Files
ik_llama.cpp/common/speculative.h
g2mt b837773728 Port speculative decoding from upstream to llama-server (#645)
* server : integrate speculative decoding

* server: Fix field names

* server: fix include, whitespace

* fix compile errors in speculative.cpp

* add llama_sampling_sample_and_accept_n to sampling

* finish porting speculative decoding in server

* port functions from common/speculative, common/sampling

* remove arg

* fix function names

* init params_dft to none

* correct value for n_ctx

* prefix kv cache tensors with model name to avoid conflict

* fix call arguments

* fix spec decoding args

* correct slot.id

* use n_max

* port the rest of sampling funcs

* fix func arguments

* slot.id starts at 1?

* Revert "prefix kv cache tensors with model name to avoid conflict"

This reverts commit fbd5dfd866.

* disable draft logging

* disable logging in speculative.cpp

in mainline, these would be LOG_DEBUG, but since ik_llama doesnt support
it, logging is disabled entirely

* add more draft model parameters

* fix

* pass flash_attn

* add speculative params for parity

* set speculative params in launch_slot_with_task instead
2025-08-16 07:26:44 +03:00

30 lines
888 B
C++

#pragma once
#include "llama.h"
#include <vector>
struct llama_speculative;
struct llama_speculative_params {
int n_draft = 16; // max drafted tokens
int n_reuse = 256;
float p_min = 0.75f; // min probability required to accept a token in the draft
};
struct llama_speculative * llama_speculative_init(struct llama_context * ctx_dft);
void llama_speculative_free(struct llama_speculative * spec);
bool llama_speculative_are_compatible(
const struct llama_context * ctx_tgt,
const struct llama_context * ctx_dft);
// sample up to n_draft tokens and add them to the batch using the draft model
std::vector<llama_token> llama_speculative_gen_draft(
struct llama_speculative * spec,
struct llama_speculative_params params,
const std::vector<llama_token> & prompt,
llama_token id_last);