE, hparams.f_logit_scale); + ml.get_key(LLM_KV_RESIDUAL_SCALE, hparams.f_residual_scale); + ml.get_key(LLM_KV_EMBEDDING_SCALE, hparams.f_embedding_scale); + ml.get_key(LLM_KV_ATTENTION_SCALE, hparams.f_attention_scale); + + switch (hparams.n_layer) { + case 32: model.type = e_model::MODEL_3B; break; + case 40: model.type = e_model::MODEL_3B; break; + // Add additional layer/vocab/etc checks here for other model sizes + default: model.type = e_model::MODEL_UNKNOWN; + } + } break; default: (void)0; } @@ -5665,6 +5734,12 @@ static void llm_load_print_meta(llama_model_loader & ml, llama_model & model) { LLAMA_LOG_INFO("%s: n_ff_exp = %d\n", __func__, hparams.n_ff_exp); LLAMA_LOG_INFO("%s: n_ff_shexp = %d\n", __func__, hparams.n_ff_shexp); } + + if (model.arch == LLM_ARCH_GRANITE || model.arch == LLM_ARCH_GRANITE_MOE) { + LLAMA_LOG_INFO("%s: f_embedding_scale = %f\n", __func__, hparams.f_embedding_scale); + LLAMA_LOG_INFO("%s: f_residual_scale = %f\n", __func__, hparams.f_residual_scale); + LLAMA_LOG_INFO("%s: f_attention_scale = %f\n", __func__, hparams.f_attention_scale); + } } // Returns false if cancelled by progress_callback @@ -5837,6 +5912,8 @@ static bool llm_load_tensors( case LLM_ARCH_LLAMA: case LLM_ARCH_REFACT: case LLM_ARCH_MINICPM: + case LLM_ARCH_GRANITE: + case LLM_ARCH_GRANITE_MOE: { model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); @@ -7669,6 +7746,11 @@ static struct ggml_tensor * llm_build_inp_embd( ggml_set_input(lctx.inp_embd); } + // For Granite architecture + if (hparams.f_embedding_scale != 0.0f) { + inpL = ggml_scale(ctx, inpL, hparams.f_embedding_scale); + } + cb(inpL, "inp_embd", -1); return inpL; @@ -8646,6 +8728,7 @@ struct llm_build_context { // KQ_mask (mask for 1 head, it will be broadcasted to all heads) struct ggml_tensor * KQ_mask = build_inp_KQ_mask(); + const float kq_scale = hparams.f_attention_scale == 0.0f ? 1.0f/sqrtf(float(n_embd_head)) : hparams.f_attention_scale; for (int il = 0; il < n_layer; ++il) { struct ggml_tensor * inpSA = inpL; @@ -8698,7 +8781,7 @@ struct llm_build_context { cur = llm_build_kv(ctx0, lctx, kv_self, gf, model.layers[il].wo, model.layers[il].bo, - Kcur, Vcur, Qcur, KQ_mask, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il); + Kcur, Vcur, Qcur, KQ_mask, n_tokens, kv_head, n_kv, kq_scale, cb, il); } if (il == n_layer - 1) { @@ -8709,6 +8792,11 @@ struct llm_build_context { inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); } + // For Granite architecture + if (hparams.f_residual_scale) { + cur = ggml_scale(ctx0, cur, hparams.f_residual_scale); + } + struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA); cb(ffn_inp, "ffn_inp", il); @@ -8745,6 +8833,11 @@ struct llm_build_context { cb(cur, "ffn_moe_out", il); } + // For Granite architecture + if (hparams.f_residual_scale) { + cur = ggml_scale(ctx0, cur, hparams.f_residual_scale); + } + cur = ggml_add(ctx0, cur, ffn_inp); cb(cur, "ffn_out", il); @@ -8764,6 +8857,12 @@ struct llm_build_context { // lm_head cur = llm_build_lora_mm(lctx, ctx0, model.output, cur); + + // For Granite architecture + if (hparams.f_logit_scale) { + cur = ggml_scale(ctx0, cur, 1.0f / hparams.f_logit_scale); + } + cb(cur, "result_output", -1); ggml_build_forward_expand(gf, cur); @@ -13942,6 +14041,8 @@ static struct ggml_cgraph * llama_build_graph( switch (model.arch) { case LLM_ARCH_LLAMA: + case LLM_ARCH_GRANITE: + case LLM_ARCH_GRANITE_MOE: { result = llm.build_llama(); } break; @@ -17195,6 +17296,8 @@ enum llama_rope_type llama_rope_type(const struct llama_model * model) { case LLM_ARCH_ARCTIC: case LLM_ARCH_DEEPSEEK2: case LLM_ARCH_CHATGLM: + case LLM_ARCH_GRANITE: + case LLM_ARCH_GRANITE_MOE: return LLAMA_ROPE_TYPE_NORM; // the pairs of head values are offset by n_rot/2 @@ -19176,6 +19279,19 @@ static int32_t llama_chat_apply_template_internal( if (add_ass) { ss << "[|assistant|]"; } + } else if (tmpl == "granite" || tmpl == "granitemoe" || tmpl_contains("<|start_of_role|>")) { + // IBM Granite template + for (const auto & message : chat) { + std::string role(message->role); + ss << "<|start_of_role|>" << role << "<|end_of_role|>"; + if (role == "assistant_tool_call") { + ss << "<|tool_call|>"; + } + ss << message->content << "<|end_of_text|>\n"; + } + if (add_ass) { + ss << "<|start_of_role|>assistant<|end_of_role|>\n"; + } } else { // template not supported return -1;