whispercpp 1.3.5 → 1.3.6

data/ext/sources/examples/talk-llama/models/paddleocr.cpp

@@ -0,0 +1,122 @@
+#include "models.h"
+
+llm_build_paddleocr::llm_build_paddleocr(const llama_model & model, const llm_graph_params & params) :
+    llm_graph_context(params) {
+
+    // NOTE: same with qwen2vl.cpp, but bias tensors are optional
+
+    const int64_t n_embd_head = hparams.n_embd_head_v();
+
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    int sections[4];
+    std::copy(std::begin(hparams.rope_sections), std::begin(hparams.rope_sections) + 4, sections);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+
+    auto * inp_attn = build_attn_inp_kv();
+
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // norm
+        {
+            cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "attn_norm", il);
+        }
+        // self-attention
+        {
+            ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur);
+            cb(Qcur, "Qcur", il);
+            if (model.layers[il].bq) {
+                Qcur = ggml_add(ctx0, Qcur, model.layers[il].bq);
+                cb(Qcur, "Qcur", il);
+            }
+            ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur);
+            cb(Kcur, "Kcur", il);
+            if (model.layers[il].bk) {
+                Kcur = ggml_add(ctx0, Kcur, model.layers[il].bk);
+                cb(Kcur, "Kcur", il);
+            }
+            ggml_tensor * Vcur = build_lora_mm(model.layers[il].wv, cur);
+            cb(Vcur, "Vcur", il);
+            if (model.layers[il].bv) {
+                Vcur = ggml_add(ctx0, Vcur, model.layers[il].bv);
+                cb(Vcur, "Vcur", il);
+            }
+            Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
+            Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
+            Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head, n_head_kv, n_tokens);
+
+            Qcur = ggml_rope_multi(
+                    ctx0, Qcur, inp_pos, nullptr,
+                    n_rot, sections, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    ext_factor, attn_factor, beta_fast, beta_slow
+                    );
+
+            Kcur = ggml_rope_multi(
+                    ctx0, Kcur, inp_pos, nullptr,
+                    n_rot, sections, rope_type, n_ctx_orig, freq_base, freq_scale,
+                    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, model.layers[il].bo,
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+        }
+        if (il == n_layer - 1) {
+            // skip computing output for unused tokens
+            cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // feed-forward network
+        {
+            cur = build_norm(ffn_inp, model.layers[il].ffn_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "ffn_norm", il);
+
+            cur = build_ffn(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_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+        cur = ggml_add(ctx0, cur, ffn_inp);
+
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+    cur = inpL;
+
+    cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+    cb(cur, "result_norm", -1);
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur);
+
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}

data/ext/sources/examples/talk-llama/models/pangu-embedded.cpp

@@ -2,10 +2,10 @@
 
 
 llm_build_pangu_embedded::llm_build_pangu_embedded(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
-    const int64_t n_embd_head = hparams.n_embd_head_v;
+    const int64_t n_embd_head = hparams.n_embd_head_v();
 
-    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
-    GGML_ASSERT(n_embd_head == hparams.n_rot);
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
 
     ggml_tensor * cur;
     ggml_tensor * inpL;

data/ext/sources/examples/talk-llama/models/phi2.cpp

@@ -2,10 +2,10 @@
 
 
 llm_build_phi2::llm_build_phi2(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
-    const int64_t n_embd_head = hparams.n_embd_head_v;
+    const int64_t n_embd_head = hparams.n_embd_head_v();
     const int64_t n_embd_gqa  = hparams.n_embd_v_gqa();
 
-    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
 
     ggml_tensor * cur;
     ggml_tensor * attn_norm_output;

data/ext/sources/examples/talk-llama/models/phi3.cpp

@@ -2,10 +2,10 @@
 
 template<bool iswa>
 llm_build_phi3<iswa>::llm_build_phi3(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
-    const int64_t n_embd_head = hparams.n_embd_head_v;
+    const int64_t n_embd_head = hparams.n_embd_head_v();
     const int64_t n_embd_gqa = hparams.n_embd_v_gqa();
 
