@fugood/llama.node 1.1.9 → 1.1.11

package/src/llama.cpp/src/llama-kv-cache.cpp

@@ -17,32 +17,25 @@
 //
 
 llama_kv_cache::llama_kv_cache(
-        const llama_model &  model,
-          layer_filter_cb && filter,
-                ggml_type    type_k,
-                ggml_type    type_v,
-                     bool    v_trans,
-                     bool    offload,
-                     bool    unified,
-                 uint32_t    kv_size,
-                 uint32_t    n_seq_max,
-                 uint32_t    n_pad,
-                 uint32_t    n_swa,
-           llama_swa_type    swa_type) :
+        const llama_model & model,
+                ggml_type   type_k,
+                ggml_type   type_v,
+                     bool   v_trans,
+                     bool   offload,
+                     bool   unified,
+                 uint32_t   kv_size,
+                 uint32_t   n_seq_max,
+                 uint32_t   n_pad,
+                 uint32_t   n_swa,
+           llama_swa_type   swa_type,
+    const layer_filter_cb & filter,
+    const  layer_reuse_cb & reuse) :
     model(model), hparams(model.hparams), v_trans(v_trans),
     n_seq_max(n_seq_max), n_stream(unified ? 1 : n_seq_max), n_pad(n_pad), n_swa(n_swa), swa_type(swa_type) {
 
     GGML_ASSERT(kv_size % n_pad == 0);
 
-    // TODO: this is temporary until we support passing reuse layer filters [KV_REUSE]
-    auto n_layer_cache = hparams.n_layer;
-    if (model.arch == LLM_ARCH_GEMMA3N) {
-        n_layer_cache = 20;
-    }
-    if (model.arch == LLM_ARCH_GLM4_MOE) {
-        // GLM-4.5: Only process up to last layer, skip final NextN layer
-        n_layer_cache = hparams.n_layer - hparams.nextn_predict_layers;
-    }
+    const uint32_t n_layer_kv = hparams.n_layer_kv();
 
     // create a context for each buffer type
     std::map<ggml_backend_buffer_type_t, ggml_context *> ctx_map;
@@ -50,7 +43,7 @@ llama_kv_cache::llama_kv_cache(
         auto it = ctx_map.find(buft);
         if (it == ctx_map.end()) {
             ggml_init_params params = {
-                /*.mem_size   =*/ size_t(2u*(1 + n_stream)*n_layer_cache*ggml_tensor_overhead()),
+                /*.mem_size   =*/ size_t(2u*(1 + n_stream)*n_layer_kv*ggml_tensor_overhead()),
                 /*.mem_buffer =*/ NULL,
                 /*.no_alloc   =*/ true,
             };
@@ -97,9 +90,14 @@ llama_kv_cache::llama_kv_cache(
                 __func__, hparams.n_embd_v_gqa_max());
     }
 
-    for (uint32_t il = 0; il < n_layer_cache; il++) {
+    for (uint32_t il = 0; il < hparams.n_layer; il++) {
+        if (!hparams.has_kv(il)) {
+            LLAMA_LOG_DEBUG("%s: layer %3d: does not have KV cache\n", __func__, il);
+            continue;
+        }
+
         if (filter && !filter(il)) {
-            LLAMA_LOG_DEBUG("%s: layer %3d: skipped\n", __func__, il);
+            LLAMA_LOG_DEBUG("%s: layer %3d: filtered\n", __func__, il);
             continue;
         }
 
@@ -147,23 +145,27 @@ llama_kv_cache::llama_kv_cache(
         layers.push_back({ il, k, v, k_stream, v_stream, });
     }
 
-    // TODO: this is temporary until we support passing reuse layer filters [KV_REUSE]
-    if (model.arch == LLM_ARCH_GEMMA3N) {
-        LLAMA_LOG_DEBUG("%s: GEMMA3N: reuse layers [%d, %d]\n", __func__, n_layer_cache, hparams.n_layer - 1);
+    if (reuse) {
+        LLAMA_LOG_DEBUG("%s: reusing layers:\n", __func__);
 
