cui-llama.rn 1.6.0 → 1.7.0

package/cpp/llama-kv-cache.h

@@ -2,32 +2,75 @@
 
 #include "llama.h"
 #include "llama-io.h"
+#include "llama-graph.h"
 #include "llama-memory.h"
 
 #include "ggml-cpp.h"
 
-#include <functional>
 #include <set>
+#include <unordered_map>
 #include <vector>
 
 struct llama_cparams;
 struct llama_hparams;
 struct llama_ubatch;
+struct llama_sbatch;
+struct llama_model;
+struct llama_context;
 
 struct llama_kv_cache : public llama_memory_i {
-    using llama_memory_i::llama_memory_i;
+    virtual ~llama_kv_cache() = default;
 
-    virtual void restore() = 0; // call if batch processing fails - restores the cache state
-    virtual void commit() = 0;  // call after successful batch processing - clears any pending state
+    // call if batch processing fails - restores the cache state
+    virtual void restore() = 0;
 
-    virtual int32_t get_n_tokens()   const = 0;
-    virtual int32_t get_used_cells() const = 0; // TODO: remove, this is too-specific to the unified cache
+    // call after successful batch processing - clears any pending state
+    virtual void commit()  = 0;
 
+    // process any pending defrag/shift/etc. operations
+    // optionally call once before processing a new batch
+    virtual bool update(llama_context & lctx) = 0;
+
+    // schedule a defrag if the fragmentation threshold is exceeded. otherwise, do nothing
+    virtual void defrag_sched(float thold) = 0;
+
+    // simulate full cache, used for allocating worst-case compute buffers
+    virtual void set_full() = 0;
+
+    //
+    // batch processing
+    //
+
+    // =============================================================================================================
+    // TODO: refactor  and simplify this
+
+    virtual llama_sbatch sbatch_init(const llama_batch & batch, bool logits_all) = 0;
+
+    // different KV caches require different batch splitting strategies
+    virtual llama_ubatch ubatch_next(llama_sbatch & sbatch, uint32_t n_ubatch, bool embd_pooled) const = 0;
+
+    // find an empty slot of size "n_tokens" in the cache
+    virtual bool find_slot(const llama_ubatch & batch) = 0;
+
+    // =============================================================================================================
+
+    // getters
     virtual bool get_can_shift() const = 0;
 
     bool get_can_edit() const override { return get_can_shift(); }
+
+    //
+    // state write/read
+    //
+
+    virtual void state_write(llama_io_write_i & io, llama_seq_id seq_id = -1) const = 0;
+    virtual void state_read (llama_io_read_i  & io, llama_seq_id seq_id = -1) = 0;
 };
 
+//
+// llama_kv_cache_guard
+//
+
 struct llama_kv_cache_guard {
     llama_kv_cache_guard(llama_kv_cache * kv) : kv(kv) {}
 
@@ -43,171 +86,430 @@ private:
     llama_kv_cache * kv;
 };
 
-struct llama_kv_cell {
-    llama_pos pos   = -1;
-    llama_pos delta =  0;
-    int32_t   src   = -1; // used by recurrent state models to copy states
-    int32_t   tail  = -1;
+//
+// llama_kv_cache_unified
+//
 
-    std::set<llama_seq_id> seq_id;
+class llama_kv_cache_unified : public llama_kv_cache {
+public:
+    static uint32_t get_padding(const llama_cparams & cparams);
 
-    bool has_seq_id(const llama_seq_id & id) const {
-        return seq_id.find(id) != seq_id.end();
-    }
+    // 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)>;
 
-    bool is_empty() const {
-        return seq_id.empty();
-    }
+    llama_kv_cache_unified(
+            const llama_model &  model,
+              layer_filter_cb && filter,
+                    lm_ggml_type    type_k,
+                    lm_ggml_type    type_v,
+                         bool    v_trans,
+                         bool    offload,
+                     uint32_t    kv_size,
+                     uint32_t    n_seq_max,
+                     uint32_t    n_pad,
+                     uint32_t    n_swa,
+               llama_swa_type    swa_type);
+
+    ~llama_kv_cache_unified() = default;
+
+    //
+    // llama_memory_i
+    //
 
