llama : add llama_memory_can_rm_suffix()

This commit is contained in:
Georgi Gerganov 2026-02-02 15:18:01 +02:00
parent 6fdddb4987
commit d30e59b62a
No known key found for this signature in database
GPG Key ID: 449E073F9DC10735
14 changed files with 49 additions and 11 deletions

View File

@ -760,6 +760,9 @@ extern "C" {
// Check if the memory supports shifting // Check if the memory supports shifting
LLAMA_API bool llama_memory_can_shift(llama_memory_t mem); LLAMA_API bool llama_memory_can_shift(llama_memory_t mem);
// Check if the memory supports removing the last tokens in the sequence
LLAMA_API bool llama_memory_can_rm_suffix(llama_memory_t mem);
// //
// State / sessions // State / sessions
// //

View File

@ -3360,6 +3360,14 @@ bool llama_memory_can_shift(llama_memory_t mem) {
return mem->get_can_shift(); return mem->get_can_shift();
} }
bool llama_memory_can_rm_suffix(llama_memory_t mem) {
if (!mem) {
return false;
}
return mem->get_can_rm_suffix();
}
// llama state API // llama state API
// deprecated // deprecated

View File

@ -221,6 +221,10 @@ bool llama_kv_cache_iswa::get_can_shift() const {
return kv_base->get_size() == kv_swa->get_size(); return kv_base->get_size() == kv_swa->get_size();
} }
bool llama_kv_cache_iswa::get_can_rm_suffix() const {
return kv_base->get_can_rm_suffix() && kv_swa->get_can_rm_suffix();
}
void llama_kv_cache_iswa::state_write(llama_io_write_i & io, llama_seq_id seq_id, llama_state_seq_flags flags) const { void llama_kv_cache_iswa::state_write(llama_io_write_i & io, llama_seq_id seq_id, llama_state_seq_flags flags) const {
if ((flags & LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY) == 0) { if ((flags & LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY) == 0) {
kv_base->state_write(io, seq_id, flags); kv_base->state_write(io, seq_id, flags);

View File

@ -43,7 +43,8 @@ public:
llama_memory_context_ptr init_update(llama_context * lctx, bool optimize) override; llama_memory_context_ptr init_update(llama_context * lctx, bool optimize) override;
bool get_can_shift() const override; bool get_can_shift() const override;
bool get_can_rm_suffix() const override;
void clear(bool data) override; void clear(bool data) override;

View File

@ -977,6 +977,10 @@ bool llama_kv_cache::get_can_shift() const {
return true; return true;
} }
bool llama_kv_cache::get_can_rm_suffix() const {
return true;
}
uint32_t llama_kv_cache::get_size() const { uint32_t llama_kv_cache::get_size() const {
const auto & cells = v_cells[seq_to_stream[0]]; const auto & cells = v_cells[seq_to_stream[0]];

View File

@ -123,7 +123,8 @@ public:
llama_memory_context_ptr init_update(llama_context * lctx, bool optimize) override; llama_memory_context_ptr init_update(llama_context * lctx, bool optimize) override;
bool get_can_shift() const override; bool get_can_shift() const override;
bool get_can_rm_suffix() const override;
void clear(bool data) override; void clear(bool data) override;

View File

@ -126,8 +126,11 @@ llama_memory_context_ptr llama_memory_hybrid_iswa::init_update(llama_context * l
} }
bool llama_memory_hybrid_iswa::get_can_shift() const { bool llama_memory_hybrid_iswa::get_can_shift() const {
// Shifting is trivially supported for recurrent return mem_attn->get_can_shift() && mem_recr->get_can_shift();
return mem_attn->get_can_shift(); }
bool llama_memory_hybrid_iswa::get_can_rm_suffix() const {
return mem_attn->get_can_rm_suffix() && mem_recr->get_can_rm_suffix();
} }
void llama_memory_hybrid_iswa::clear(bool data) { void llama_memory_hybrid_iswa::clear(bool data) {

