vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

speculative : refactor and add a simpler example (#10362)

Browse source 
* speculative : refactor and add a simpler example

ggml-ci

* speculative : clean-up and add comments and TODOs [no ci]

* speculative : manage context in common_speculative

ggml-ci

* speculative : simplify

ggml-ci

* speculative : simplify (cont)

ggml-ci

* speculative : add --draft-min CLI arg

* speculative : minor fixup

* make : build fixes

* speculative : do not redraft previous drafts

ggml-ci

* speculative : fix the draft sampling

ggml-ci

* speculative : fix compile warning

* common : refactor args

ggml-ci

* common : change defaults [no ci]

* common : final touches

ggml-ci
This commit is contained in:
Georgi Gerganov 2024-11-25 09:58:41 +02:00 • committed by GitHub
parent cce5a90075
commit d9d54e498d
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
28 changed files with 1028 additions and 326 deletions
Download patch file Download diff file Expand all files Collapse all files

23
common/sampling.h
View file

@ -36,7 +36,7 @@ struct common_sampler;
// llama_sampler API overloads
struct common_sampler * common_sampler_init(const struct llama_model * model, const struct common_sampler_params & params);
struct common_sampler * common_sampler_init(const struct llama_model * model, const struct common_params_sampling & params);
void common_sampler_free(struct common_sampler * gsmpl);
@ -60,6 +60,27 @@ void common_perf_print(const struct llama_context * ctx, const struct common_sam
//
llama_token common_sampler_sample(struct common_sampler * gsmpl, struct llama_context * ctx, int idx, bool grammar_first = false);
// generalized version of common_sampler_sample
//
// will cross-reference the sampled tokens with a batch of draft tokens and accept those that match
// if the sampler disagrees at some point, we stop and return the accepted tokens up to now
//
// common_sampler_sample_n(gsmpl, ctx, { idx }, {});
//
// is equivalent to
//
// common_sampler_sample(gsmpl, ctx, idx);
// common_sampler_accept(gsmpl, token, true);
//
// requires: idxs.size() == draft.size() + 1
//
// returns at least 1 token, up to idxs.size()
//
std::vector<llama_token> common_sampler_sample_and_accept_n(struct common_sampler * gsmpl, struct llama_context * ctx, const std::vector<int> & idxs, const llama_tokens & draft, bool grammar_first = false);
// assume idxs == [ 0, 1, 2, ..., draft.size() ]
std::vector<llama_token> common_sampler_sample_and_accept_n(struct common_sampler * gsmpl, struct llama_context * ctx, const llama_tokens & draft, bool grammar_first = false);
uint32_t common_sampler_get_seed(const struct common_sampler * gsmpl);
// helpers