-    bool is_same_seq(const llama_kv_cell & other) const {
-        return seq_id == other.seq_id;
-    }
-};
+    void clear() override;
 
-// ring-buffer of cached KV data
-// TODO: pimpl
-// TODO: add notion of max sequences
-class llama_kv_cache_unified : public llama_kv_cache {
-public:
-    // can be used to query data from the model if needed
-    struct callbacks {
-        std::function<lm_ggml_tensor * (uint32_t n_ctx_per_seq, int il)> get_rope_factors;
+    bool seq_rm  (llama_seq_id seq_id,                              llama_pos p0, llama_pos p1) override;
+    void seq_cp  (llama_seq_id seq_id_src, llama_seq_id seq_id_dst, llama_pos p0, llama_pos p1) override;
+    void seq_keep(llama_seq_id seq_id)                                                          override;
+    void seq_add (llama_seq_id seq_id,                              llama_pos p0, llama_pos p1, llama_pos delta) override;
+    void seq_div (llama_seq_id seq_id,                              llama_pos p0, llama_pos p1, int d) override;
+
+    llama_pos seq_pos_min(llama_seq_id seq_id) const override;
+    llama_pos seq_pos_max(llama_seq_id seq_id) const override;
+
+    //
+    // llama_kv_cache
+    //
+
+    void restore() override;
+    void commit()  override;
+
+    bool update(llama_context & ctx) override;
+
+    void defrag_sched(float thold) override;
+
+    void set_full() override;
+
+    llama_sbatch sbatch_init(const llama_batch & batch, bool logits_all) override;
+    llama_ubatch ubatch_next(llama_sbatch & sbatch, uint32_t n_ubatch, bool embd_pooled) const override;
+
+    // updates the cache head
+    // Note: On success, it's important that cache.head points
+    // to the first cell of the slot.
+    bool find_slot(const llama_ubatch & batch) override;
+
+    bool get_can_shift() const override;
+
+    // state write/load
+
+    void state_write(llama_io_write_i & io, llama_seq_id seq_id = -1) const override;
+    void state_read (llama_io_read_i  & io, llama_seq_id seq_id = -1)       override;
+
+    //
+    // llama_kv_cache_unified specific API
+    //
+
+    uint32_t get_n() const;
+    uint32_t get_size() const;
+
+    // get views of the current state of the cache
+    lm_ggml_tensor * get_k(lm_ggml_context * ctx, int32_t il) const;
+    lm_ggml_tensor * get_v(lm_ggml_context * ctx, int32_t il) const;
+
+    // store k_cur and v_cur in the cache based on the current head location
+    lm_ggml_tensor * cpy_k(lm_ggml_context * ctx, lm_ggml_tensor * k_cur, int32_t il) const;
+    lm_ggml_tensor * cpy_v(lm_ggml_context * ctx, lm_ggml_tensor * v_cur, int32_t il) const;
+
+    void prune_swa(llama_seq_id seq_id, llama_pos pmin, llama_pos pmax);
+
+    void set_input_kq_mask   (lm_ggml_tensor * dst, const llama_ubatch * ubatch, bool causal_attn) const;
+    void set_input_k_shift   (lm_ggml_tensor * dst) const;
+    void set_input_pos_bucket(lm_ggml_tensor * dst, const llama_ubatch * ubatch) const;
+
+private:
+    const llama_model & model;
+    const llama_hparams & hparams;
+
+    struct kv_cell {
+        llama_pos pos   = -1;
+        llama_pos delta =  0;
+
+        // TODO: replace with bitset uint64_t
+        std::set<llama_seq_id> seq_id;
+
+        bool has_seq_id(const llama_seq_id & id) const {
+            return seq_id.find(id) != seq_id.end();
+        }
+
+        bool is_empty() const {
+            return seq_id.empty();
+        }
+
+        bool is_same_seq(const kv_cell & other) const {
+            return seq_id == other.seq_id;
+        }
     };
 