-        for (uint32_t il = n_layer_cache; il < hparams.n_layer; il++) {
-            if (filter && !filter(il)) {
-                LLAMA_LOG_DEBUG("%s: layer %3d: skipped\n", __func__, il);
+        for (uint32_t il = 0; il < hparams.n_layer; il++) {
+            const int32_t il_reuse = reuse(il);
+
+            if (il_reuse < 0) {
+                LLAMA_LOG_DEBUG("%s: - layer %3d: no reuse\n", __func__, il);
                 continue;
             }
 
-            const bool     is_swa   = hparams.is_swa(il);
-            const uint32_t il_reuse = n_layer_cache - (is_swa ? 2 : 1);
+            if (filter && !filter(il)) {
+                LLAMA_LOG_DEBUG("%s: - layer %3d: filtered\n", __func__, il);
+                continue;
+            }
 
             GGML_ASSERT(map_layer_ids.find(il_reuse) != map_layer_ids.end());
+
             map_layer_ids[il] = map_layer_ids[il_reuse];
 
-            LLAMA_LOG_DEBUG("%s: layer %3d: reuse layer %d, isw = %d\n", __func__, il, il_reuse, is_swa);
+            LLAMA_LOG_DEBUG("%s: - layer %3d: reuse layer %d, is_swa = %d\n", __func__, il, il_reuse, hparams.is_swa(il));
         }
     }
 
@@ -195,18 +197,6 @@ llama_kv_cache::llama_kv_cache(
 
     const char * LLAMA_KV_CACHE_DEBUG = getenv("LLAMA_KV_CACHE_DEBUG");
     debug = LLAMA_KV_CACHE_DEBUG ? atoi(LLAMA_KV_CACHE_DEBUG) : 0;
-
-    const char * LLAMA_SET_ROWS = getenv("LLAMA_SET_ROWS");
-    supports_set_rows = LLAMA_SET_ROWS ? atoi(LLAMA_SET_ROWS) != 0 : supports_set_rows;
-
-    if (!supports_set_rows) {
-        // ref: https://github.com/ggml-org/llama.cpp/pull/14363
-        GGML_ASSERT(unified && "cannot use non-unified KV cache without ggml_set_rows() support");
-    }
-
-    if (!supports_set_rows) {
-        LLAMA_LOG_WARN("%s: LLAMA_SET_ROWS=0, using old ggml_cpy() method for backwards compatibility\n", __func__);
-    }
 }
 
 void llama_kv_cache::clear(bool data) {
@@ -549,11 +539,8 @@ llama_kv_cache::slot_info_vec_t llama_kv_cache::prepare(const std::vector<llama_
     bool success = true;
 
     for (const auto & ubatch : ubatches) {
-        // non-continuous slots require support for ggml_set_rows()
-        const bool cont = supports_set_rows ? false : true;
-
         // only find a suitable slot for the ubatch. don't modify the cells yet
-        const auto sinfo_new = find_slot(ubatch, cont);
+        const auto sinfo_new = find_slot(ubatch, false);
         if (sinfo_new.empty()) {
             success = false;
             break;
@@ -769,8 +756,8 @@ llama_kv_cache::slot_info llama_kv_cache::find_slot(const llama_ubatch & ubatch,
             GGML_ASSERT(ubatch.seq_id  [s*n_tokens][0] == seq_id);
         }
 
-        res.s0 = std::min<llama_seq_id>(res.s0, seq_to_stream[seq_id]);
-        res.s1 = std::max<llama_seq_id>(res.s1, seq_to_stream[seq_id]);
+        res.s0 = std::min<uint32_t>(res.s0, seq_to_stream[seq_id]);
+        res.s1 = std::max<uint32_t>(res.s1, seq_to_stream[seq_id]);
 
         res.strm[s] = seq_to_stream[seq_id];
         res.idxs[s].reserve(n_tokens);
@@ -962,11 +949,11 @@ bool llama_kv_cache::get_has_shift() const {
     return result;
 }
 
-uint32_t llama_kv_cache::get_n_kv() const {
+uint32_t llama_kv_cache::get_n_kv(const slot_info & sinfo) const {
     uint32_t result = 0;
 
