Add Gemma3 support (text only) (#276)

* WIP Gemma3: not working

* gemma3: build_gemma3 seems to be working now

* Revert changes to convert_hf_to_gguf.py

It wasn't working, so I guess, it is better to leave the
conversion up tp upstream.

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>

1 files changed, 208 insertions, 0 deletions

diff --git a/src/llama.cpp b/src/llama.cpp
index 186cb5a5..54e0bb01 100644
--- a/src/llama.cpp
+++ b/src/llama.cpp
@@ -203,6 +203,7 @@ enum llm_arch {
     LLM_ARCH_MINICPM,
     LLM_ARCH_GEMMA,
     LLM_ARCH_GEMMA2,
+    LLM_ARCH_GEMMA3,
     LLM_ARCH_STARCODER2,
     LLM_ARCH_MAMBA,
     LLM_ARCH_XVERSE,
@@ -250,6 +251,7 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
     { LLM_ARCH_MINICPM,         "minicpm"      },
     { LLM_ARCH_GEMMA,           "gemma"        },
     { LLM_ARCH_GEMMA2,          "gemma2"       },
+    { LLM_ARCH_GEMMA3,          "gemma3"       },
     { LLM_ARCH_STARCODER2,      "starcoder2"   },
     { LLM_ARCH_MAMBA,           "mamba"        },
     { LLM_ARCH_XVERSE,          "xverse"       },
@@ -1057,6 +1059,26 @@ static const std::map<llm_arch, std::map<llm_tensor, std::string>> LLM_TENSOR_NA
         },
     },
     {
+        LLM_ARCH_GEMMA3,
+        {
+            { LLM_TENSOR_TOKEN_EMBD,      "token_embd" },
+            { LLM_TENSOR_OUTPUT_NORM,     "output_norm" },
+            { LLM_TENSOR_ATTN_NORM,       "blk.%d.attn_norm" },
+            { LLM_TENSOR_ATTN_Q,          "blk.%d.attn_q" },
+            { LLM_TENSOR_ATTN_Q_NORM,     "blk.%d.attn_q_norm" },
+            { LLM_TENSOR_ATTN_K,          "blk.%d.attn_k" },
+            { LLM_TENSOR_ATTN_K_NORM,     "blk.%d.attn_k_norm" },
+            { LLM_TENSOR_ATTN_V,          "blk.%d.attn_v" },
+            { LLM_TENSOR_ATTN_OUT,        "blk.%d.attn_output" },
+            { LLM_TENSOR_ATTN_POST_NORM,  "blk.%d.post_attention_norm" },
+            { 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_TENSOR_FFN_POST_NORM,   "blk.%d.post_ffw_norm" },
+        },
+    },
+    {
         LLM_ARCH_STARCODER2,
         {
             { LLM_TENSOR_TOKEN_EMBD,      "token_embd" },
@@ -2313,6 +2335,7 @@ struct llama_hparams {
     uint32_t n_layer;
     uint32_t n_rot;
     uint32_t n_swa = 0; // sliding window attention (SWA)
+    uint32_t n_swa_pattern = 1; // by default, all layers use non-sliding-window attention
     uint32_t n_embd_head_k; // dimension of keys (d_k). d_q is assumed to be the same, but there are n_head q heads, and only n_head_kv k-v heads
     uint32_t n_embd_head_v; // dimension of values (d_v) aka n_embd_head
     uint32_t n_expert = 0;
@@ -2342,7 +2365,9 @@ struct llama_hparams {
 
     float    rope_attn_factor = 1.0f;
     float    rope_freq_base_train;
+    float    rope_freq_base_train_swa;
     float    rope_freq_scale_train;
+    float    rope_freq_scale_train_swa;
     uint32_t n_ctx_orig_yarn;
     float    rope_yarn_log_mul;
 
@@ -4963,6 +4988,10 @@ static void llm_load_hparams(
     }
     hparams.rope_freq_scale_train = ropescale == 0.0f ? 1.0f : 1.0f/ropescale;
 
+    // by default assume that the sliding-window layers use the same scaling type as the non-sliding-window layers
+    hparams.rope_freq_base_train_swa  = hparams.rope_freq_base_train;
+    hparams.rope_freq_scale_train_swa = hparams.rope_freq_scale_train;
+
     ml.get_key(LLM_KV_ROPE_SCALING_ATTN_FACTOR, hparams.rope_attn_factor, false);
 