View File

@ -55,7 +55,8 @@ public:
llama_memory_context_ptr init_update(llama_context * lctx, bool optimize) override; llama_memory_context_ptr init_update(llama_context * lctx, bool optimize) override;
bool get_can_shift() const override; bool get_can_shift() const override;
bool get_can_rm_suffix() const override;
void clear(bool data) override; void clear(bool data) override;

View File

@ -120,8 +120,11 @@ llama_memory_context_ptr llama_memory_hybrid::init_update(llama_context * lctx,
} }
bool llama_memory_hybrid::get_can_shift() const { bool llama_memory_hybrid::get_can_shift() const {
// Shifting is trivially supported for recurrent return mem_attn->get_can_shift() && mem_recr->get_can_shift();
return mem_attn->get_can_shift(); }
bool llama_memory_hybrid::get_can_rm_suffix() const {
return mem_attn->get_can_rm_suffix() && mem_recr->get_can_rm_suffix();
} }
void llama_memory_hybrid::clear(bool data) { void llama_memory_hybrid::clear(bool data) {

View File

@ -55,7 +55,8 @@ public:
llama_memory_context_ptr init_update(llama_context * lctx, bool optimize) override; llama_memory_context_ptr init_update(llama_context * lctx, bool optimize) override;
bool get_can_shift() const override; bool get_can_shift() const override;
bool get_can_rm_suffix() const override;
void clear(bool data) override; void clear(bool data) override;

View File

@ -665,6 +665,10 @@ bool llama_memory_recurrent::get_can_shift() const {
return true; return true;
} }
bool llama_memory_recurrent::get_can_rm_suffix() const {
return false;
}
size_t llama_memory_recurrent::total_size() const { size_t llama_memory_recurrent::total_size() const {
size_t size = 0; size_t size = 0;
for (const auto & [_, buf] : ctxs_bufs) { for (const auto & [_, buf] : ctxs_bufs) {

View File

@ -58,7 +58,9 @@ public:
// find a contiguous slot of memory cells and emplace the ubatch there // find a contiguous slot of memory cells and emplace the ubatch there
bool find_slot(const llama_ubatch & ubatch); bool find_slot(const llama_ubatch & ubatch);
bool get_can_shift() const override; bool get_can_shift() const override;
bool get_can_rm_suffix() const override;
// state write/load // state write/load

View File

@ -91,7 +91,8 @@ struct llama_memory_i {
virtual llama_memory_context_ptr init_update(llama_context * lctx, bool optimize) = 0; virtual llama_memory_context_ptr init_update(llama_context * lctx, bool optimize) = 0;
// getters // getters
virtual bool get_can_shift() const = 0; virtual bool get_can_shift() const = 0;
virtual bool get_can_rm_suffix() const = 0;
// //
// ops // ops

View File

@ -752,7 +752,7 @@ private:
slot.prompt.tokens.has_mtmd = mctx != nullptr; slot.prompt.tokens.has_mtmd = mctx != nullptr;
// try speculative decoding // try speculative decoding
{ if (llama_memory_can_rm_suffix(llama_get_memory(ctx))) {
slot.spec = common_speculative_init(params_base.speculative, slot.ctx); slot.spec = common_speculative_init(params_base.speculative, slot.ctx);
if (slot.spec) { if (slot.spec) {
if (mctx) { if (mctx) {
@ -763,6 +763,8 @@ private:
} else { } else {
SLT_INF(slot, "%s", "speculative decoding context not initialized\n"); SLT_INF(slot, "%s", "speculative decoding context not initialized\n");
} }
} else {
SLT_WRN(slot, "%s", "speculative decoding not supported by this context (no memory_rm_suffix support)\n");
} }
SLT_INF(slot, "new slot, n_ctx = %d\n", slot.n_ctx); SLT_INF(slot, "new slot, n_ctx = %d\n", slot.n_ctx);