-    llama_kv_cache_unified(
-            const llama_hparams & hparams,
-            callbacks             cbs);
+    struct kv_layer {
+        // layer index in the model
+        // note: can be different from the layer index in the KV cache
+        uint32_t il;
+
+        lm_ggml_tensor * k;
+        lm_ggml_tensor * v;
+    };
+
+    bool has_shift = false;
+    bool do_defrag = false;
+    bool v_trans   = true;  // the value tensor is transposed
+
+    uint32_t head = 0; // the location where the batch will be placed in the cache (see find_slot())
+    uint32_t size = 0; // total number of cells, shared across all sequences
+    uint32_t used = 0; // used cells (i.e. at least one seq_id) (TODO: add `struct kv_cells` and keep track automaticallt)
+
+    // computed before each graph build
+    uint32_t n = 0;
+
+    const uint32_t n_seq_max = 1;
+
+    // required padding
+    const uint32_t n_pad = 1;
+
+    // SWA
+    const uint32_t n_swa = 0;
+
+    const llama_swa_type swa_type = LLAMA_SWA_TYPE_NONE;
+
+    std::vector<lm_ggml_context_ptr>        ctxs;
+    std::vector<lm_ggml_backend_buffer_ptr> bufs;
+
+    std::vector<kv_cell>  cells;  // TODO: replace with `struct kv_cells`
+    std::vector<kv_layer> layers;
 
-    virtual ~llama_kv_cache_unified() = default;
+    // model layer id -> KV cache layer id
+    std::unordered_map<int32_t, int32_t> map_layer_ids;
+
+    // recovery information used to restore the KV cells to their original state in case of a failure
+    struct {
+        void clear() {
+            cells.clear();
+        }
+
+        std::unordered_map<uint32_t, kv_cell> cells;
+    } recovery;
+
+    // defrag
+    struct {
+        std::vector<uint32_t> ids;
+    } defrag_info;
+
+    // return true if cells have been moved
+    bool defrag_prepare(int32_t n_max_nodes);
+
+    // find how many cells are currently in use
+    uint32_t cell_max() const;
+
+    size_t total_size() const;
+
+    size_t size_k_bytes() const;
+    size_t size_v_bytes() const;
+
+    bool is_masked_swa(llama_pos p0, llama_pos p1) const;
+
+    lm_ggml_tensor * build_rope_shift(
+            const llama_cparams & cparams,
+                   lm_ggml_context * ctx,
+                    lm_ggml_tensor * cur,
+                    lm_ggml_tensor * shift,
+                    lm_ggml_tensor * factors,
+                          float   freq_base,
+                          float   freq_scale) const;
+
+    llm_graph_result_ptr build_graph_shift(
+            const llama_cparams & cparams,
+                   lm_ggml_context * ctx,
+                    lm_ggml_cgraph * gf) const;
+
+    llm_graph_result_ptr build_graph_defrag(
+            const llama_cparams & cparams,
+                   lm_ggml_context * ctx,
+                    lm_ggml_cgraph * gf) const;
+
+    void state_write_meta(llama_io_write_i & io, const std::vector<std::pair<uint32_t, uint32_t>> & cell_ranges, llama_seq_id seq_id = -1) const;
+    void state_write_data(llama_io_write_i & io, const std::vector<std::pair<uint32_t, uint32_t>> & cell_ranges) const;
 