-    for (uint32_t s = 0; s < n_stream; ++s) {
-        const auto & cells = v_cells[s];
+    for (uint32_t s = 0; s < sinfo.n_stream(); ++s) {
+        const auto & cells = v_cells[sinfo.strm[s]];
 
         result = std::max(std::min(cells.size(), std::max(n_pad, GGML_PAD(cells.used_max_p1(), n_pad))), result);
     }
@@ -974,10 +961,6 @@ uint32_t llama_kv_cache::get_n_kv() const {
     return result;
 }
 
-bool llama_kv_cache::get_supports_set_rows() const {
-    return supports_set_rows;
-}
-
 ggml_tensor * llama_kv_cache::get_k(ggml_context * ctx, int32_t il, uint32_t n_kv, const slot_info & sinfo) const {
     const int32_t ikv = map_layer_ids.at(il);
 
@@ -1015,52 +998,42 @@ ggml_tensor * llama_kv_cache::get_v(ggml_context * ctx, int32_t il, uint32_t n_k
         // note: v->nb[1] <= v->nb[2]
         return ggml_view_4d(ctx, v,
                 hparams.n_embd_head_v, hparams.n_head_kv(il), n_kv, ns,
-                ggml_row_size(v->type, hparams.n_embd_head_v),            // v->nb[1]
-                ggml_row_size(v->type, n_embd_v_gqa),         // v->nb[2]
-                ggml_row_size(v->type, n_embd_v_gqa*kv_size), // v->nb[3]
+                ggml_row_size(v->type, hparams.n_embd_head_v),          // v->nb[1]
+                ggml_row_size(v->type, n_embd_v_gqa),                   // v->nb[2]
+                ggml_row_size(v->type, n_embd_v_gqa*kv_size),           // v->nb[3]
                 ggml_row_size(v->type, n_embd_v_gqa*kv_size)*sinfo.s0);
     }
 
     // note: v->nb[1] > v->nb[2]
     return ggml_view_4d(ctx, v,
             n_kv, hparams.n_head_kv(il), hparams.n_embd_head_v, ns,
-            ggml_row_size(v->type, kv_size*hparams.n_embd_head_v),    // v->nb[1]
-            ggml_row_size(v->type, kv_size),                          // v->nb[2]
-            ggml_row_size(v->type, kv_size*n_embd_v_gqa), // v->nb[3]
+            ggml_row_size(v->type, kv_size*hparams.n_embd_head_v),  // v->nb[1]
+            ggml_row_size(v->type, kv_size),                        // v->nb[2]
+            ggml_row_size(v->type, kv_size*n_embd_v_gqa),           // v->nb[3]
             ggml_row_size(v->type, kv_size*n_embd_v_gqa)*sinfo.s0);
 }
 
 ggml_tensor * llama_kv_cache::cpy_k(ggml_context * ctx, ggml_tensor * k_cur, ggml_tensor * k_idxs, int32_t il, const slot_info & sinfo) const {
+    GGML_UNUSED(sinfo);
+
     const int32_t ikv = map_layer_ids.at(il);
 
     auto * k = layers[ikv].k;
 
-    const int64_t n_embd_k_gqa = k->ne[0];
     const int64_t n_tokens = k_cur->ne[2];
 
     k_cur = ggml_reshape_2d(ctx, k_cur, k->ne[0], n_tokens);
 
-    if (k_idxs && supports_set_rows) {
-        if (k->ne[2] > 1) {
-            k = ggml_reshape_2d(ctx, k, k->ne[0], k->ne[1]*k->ne[2]);
-        }
-
-        return ggml_set_rows(ctx, k, k_cur, k_idxs);
+    if (k->ne[2] > 1) {
+        k = ggml_reshape_2d(ctx, k, k->ne[0], k->ne[1]*k->ne[2]);
     }
 