-    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
 
     ggml_tensor * cur;
     ggml_tensor * inpL;
@@ -114,7 +114,7 @@ llm_build_phi3<iswa>::llm_build_phi3(const llama_model & model, const llm_graph_
                     nullptr,
                     n_expert, n_expert_used,
                     LLM_FFN_SILU, true,
-                    false, 0.0,
+                    hparams.expert_weights_scale,
                     LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
                     il);
             cb(cur, "ffn_moe_out", il);

data/ext/sources/examples/talk-llama/models/plamo.cpp

@@ -1,10 +1,10 @@
 #include "models.h"
 
 llm_build_plamo::llm_build_plamo(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
-    const int64_t n_embd_head = hparams.n_embd_head_v;
+    const int64_t n_embd_head = hparams.n_embd_head_v();
 
-    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
-    GGML_ASSERT(n_embd_head == hparams.n_rot);
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
 
     ggml_tensor * cur;
     ggml_tensor * inpL;

data/ext/sources/examples/talk-llama/models/plamo2.cpp

@@ -1,7 +1,9 @@
 #include "models.h"
 
+#include "llama-memory-recurrent.h"
+
 llm_build_plamo2::llm_build_plamo2(const llama_model & model, const llm_graph_params & params) :
-    llm_graph_context_mamba(params) {
+    llm_build_mamba_base(params) {
     ggml_tensor * cur;
     ggml_tensor * inpL;
 
@@ -25,7 +27,7 @@ llm_build_plamo2::llm_build_plamo2(const llama_model & model, const llm_graph_pa
         cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
 
         // check if this layer is Mamba or Attention
-        bool is_mamba_layer = hparams.is_recurrent(il);
+        const bool is_mamba_layer = hparams.is_recurrent(il);
 
         if (is_mamba_layer) {
             // PLaMo-2 Mamba layer
@@ -104,9 +106,9 @@ ggml_tensor * llm_build_plamo2::build_plamo2_attn_layer(llm_graph_input_attn_kv
         cb(qkv, "wqkv", il);
 
         // split QKV tensor into Q, K, V
-        const int64_t n_embd_head_q = hparams.n_embd_head_k;
-        const int64_t n_embd_head_k = hparams.n_embd_head_k;
-        const int64_t n_embd_head_v = hparams.n_embd_head_v;
+        const int64_t n_embd_head_q = hparams.n_embd_head_k();
+        const int64_t n_embd_head_k = hparams.n_embd_head_k();
+        const int64_t n_embd_head_v = hparams.n_embd_head_v();
         int32_t       n_head        = hparams.n_head(il);
         int32_t       n_head_kv     = hparams.n_head_kv(il);
 
@@ -169,6 +171,8 @@ ggml_tensor * llm_build_plamo2::build_plamo2_mamba_layer(llm_graph_input_rs * in
     GGML_ASSERT(n_seqs != 0);
     GGML_ASSERT(ubatch.equal_seqs());
     GGML_ASSERT(ubatch.n_tokens == n_seq_tokens * n_seqs);
+    GGML_ASSERT(d_inner % n_head == 0);
+    GGML_ASSERT(n_group == 0);
 
     ggml_tensor * conv_states_all = mctx_cur->get_r_l(il);
     ggml_tensor * ssm_states_all  = mctx_cur->get_s_l(il);

data/ext/sources/examples/talk-llama/models/plamo3.cpp

@@ -3,8 +3,8 @@
 template <bool iswa>
 llm_build_plamo3<iswa>::llm_build_plamo3(const llama_model & model, const llm_graph_params & params) :
     llm_graph_context(params) {
-    const int64_t head_dim_q = hparams.n_embd_head_k;
-    const int64_t head_dim_v = hparams.n_embd_head_v;
+    const int64_t head_dim_q = hparams.n_embd_head_k();
+    const int64_t head_dim_v = hparams.n_embd_head_v();
 
     ggml_tensor * cur;
     ggml_tensor * inpL = build_inp_embd(model.tok_embd);

data/ext/sources/examples/talk-llama/models/plm.cpp

@@ -1,10 +1,11 @@
 #include "models.h"
 
 llm_build_plm::llm_build_plm(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
-    const float kq_scale = 1.0f/sqrtf(float(hparams.n_embd_head_k));
+    const float kq_scale = 1.0f/sqrtf(float(hparams.n_embd_head_k()));
+
+    const uint32_t n_embd_head_qk_rope = hparams.n_rot();
+    const uint32_t n_embd_head_qk_nope = hparams.n_embd_head_k() - hparams.n_rot();
 