-    // TODO: become constructor
-    bool init(
-            const llama_model & model,   // TODO: do not reference the model
-          const llama_cparams & cparams,
+    bool state_read_meta(llama_io_read_i & io, uint32_t cell_count, llama_seq_id dest_seq_id = -1);
+    bool state_read_data(llama_io_read_i & io, uint32_t cell_count);
+};
+
+//
+// llama_kv_cache_unified_iswa
+//
+
+// utilizes two instances of llama_kv_cache_unified
+//   the first instance is for the non-SWA layers of the model and the second instance is for the SWA layers
+//   upon successful commit, the SWA cache removes old tokens outside the n_swa window
+
+class llama_kv_cache_unified_iswa : public llama_kv_cache {
+public:
+    llama_kv_cache_unified_iswa(
+            const llama_model & model,
                     lm_ggml_type   type_k,
                     lm_ggml_type   type_v,
+                         bool   v_trans,
+                         bool   offload,
+                         bool   swa_full,
                      uint32_t   kv_size,
-                         bool   offload);
+                     uint32_t   n_seq_max,
+                     uint32_t   n_batch,
+                     uint32_t   n_pad);
 
-    int32_t get_n_tokens()   const override;
-    int32_t get_used_cells() const override;
-
-    size_t total_size() const;
+    ~llama_kv_cache_unified_iswa() = default;
 
-    // TODO: better data structures to reduce the cost of this operation
-    llama_pos pos_max() const;
+    //
+    // llama_memory_i
+    //
 
     void clear() override;
-    void defrag() override;
-
-    virtual void restore() override;
-    virtual void commit() override;
 
     bool seq_rm  (llama_seq_id seq_id,                              llama_pos p0, llama_pos p1) override;
     void seq_cp  (llama_seq_id seq_id_src, llama_seq_id seq_id_dst, llama_pos p0, llama_pos p1) override;
-    void seq_keep(llama_seq_id seq_id) override;
+    void seq_keep(llama_seq_id seq_id)                                                          override;
     void seq_add (llama_seq_id seq_id,                              llama_pos p0, llama_pos p1, llama_pos delta) override;
     void seq_div (llama_seq_id seq_id,                              llama_pos p0, llama_pos p1, int d) override;
 
+    llama_pos seq_pos_min(llama_seq_id seq_id) const override;
     llama_pos seq_pos_max(llama_seq_id seq_id) const override;
 
+    //
+    // llama_kv_cache
+    //
+
+    void restore() override;
+    void commit()  override;
+
+    bool update(llama_context & ctx) override;
+
+    void defrag_sched(float thold) override;
+
+    void set_full() override;
+
+    llama_sbatch sbatch_init(const llama_batch & batch, bool logits_all) override;
+    llama_ubatch ubatch_next(llama_sbatch & sbatch, uint32_t n_ubatch, bool embd_pooled) const override;
+
+    bool find_slot(const llama_ubatch & batch) override;
+
     bool get_can_shift() const override;
 
-    // find an empty slot of size "n_tokens" in the cache
-    // updates the cache head
-    // Note: On success, it's important that cache.head points
-    // to the first cell of the slot.
-    bool find_slot(const llama_ubatch & batch);
+    // state write/load
 
-    // TODO: maybe not needed
-    uint32_t get_padding(const llama_cparams & cparams) const;
+    void state_write(llama_io_write_i & io, llama_seq_id seq_id = -1) const override;
+    void state_read (llama_io_read_i  & io, llama_seq_id seq_id = -1)       override;
 
-    // find how many cells are currently in use
-    uint32_t cell_max() const;
+    //
+    // llama_kv_cache_unified_iswa specific API
+    //
 
-    size_t size_k_bytes() const;
-    size_t size_v_bytes() const;
+    llama_kv_cache_unified * get_kv_base() const;
+    llama_kv_cache_unified * get_kv_swa () const;
 
-    // defrag
+private:
+    const llama_hparams & hparams;
+
+    bool do_prune = true;
 
     struct {
-        std::vector<uint32_t> ids;
-    } defrag_info;
+        struct entry {
+            llama_pos pmin;
+            llama_pos pmax;
+        };
 
-    // return true if cells have been moved
-    bool defrag_prepare(int32_t n_max_nodes);
+        void clear() {
+            pos.clear();
+        }
 
-    // commit/restore cache
+        // used to perform SWA pruning of old tokens
+        std::unordered_map<llama_seq_id, entry> pos;
+    } pending;
 