-    // TODO: fallback to old ggml_cpy() method for backwards compatibility
-    //       will be removed when ggml_set_rows() is adopted by all backends
-
-    GGML_ASSERT(n_stream == 1 && "n_stream > 1 not supported without LLAMA_SET_ROWS");
-
-    ggml_tensor * k_view = ggml_view_1d(ctx, k,
-            n_tokens*n_embd_k_gqa,
-            ggml_row_size(k->type, n_embd_k_gqa)*sinfo.head());
-
-    return ggml_cpy(ctx, k_cur, k_view);
+    return ggml_set_rows(ctx, k, k_cur, k_idxs);
 }
 
 ggml_tensor * llama_kv_cache::cpy_v(ggml_context * ctx, ggml_tensor * v_cur, ggml_tensor * v_idxs, int32_t il, const slot_info & sinfo) const {
+    GGML_UNUSED(sinfo);
+
     const int32_t ikv = map_layer_ids.at(il);
 
     auto * v = layers[ikv].v;
@@ -1070,48 +1043,25 @@ ggml_tensor * llama_kv_cache::cpy_v(ggml_context * ctx, ggml_tensor * v_cur, ggm
 
     v_cur = ggml_reshape_2d(ctx, v_cur, n_embd_v_gqa, n_tokens);
 
-    if (v_idxs && supports_set_rows) {
-        if (!v_trans) {
-            if (v->ne[2] > 1) {
-                v = ggml_reshape_2d(ctx, v, v->ne[0], v->ne[1]*v->ne[2]);
-            }
-
-            return ggml_set_rows(ctx, v, v_cur, v_idxs);
-        }
-
-        // [TAG_V_CACHE_VARIABLE]
-        if (n_embd_v_gqa < v->ne[0]) {
-            v_cur = ggml_pad(ctx, v_cur, v->ne[0] - n_embd_v_gqa, 0, 0, 0);
+    if (!v_trans) {
+        if (v->ne[2] > 1) {
+            v = ggml_reshape_2d(ctx, v, v->ne[0], v->ne[1]*v->ne[2]);
         }
 
-        // the row becomes a single element
-        ggml_tensor * v_view = ggml_reshape_2d(ctx, v, 1, v->ne[0]*v->ne[1]*v->ne[2]);
-
-        v_cur = ggml_reshape_2d(ctx, v_cur, 1, v_cur->ne[0]*v_cur->ne[1]);
-
-        return ggml_set_rows(ctx, v_view, v_cur, v_idxs);
+        return ggml_set_rows(ctx, v, v_cur, v_idxs);
     }
 
-    // TODO: fallback to old ggml_cpy() method for backwards compatibility
-    //       will be removed when ggml_set_rows() is adopted by all backends
-
-    GGML_ASSERT(n_stream == 1 && "n_stream > 1 not supported without LLAMA_SET_ROWS");
-
-    ggml_tensor * v_view = nullptr;
+    // [TAG_V_CACHE_VARIABLE]
+    if (n_embd_v_gqa < v->ne[0]) {
+        v_cur = ggml_pad(ctx, v_cur, v->ne[0] - n_embd_v_gqa, 0, 0, 0);
+    }
 
-    if (!v_trans) {
-        v_view = ggml_view_1d(ctx, v,
-                n_tokens*n_embd_v_gqa,
-                ggml_row_size(v->type, n_embd_v_gqa)*sinfo.head());
-    } else {
-        v_cur = ggml_transpose(ctx, v_cur);
+    // the row becomes a single element
+    ggml_tensor * v_view = ggml_reshape_2d(ctx, v, 1, v->ne[0]*v->ne[1]*v->ne[2]);
 
-        v_view = ggml_view_2d(ctx, v, n_tokens, n_embd_v_gqa,
-                (v->ne[1]    )*ggml_element_size(v),
-                (sinfo.head())*ggml_element_size(v));
-    }
+    v_cur = ggml_reshape_2d(ctx, v_cur, 1, v_cur->ne[0]*v_cur->ne[1]);
 
-    return ggml_cpy(ctx, v_cur, v_view);
+    return ggml_set_rows(ctx, v_view, v_cur, v_idxs);
 }
 
 ggml_tensor * llama_kv_cache::build_input_k_idxs(ggml_context * ctx, const llama_ubatch & ubatch) const {
@@ -1141,10 +1091,6 @@ ggml_tensor * llama_kv_cache::build_input_v_idxs(ggml_context * ctx, const llama
 }
 
 void llama_kv_cache::set_input_k_idxs(ggml_tensor * dst, const llama_ubatch * ubatch, const slot_info & sinfo) const {
-    if (!supports_set_rows) {
-        return;
-    }
-
     const uint32_t n_tokens = ubatch->n_tokens;
     GGML_ASSERT(n_tokens == (int64_t) sinfo.size()*sinfo.n_stream());
 