-    const uint32_t n_embd_head_qk_rope = hparams.n_rot;
-    const uint32_t n_embd_head_qk_nope = hparams.n_embd_head_k - hparams.n_rot;
     const uint32_t kv_lora_rank = hparams.n_lora_kv;
 
     ggml_tensor * cur;
@@ -37,15 +38,15 @@ llm_build_plm::llm_build_plm(const llama_model & model, const llm_graph_params &
 
             // split into {n_head * n_embd_head_qk_nope, n_tokens}
             ggml_tensor * q_nope = ggml_view_3d(ctx0, q, n_embd_head_qk_nope, n_head, n_tokens,
-                    ggml_row_size(q->type, hparams.n_embd_head_k),
-                    ggml_row_size(q->type, hparams.n_embd_head_k * n_head),
+                    ggml_row_size(q->type, hparams.n_embd_head_k()),
+                    ggml_row_size(q->type, hparams.n_embd_head_k() * n_head),
                     0);
             cb(q_nope, "q_nope", il);
 
             // and {n_head * n_embd_head_qk_rope, n_tokens}
             ggml_tensor * q_pe = ggml_view_3d(ctx0, q, n_embd_head_qk_rope, n_head, n_tokens,
-                    ggml_row_size(q->type, hparams.n_embd_head_k),
-                    ggml_row_size(q->type, hparams.n_embd_head_k * n_head),
+                    ggml_row_size(q->type, hparams.n_embd_head_k()),
+                    ggml_row_size(q->type, hparams.n_embd_head_k() * n_head),
                     ggml_row_size(q->type, n_embd_head_qk_nope));
             cb(q_pe, "q_pe", il);
 
@@ -77,23 +78,23 @@ llm_build_plm::llm_build_plm(const llama_model & model, const llm_graph_params &
 
             // split into {n_head * n_embd_head_qk_nope, n_tokens}
             ggml_tensor * k_nope = ggml_view_3d(ctx0, kv, n_embd_head_qk_nope, n_head, n_tokens,
-                    ggml_row_size(kv->type, n_embd_head_qk_nope + hparams.n_embd_head_v),
-                    ggml_row_size(kv->type, n_head * (n_embd_head_qk_nope + hparams.n_embd_head_v)),
+                    ggml_row_size(kv->type, n_embd_head_qk_nope + hparams.n_embd_head_v()),
+                    ggml_row_size(kv->type, n_head * (n_embd_head_qk_nope + hparams.n_embd_head_v())),
                     0);
             cb(k_nope, "k_nope", il);
 
             // and {n_head * n_embd_head_v, n_tokens}
-            ggml_tensor * v_states = ggml_view_3d(ctx0, kv, hparams.n_embd_head_v, n_head, n_tokens,
-                    ggml_row_size(kv->type, (n_embd_head_qk_nope + hparams.n_embd_head_v)),
-                    ggml_row_size(kv->type, (n_embd_head_qk_nope + hparams.n_embd_head_v)*n_head),
+            ggml_tensor * v_states = ggml_view_3d(ctx0, kv, hparams.n_embd_head_v(), n_head, n_tokens,
+                    ggml_row_size(kv->type, (n_embd_head_qk_nope + hparams.n_embd_head_v())),
+                    ggml_row_size(kv->type, (n_embd_head_qk_nope + hparams.n_embd_head_v())*n_head),
                     ggml_row_size(kv->type, (n_embd_head_qk_nope)));
             cb(v_states, "v_states", il);
 
             v_states = ggml_cont(ctx0, v_states);
             cb(v_states, "v_states", il);
 
-            v_states = ggml_view_2d(ctx0, v_states, hparams.n_embd_head_v * n_head, n_tokens,
-                    ggml_row_size(kv->type, hparams.n_embd_head_v * n_head),
+            v_states = ggml_view_2d(ctx0, v_states, hparams.n_embd_head_v() * n_head, n_tokens,
+                    ggml_row_size(kv->type, hparams.n_embd_head_v() * n_head),
                     0);
             cb(v_states, "v_states", il);
 

data/ext/sources/examples/talk-llama/models/qwen.cpp

@@ -2,9 +2,9 @@
 
 
 llm_build_qwen::llm_build_qwen(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
-    const int64_t n_embd_head = hparams.n_embd_head_v;
+    const int64_t n_embd_head = hparams.n_embd_head_v();
 