-    struct slot_range {
-        uint32_t c0 = 0; // note: these are cell indices, not sequence positions
-        uint32_t c1 = 0;
+    std::unique_ptr<llama_kv_cache_unified> kv_base;
+    std::unique_ptr<llama_kv_cache_unified> kv_swa;
+};
+
+//
+// llama_kv_cache_recurrent
+//
+
+class llama_kv_cache_recurrent : public llama_kv_cache {
+public:
+    struct kv_cell {
+        llama_pos pos  = -1;
+        int32_t   src  = -1; // used to copy states
+        int32_t   tail = -1;
+
+        std::set<llama_seq_id> seq_id;
+
+        bool has_seq_id(const llama_seq_id & id) const {
+            return seq_id.find(id) != seq_id.end();
+        }
+
+        bool is_empty() const {
+            return seq_id.empty();
+        }
+
+        bool is_same_seq(const kv_cell & other) const {
+            return seq_id == other.seq_id;
+        }
     };
 
-    // pending cell updates that are not yet committed
-    struct {
-        std::vector<slot_range> ranges;
-    } pending;
+    llama_kv_cache_recurrent(
+            const llama_model & model,
+                    lm_ggml_type   type_k,
+                    lm_ggml_type   type_v,
+                         bool   offload,
+                     uint32_t   kv_size,
+                     uint32_t   n_seq_max);
 
-    // state write/load
+    ~llama_kv_cache_recurrent() = default;
 
-    void state_write(llama_io_write_i & io, llama_seq_id seq_id = -1) const;
-    void state_read (llama_io_read_i  & io, llama_seq_id seq_id = -1);
+    //
+    // llama_memory_i
+    //
 
-    // members
+    void clear() override;
 
-    const llama_hparams & hparams;
+    bool seq_rm  (llama_seq_id seq_id,                              llama_pos p0, llama_pos p1) override;
+    void seq_cp  (llama_seq_id seq_id_src, llama_seq_id seq_id_dst, llama_pos p0, llama_pos p1) override;
+    void seq_keep(llama_seq_id seq_id)                                                          override;
+    void seq_add (llama_seq_id seq_id,                              llama_pos p0, llama_pos p1, llama_pos delta) override;
+    void seq_div (llama_seq_id seq_id,                              llama_pos p0, llama_pos p1, int d) override;
 
-    callbacks cbs;
+    llama_pos seq_pos_min(llama_seq_id seq_id) const override;
+    llama_pos seq_pos_max(llama_seq_id seq_id) const override;
 
-    bool has_shift = false;
-    bool do_defrag = false;
+    //
+    // llama_kv_cache
+    //
 
-    // TODO: remove this and implement llama_kv_cache_recurrent instead
-    bool recurrent = false; // with recurrent state models, a cell can hold the state for more than one past token
+    void restore() override;
+    void commit()  override;
 
-    bool v_trans   = true;  // the value tensor is transposed
-    bool can_shift = false;
+    bool update(llama_context & ctx) override;
+
+    void defrag_sched(float thold) override;
+
+    void set_full() override;
 
-    // Note: The value of head isn't only used to optimize searching
-    // for a free KV slot. llama_decode_impl also uses it, so it
-    // cannot be freely changed after a slot has been allocated.
-    uint32_t head = 0;
-    uint32_t size = 0;
+    llama_sbatch sbatch_init(const llama_batch & batch, bool logits_all) override;
+    llama_ubatch ubatch_next(llama_sbatch & sbatch, uint32_t n_ubatch, bool embd_pooled) const override;
+
+    bool find_slot(const llama_ubatch & batch) override;
+
+    bool get_can_shift() const override;
+
+    // TODO: temporary methods - they are not really const as they do const_cast<>, fix this
+    int32_t s_copy(int i) const;
+    float   s_mask(int i) const;
+
+    // state write/load
+
+    void state_write(llama_io_write_i & io, llama_seq_id seq_id = -1) const override;
+    void state_read (llama_io_read_i  & io, llama_seq_id seq_id = -1) override;
+
+    uint32_t head = 0; // the location where the batch will be placed in the cache (see find_slot())
+    uint32_t size = 0; // total number of cells, shared across all sequences
     uint32_t used = 0; // used cells (i.e. at least one seq_id)
 