@@ -1161,10 +1107,6 @@ void llama_kv_cache::set_input_k_idxs(ggml_tensor * dst, const llama_ubatch * ub
 }
 
 void llama_kv_cache::set_input_v_idxs(ggml_tensor * dst, const llama_ubatch * ubatch, const slot_info & sinfo) const {
-    if (!supports_set_rows) {
-        return;
-    }
-
     const uint32_t n_tokens = ubatch->n_tokens;
     GGML_ASSERT(n_tokens == (int64_t) sinfo.size()*sinfo.n_stream());
 
@@ -1983,8 +1925,7 @@ bool llama_kv_cache_context::apply() {
     }
 
     kv->apply_ubatch(sinfos[i_cur], ubatches[i_cur]);
-
-    n_kv = kv->get_n_kv();
+    n_kv = kv->get_n_kv(sinfos[i_cur]);
 
     return true;
 }
@@ -2003,10 +1944,6 @@ uint32_t llama_kv_cache_context::get_n_kv() const {
     return n_kv;
 }
 
-bool llama_kv_cache_context::get_supports_set_rows() const {
-    return kv->get_supports_set_rows();
-}
-
 ggml_tensor * llama_kv_cache_context::get_k(ggml_context * ctx, int32_t il) const {
     return kv->get_k(ctx, il, n_kv, sinfos[i_cur]);
 }

package/src/llama.cpp/src/llama-kv-cache.h

@@ -21,9 +21,6 @@ class llama_kv_cache : public llama_memory_i {
 public:
     static uint32_t get_padding(const llama_cparams & cparams);
 
-    // this callback is used to filter out layers that should not be included in the cache
-    using layer_filter_cb = std::function<bool(int32_t il)>;
-
     struct stream_copy_info {
         bool empty() const {
             assert(ssrc.size() == sdst.size());
@@ -41,8 +38,8 @@ public:
         using idx_vec_t = std::vector<uint32_t>;
 
         // number of streams: ns = s1 - s0 + 1
-        llama_seq_id s0;
-        llama_seq_id s1;
+        uint32_t s0;
+        uint32_t s1;
 
         std::vector<llama_seq_id> strm; // [ns]
         std::vector<idx_vec_t>    idxs; // [ns]
@@ -82,18 +79,19 @@ public:
     using slot_info_vec_t = std::vector<slot_info>;
 
     llama_kv_cache(
-            const llama_model &  model,
-              layer_filter_cb && filter,
-                    ggml_type    type_k,
-                    ggml_type    type_v,
-                         bool    v_trans,
-                         bool    offload,
-                         bool    unified,
-                     uint32_t    kv_size,
-                     uint32_t    n_seq_max,
-                     uint32_t    n_pad,
-                     uint32_t    n_swa,
-               llama_swa_type    swa_type);
+            const llama_model & model,
+                    ggml_type   type_k,
+                    ggml_type   type_v,
+                         bool   v_trans,
+                         bool   offload,
+                         bool   unified,
+                     uint32_t   kv_size,
+                     uint32_t   n_seq_max,
+                     uint32_t   n_pad,
+                     uint32_t   n_swa,
+               llama_swa_type   swa_type,
+        const layer_filter_cb & filter,
+        const  layer_reuse_cb & reuse);
 
     ~llama_kv_cache() = default;
 
@@ -141,10 +139,7 @@ public:
     // graph_build API
     //
 
-    uint32_t get_n_kv() const;
-
-    // TODO: temporary
-    bool get_supports_set_rows() const;
+    uint32_t get_n_kv(const slot_info & sinfo) const;
 
     // get views of the current state of the cache
     ggml_tensor * get_k(ggml_context * ctx, int32_t il, uint32_t n_kv, const slot_info & sinfo) const;
@@ -217,10 +212,6 @@ private:
     // env: LLAMA_KV_CACHE_DEBUG
     int debug = 0;
 