-    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
 
     ggml_tensor * cur;
     ggml_tensor * inpL;

data/ext/sources/examples/talk-llama/models/qwen2.cpp

@@ -1,10 +1,10 @@
 #include "models.h"
 
 llm_build_qwen2::llm_build_qwen2(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
-    const int64_t n_embd_head = hparams.n_embd_head_v;
+    const int64_t n_embd_head = hparams.n_embd_head_v();
 
-    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
-    GGML_ASSERT(n_embd_head == hparams.n_rot);
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
 
     ggml_tensor * cur;
     ggml_tensor * inpL;

data/ext/sources/examples/talk-llama/models/qwen2moe.cpp

@@ -1,10 +1,10 @@
 #include "models.h"
 
 llm_build_qwen2moe::llm_build_qwen2moe(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
-    const int64_t n_embd_head = hparams.n_embd_head_v;
+    const int64_t n_embd_head = hparams.n_embd_head_v();
 
-    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
-    GGML_ASSERT(n_embd_head == hparams.n_rot);
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
 
     ggml_tensor * cur;
     ggml_tensor * inpL;
@@ -94,7 +94,7 @@ llm_build_qwen2moe::llm_build_qwen2moe(const llama_model & model, const llm_grap
                     nullptr,
                     n_expert, n_expert_used,
                     LLM_FFN_SILU, false,
-                    false, 0.0,
+                    hparams.expert_weights_scale,
                     LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX,
                     il);
         cb(moe_out, "ffn_moe_out", il);

data/ext/sources/examples/talk-llama/models/qwen2vl.cpp

@@ -1,10 +1,10 @@
 #include "models.h"
 
 llm_build_qwen2vl::llm_build_qwen2vl(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
-    const int64_t n_embd_head = hparams.n_embd_head_v;
+    const int64_t n_embd_head = hparams.n_embd_head_v();
 
-    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
-    GGML_ASSERT(n_embd_head == hparams.n_rot);
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
 
     ggml_tensor * cur;
     ggml_tensor * inpL;

data/ext/sources/examples/talk-llama/models/qwen3.cpp

@@ -1,10 +1,10 @@
 #include "models.h"
 
 llm_build_qwen3::llm_build_qwen3(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
-    const int64_t n_embd_head = hparams.n_embd_head_v;
+    const int64_t n_embd_head = hparams.n_embd_head_v();
 
-    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
-    GGML_ASSERT(n_embd_head == hparams.n_rot);
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k());
+    GGML_ASSERT(n_embd_head == n_rot);
 
     ggml_tensor * cur;
     ggml_tensor * inpL;
@@ -30,13 +30,13 @@ llm_build_qwen3::llm_build_qwen3(const llama_model & model, const llm_graph_para
         // self-attention
         {
             // compute Q and K and RoPE them
-            ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur);
+            ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur, model.layers[il].wq_s);
             cb(Qcur, "Qcur", il);
 
-            ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur);
+            ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur, model.layers[il].wk_s);
             cb(Kcur, "Kcur", il);
 
-            ggml_tensor * Vcur = build_lora_mm(model.layers[il].wv, cur);
+            ggml_tensor * Vcur = build_lora_mm(model.layers[il].wv, cur, model.layers[il].wv_s);
             cb(Vcur, "Vcur", il);
 
             Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head,    n_tokens);
@@ -68,6 +68,9 @@ llm_build_qwen3::llm_build_qwen3(const llama_model & model, const llm_graph_para
             cur = build_attn(inp_attn,
                     model.layers[il].wo, model.layers[il].bo,
                     Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+            if (model.layers[il].wo_s) {
+                cur = ggml_mul(ctx0, cur, model.layers[il].wo_s);
+            }
         }
         if (il == n_layer - 1 && inp_out_ids) {
             cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
@@ -83,9 +86,9 @@ llm_build_qwen3::llm_build_qwen3(const llama_model & model, const llm_graph_para
         cb(cur, "ffn_norm", il);
 
         cur = build_ffn(cur,
-                model.layers[il].ffn_up,   NULL, NULL,
-                model.layers[il].ffn_gate, NULL, NULL,
-                model.layers[il].ffn_down, NULL, NULL,
+                model.layers[il].ffn_up,   NULL, model.layers[il].ffn_up_s,
+                model.layers[il].ffn_gate, NULL, model.layers[il].ffn_gate_s,
+                model.layers[il].ffn_down, NULL, model.layers[il].ffn_down_s,
                 NULL,
                 LLM_FFN_SILU, LLM_FFN_PAR, il);
         cb(cur, "ffn_out", il);