     // computed before each graph build
     uint32_t n = 0;
 
-    std::vector<llama_kv_cell> cells;
+    std::vector<kv_cell> cells;
 
     std::vector<lm_ggml_tensor *> k_l; // per layer
     std::vector<lm_ggml_tensor *> v_l;
 
 private:
-    lm_ggml_type type_k = LM_GGML_TYPE_F16;
-    lm_ggml_type type_v = LM_GGML_TYPE_F16;
+    //const llama_model & model;
+    const llama_hparams & hparams;
+
+    // commit/restore cache
+    // TODO: rework for recurrent cache
+    struct slot_range {
+        uint32_t c0 = 0; // note: these are cell indices, not sequence positions
+        uint32_t c1 = 0;
+    };
+
+    // pending cell updates that are not yet committed
+    struct {
+        std::vector<slot_range> ranges;
+    } pending;
+
+    const uint32_t n_seq_max = 1;
 
     std::vector<lm_ggml_context_ptr>        ctxs;
     std::vector<lm_ggml_backend_buffer_ptr> bufs;
 
+    // find how many cells are currently in use
+    uint32_t cell_max() const;
+
+    size_t total_size() const;
+
+    size_t size_k_bytes() const;
+    size_t size_v_bytes() const;
+
     void state_write_meta(llama_io_write_i & io, const std::vector<std::pair<uint32_t, uint32_t>> & cell_ranges, llama_seq_id seq_id = -1) const;
     void state_write_data(llama_io_write_i & io, const std::vector<std::pair<uint32_t, uint32_t>> & cell_ranges) const;
 
     bool state_read_meta(llama_io_read_i & io, uint32_t cell_count, llama_seq_id dest_seq_id = -1);
     bool state_read_data(llama_io_read_i & io, uint32_t cell_count);
 };
-
-// TODO: temporary reusing llama_kv_cache_unified -- implement recurrent cache and simplify llama_kv_cache_unified
-//class llama_kv_cache_recurrent : public llama_kv_cache_unified {
-//public:
-//    using llama_kv_cache_unified::llama_kv_cache_unified;
-//};
-
-//
-// kv cache view
-//
-
-llama_kv_cache_view llama_kv_cache_view_init(const llama_kv_cache & kv, int32_t n_seq_max);
-
-void llama_kv_cache_view_update(llama_kv_cache_view * view, const llama_kv_cache * kv);

package/cpp/llama-memory.h

@@ -2,12 +2,22 @@
 
 #include "llama.h"
 
+struct llama_memory_params {
+    // kv cache
+    lm_ggml_type type_k;
+    lm_ggml_type type_v;
+
+    // use full-size SWA cache
+    bool swa_full;
+};
+
 // general concept of LLM memory
 // the KV cache is a type of LLM memory, but there can be other types
 class llama_memory_i {
 public:
+    virtual ~llama_memory_i() = default;
+
     virtual void clear() = 0;
-    virtual void defrag() = 0;
 
     virtual bool seq_rm  (llama_seq_id seq_id,                              llama_pos p0, llama_pos p1) = 0;
     virtual void seq_cp  (llama_seq_id seq_id_src, llama_seq_id seq_id_dst, llama_pos p0, llama_pos p1) = 0;
@@ -15,6 +25,7 @@ public:
     virtual void seq_add (llama_seq_id seq_id,                              llama_pos p0, llama_pos p1, llama_pos delta) = 0;
     virtual void seq_div (llama_seq_id seq_id,                              llama_pos p0, llama_pos p1, int d) = 0;
 
+    virtual llama_pos seq_pos_min(llama_seq_id seq_id) const = 0;
     virtual llama_pos seq_pos_max(llama_seq_id seq_id) const = 0;
 
     virtual bool get_can_edit() const = 0;