-    // env: LLAMA_SET_ROWS (temporary)
-    // ref: https://github.com/ggml-org/llama.cpp/pull/14285
-    bool supports_set_rows = true;
-
     const llama_swa_type swa_type = LLAMA_SWA_TYPE_NONE;
 
     std::vector<ggml_context_ptr>        ctxs;
@@ -320,9 +311,6 @@ public:
 
     uint32_t get_n_kv() const;
 
-    // TODO: temporary
-    bool get_supports_set_rows() const;
-
     // get views of the current state of the cache
     ggml_tensor * get_k(ggml_context * ctx, int32_t il) const;
     ggml_tensor * get_v(ggml_context * ctx, int32_t il) const;

package/src/llama.cpp/src/llama-memory-hybrid.cpp

@@ -9,32 +9,29 @@
 //
 
 llama_memory_hybrid::llama_memory_hybrid(
-    const llama_model & model,
-                         /* attn */
-            ggml_type    type_k,
-            ggml_type    type_v,
-                 bool    v_trans,
-             uint32_t    kv_size,
-             uint32_t    n_pad,
-             uint32_t    n_swa,
-       llama_swa_type    swa_type,
-                         /* recurrent */
-            ggml_type    type_r,
-            ggml_type    type_s,
-             uint32_t    rs_size,
-                         /* common */
-             uint32_t    n_seq_max,
-                 bool    offload,
-                 bool    unified,
-                         /* layer filters */
-      layer_filter_cb && filter_attn,
-      layer_filter_cb && filter_recr) :
+        const llama_model & model,
+                            /* attn */
+                ggml_type   type_k,
+                ggml_type   type_v,
+                     bool   v_trans,
+                 uint32_t   kv_size,
+                 uint32_t   n_pad,
+                 uint32_t   n_swa,
+           llama_swa_type   swa_type,
+                            /* recurrent */
+                ggml_type   type_r,
+                ggml_type   type_s,
+                 uint32_t   rs_size,
+                            /* common */
+                 uint32_t   n_seq_max,
+                     bool   offload,
+                     bool   unified,
+                            /* layer filters */
+    const layer_filter_cb & filter_attn,
+    const layer_filter_cb & filter_recr) :
     hparams(model.hparams),
     mem_attn(new llama_kv_cache(
         model,
-        filter_attn == nullptr ?
-            [&](int32_t il) { return !hparams.is_recurrent(il); }
-            : filter_attn,
         type_k,
         type_v,
         v_trans,
@@ -44,18 +41,22 @@ llama_memory_hybrid::llama_memory_hybrid(
         n_seq_max,
         n_pad,
         n_swa,
-        swa_type
+        swa_type,
+        filter_attn == nullptr ?
+            [&](int32_t il) { return !hparams.is_recurrent(il); }
+            : filter_attn,
+        nullptr
     )),
     mem_recr(new llama_memory_recurrent(
         model,
-        filter_recr == nullptr ?
-            [&](int32_t il) { return hparams.is_recurrent(il); }
-            : filter_recr,
         type_r,
         type_s,
         offload,
         rs_size,
-        n_seq_max
+        n_seq_max,
+        filter_recr == nullptr ?
+            [&](int32_t il) { return hparams.is_recurrent(il); }
+            : filter_recr
     )) {}
 
 llama_memory_context_ptr llama_memory_hybrid::init_batch(llama_batch_allocr & balloc, uint32_t n_ubatch, bool embd_all) {

package/src/llama.cpp/src/llama-memory-hybrid.h

@@ -18,31 +18,27 @@
 
 class llama_memory_hybrid : public llama_memory_i {
 public:
-
-    // this callback is used to filter out layers that should not be included in the cache
-    using layer_filter_cb = std::function<bool(int32_t il)>;
-
     llama_memory_hybrid(
         const llama_model & model,
                             /* attn */
-                ggml_type    type_k,
-                ggml_type    type_v,
-                     bool    v_trans,
-                 uint32_t    kv_size,
-                 uint32_t    n_pad,
-                 uint32_t    n_swa,
-           llama_swa_type    swa_type,
-                             /* recurrent */
-                ggml_type    type_r,
-                ggml_type    type_s,
-                 uint32_t    rs_size,
-                             /* common */
-                 uint32_t    n_seq_max,
-                     bool    offload,
-                     bool    unified,
-                             /* layer filters */
-          layer_filter_cb && filter_attn = nullptr,
-          layer_filter_cb && filter_recr = nullptr);
+                ggml_type   type_k,
+                ggml_type   type_v,
+                     bool   v_trans,
+                 uint32_t   kv_size,
+                 uint32_t   n_pad,
+                 uint32_t   n_swa,
+           llama_swa_type   swa_type,
+                            /* recurrent */
+                ggml_type   type_r,
+                ggml_type   type_s,
+                 uint32_t   rs_size,
+                            /* common */
+                 uint32_t   n_seq_max,
+                     bool   offload,
+                     bool   unified,
+                            /* layer filters */
+    const layer_filter_cb & filter_attn = nullptr,
+    const layer_filter_cb & filter_recr = nullptr);
 
     ~llama_memory_hybrid() = default;
 

package/src/llama.cpp/src/llama-memory-recurrent.cpp

@@ -16,13 +16,13 @@
 //
 
 llama_memory_recurrent::llama_memory_recurrent(
-        const llama_model &  model,
-          layer_filter_cb && filter,
-                ggml_type    type_r,
-                ggml_type    type_s,
-                     bool    offload,
-                 uint32_t    mem_size,
-                 uint32_t    n_seq_max) : hparams(model.hparams), n_seq_max(n_seq_max) {
+        const llama_model & model,
+                ggml_type   type_r,
+                ggml_type   type_s,
+                     bool   offload,
+                 uint32_t   mem_size,
+                 uint32_t   n_seq_max,
+    const layer_filter_cb & filter) : hparams(model.hparams), n_seq_max(n_seq_max) {
     const int32_t n_layer = hparams.n_layer;
 
     head = 0;

package/src/llama.cpp/src/llama-memory-recurrent.h

@@ -15,18 +15,14 @@
 //       see the implementation of llama_kv_cache_context_i for an example how to do it
 class llama_memory_recurrent : public llama_memory_i {
 public:
-
-    // this callback is used to filter out layers that should not be included in the cache
-    using layer_filter_cb = std::function<bool(int32_t il)>;
-
     llama_memory_recurrent(
-            const llama_model &  model,
-              layer_filter_cb && filter,
-                    ggml_type    type_r,
-                    ggml_type    type_s,
-                         bool    offload,
-                     uint32_t    mem_size,
-                     uint32_t    n_seq_max);
+            const llama_model & model,
+                    ggml_type   type_r,
+                    ggml_type   type_s,
+                         bool   offload,
+                     uint32_t   mem_size,
+                     uint32_t   n_seq_max,
+        const layer_filter_cb & filter);
 
     ~llama_memory_recurrent() = default;
 

package/src/llama.cpp/src/llama-memory.h

@@ -3,6 +3,7 @@
 #include "llama.h"
 
 #include <memory>
+#include <functional>
 
 struct llama_ubatch;
 
@@ -64,6 +65,13 @@ using llama_memory_context_ptr = std::unique_ptr<llama_memory_context_i>;
 // general concept of LLM memory
 // the KV cache is a type of LLM memory, but there can be other types
 struct llama_memory_i {
+    // this callback is used to filter out layers that should not be included in the cache
+    using layer_filter_cb = std::function<bool(int32_t il)>;
+
+    // this callback is used to specify which layers should reuse memory from other layers
+    // return negative value to indicate that the layer il should not reuse memory
+    using layer_reuse_cb = std::function<int32_t(int32_t il)>;
+
     virtual ~llama_memory_i() = default;
 
     // split the input batch into a set of ubatches and verify that they can fit into the cache

package/src/llama.cpp/src/llama-model-loader.cpp

@@ -788,6 +788,7 @@ const struct ggml_tensor * llama_model_loader::check_tensor_dims(const std::stri
 }
 
 struct ggml_tensor * llama_model_loader::create_tensor(struct ggml_context * ctx, const std::string & name, const std::initializer_list<int64_t> & ne, int flags) {
+    LLAMA_LOG_DEBUG("%s: loading tensor %s\n", __func__, name.c_str());
     const struct ggml_tensor * cur = check_tensor_dims(name, ne, !(flags & TENSOR_NOT_REQUIRED));
 
     if (cur == NULL) {