     // non-transformer models do not have attention heads
@@ -5310,6 +5339,28 @@ static void llm_load_hparams(
                     default: model.type = e_model::MODEL_UNKNOWN;
                }
             } break;
+        case LLM_ARCH_GEMMA3:
+            {
+                hparams.n_swa_pattern = 6;
+
+                hparams.rope_freq_base_train_swa  = 10000.0f;
+                hparams.rope_freq_scale_train_swa = 1.0f;
+
+                ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW,    hparams.n_swa);
+                ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+                switch (hparams.n_layer) {
+                    case 26: model.type = e_model::MODEL_2B; break;
+                    case 34: model.type = e_model::MODEL_4B; break;
+                    case 48: model.type = e_model::MODEL_12B; break;
+                    case 62: model.type = e_model::MODEL_27B; break;
+                    default: model.type = e_model::MODEL_UNKNOWN;
+                }
+
+                hparams.f_attention_scale = model.type == e_model::MODEL_27B
+                    ? 1.0f / std::sqrt(float(hparams.n_embd / hparams.n_head(0)))
+                    : 1.0f / std::sqrt(float(hparams.n_embd_head_k));
+            } break;
         case LLM_ARCH_STARCODER2:
             {
                 ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
@@ -7411,6 +7462,38 @@ static bool llm_load_tensors(
                         layer.ffn_post_norm = create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd});
                     }
                 } break;
+            case LLM_ARCH_GEMMA3:
+                {
+                    model.tok_embd = create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    model.output_norm = create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    model.output      = create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD,  "weight"), {n_embd, n_vocab},
+                            llama_model_loader::TENSOR_DUPLICATED); // same as tok_embd, duplicated to allow offloading
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        ggml_context * ctx_layer = ctx_for_layer(i);
+                        ggml_context * ctx_split = ctx_for_layer_split(i);
+                        auto & layer = model.layers[i];
+
+                        layer.attn_norm = create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wq = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd_head_k * n_head}, 0);
+                        layer.wk = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_k_gqa}, 0);
+                        layer.wv = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_v_gqa}, 0);
+                        layer.wo = create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+                        layer.attn_post_norm = create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_k_norm    = create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM,    "weight", i), {n_embd_head_k}, 0);
+                        layer.attn_q_norm    = create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM,    "weight", i), {n_embd_head_k}, 0);
+
+                        layer.ffn_norm = create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_gate = create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_up   = create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_post_norm = create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, 0);
+                    }
+                } break;
             case LLM_ARCH_STARCODER2:
                 {
                     model.tok_embd = create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
@@ -12664,6 +12747,126 @@ struct llm_build_context {
         return gf;
     }
 
+    struct ggml_cgraph * build_gemma3() {
+        struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, llama_model_max_nodes(model), false);
+
+        const int64_t n_embd_head_k = hparams.n_embd_head_k;
+
+        struct ggml_tensor * cur;
+        struct ggml_tensor * inpL;
+
+        inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
+
+        // important: do not normalize weights for raw embeddings input (i.e. encoded image emdeddings)
+        if (batch.token) {
+            inpL = ggml_scale(ctx0, inpL, sqrtf(n_embd));
+            cb(inpL, "inp_scaled", -1);
+        }
+
+        // inp_pos - contains the positions
+        struct ggml_tensor * inp_pos = build_inp_pos();
+
+        // KQ_mask (mask for 1 head, it will be broadcasted to all heads)
+        // gemma3 requires different mask for layers using sliding window (SWA)
+        struct ggml_tensor * KQ_mask     = build_inp_KQ_mask(true);
+        struct ggml_tensor * KQ_mask_swa = build_inp_KQ_mask_swa(true);
+
+        // "5-to-1 interleaved attention"
+        // 5 layers of local attention followed by 1 layer of global attention
+        static const int sliding_window_pattern = 6;
+
+        for (int il = 0; il < n_layer; ++il) {
+            const bool is_sliding          = (il + 1) % sliding_window_pattern;
+            const float freq_base_l        = is_sliding ? 10000.0f    : freq_base;
+            const float freq_scale_l       = is_sliding ? 1.0f        : freq_scale;
+            struct ggml_tensor * KQ_mask_l = is_sliding ? KQ_mask_swa : KQ_mask;
+
+            // norm
+            cur = llm_build_norm(ctx0, inpL, hparams, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, cb, il);
+            cb(cur, "attn_norm", il);
+
+            // self-attention
+            {
+                // compute Q and K and RoPE them
+                struct ggml_tensor * Qcur = llm_build_lora_mm(lctx, ctx0, model.layers[il].wq, cur);
+                cb(Qcur, "Qcur", il);
+
+                struct ggml_tensor * Kcur = llm_build_lora_mm(lctx, ctx0, model.layers[il].wk, cur);
+                cb(Kcur, "Kcur", il);
+
+                struct ggml_tensor * Vcur = llm_build_lora_mm(lctx, ctx0, model.layers[il].wv, cur);
+                cb(Vcur, "Vcur", il);
+
+                Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head_k, n_head,    n_tokens);
+                Qcur = llm_build_norm(ctx0, Qcur, hparams, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, cb, il);
+                cb(Qcur, "Qcur_normed", il);
+
+                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);
+                cb(Qcur, "Qcur", il);
+
+                Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head_k, n_head_kv, n_tokens);
+                Kcur = llm_build_norm(ctx0, Kcur, hparams, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, cb, il);
+                cb(Kcur, "Kcur_normed", il);
+
+                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(Kcur, "Kcur", il);
+
+                cur = llm_build_kv(ctx0, lctx, kv_self, gf, model.layers[il].wo, NULL,
+                        Kcur, Vcur, Qcur, KQ_mask_l, n_tokens, kv_head, n_kv, hparams.f_attention_scale, cb, il);
+            }
+
+            cur = llm_build_norm(ctx0, cur, hparams, model.layers[il].attn_post_norm, NULL, LLM_NORM_RMS, cb, il);
+            cb(cur, "attn_post_norm", il);
+
+            if (il == n_layer - 1) {
+                // skip computing output for unused tokens
+                struct ggml_tensor * inp_out_ids = build_inp_out_ids();
+                cur  = ggml_get_rows(ctx0,  cur, inp_out_ids);
+                inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+            }
+
+            struct ggml_tensor * sa_out = ggml_add(ctx0, cur, inpL);
+            cb(sa_out, "sa_out", il);
+
+            cur = llm_build_norm(ctx0, sa_out, hparams, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, cb, il);
+            cb(cur, "ffn_norm", il);
+
+            // feed-forward network
+            cur = llm_build_ffn(ctx0, lctx, cur,
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_GELU, LLM_FFN_PAR, cb, il);
+            cb(cur, "ffn_out", il);
+
+            cur = llm_build_norm(ctx0, cur, hparams, model.layers[il].ffn_post_norm, NULL, LLM_NORM_RMS, cb, -1);
+            cb(cur, "ffn_post_norm", -1);
+
+            cur = ggml_add(ctx0, cur, sa_out);
+            cur = lctx.cvec.apply_to(ctx0, cur, il);
+            cb(cur, "l_out", il);
+
+            // input for next layer
+            inpL = cur;
+        }
+
+        cur = inpL;
+
+        cur = llm_build_norm(ctx0, cur, hparams, model.output_norm, NULL, LLM_NORM_RMS, cb, -1);
+        cb(cur, "result_norm", -1);
+
+        // lm_head
+        cur = llm_build_lora_mm(lctx, ctx0, model.output, cur);
+
+        cb(cur, "result_output", -1);
+
+        ggml_build_forward_expand(gf, cur);
+
+        return gf;
+    }
 
     struct ggml_cgraph * build_starcoder2() {
         struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, llama_model_max_nodes(model), false);
@@ -14996,6 +15199,10 @@ static struct ggml_cgraph * llama_build_graph(
             {
                 result = llm.build_gemma2();
             } break;
+        case LLM_ARCH_GEMMA3:
+            {
+                result = llm.build_gemma3();
+            } break;
         case LLM_ARCH_STARCODER2:
             {
                 result = llm.build_starcoder2();
@@ -18826,6 +19033,7 @@ enum llama_rope_type llama_rope_type(const struct llama_model * model) {
         case LLM_ARCH_PHI3:
         case LLM_ARCH_GEMMA:
         case LLM_ARCH_GEMMA2:
+        case LLM_ARCH_GEMMA3:
         case LLM_ARCH_STARCODER2:
         case LLM_ARCH_OPENELM:
         case LLM_ARCH_GPTNEOX: