tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-13 11:57:43 -04:00
Code Issues Packages Projects Releases Wiki Activity

sampling : avoid expensive softmax during greedy sampling (#9605)

Browse Source 
* sampling : avoid expensive softmax during greedy sampling

ggml-ci

* speculative : fix default RNG seed + set sparams.n_probs

* Update tests/test-sampling.cpp

Co-authored-by: slaren <slarengh@gmail.com>

* sampling : add clarifying comment [no ci]

---------

Co-authored-by: slaren <slarengh@gmail.com>
This commit is contained in:
Georgi Gerganov
2024-09-24 09:03:17 +03:00
committed by GitHub
parent c087b6f11d
commit b0f27361f3
5 changed files with 59 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
common/sampling.cpp
View File

@@ -209,7 +209,15 @@ struct gpt_sampler * gpt_sampler_init(const struct llama_model * model, const st
GGML_ASSERT(false && "unknown mirostat version");
}
} else {
llama_sampler_chain_add(result->chain, llama_sampler_init_softmax());
if (params.n_probs > 0) {
// some use cases require to sample greedily, but still obtain the probabilities of the top tokens
// ref: https://github.com/ggerganov/llama.cpp/pull/9605
//
// the following will not produce exactly the same probs as applyging softmax to the full vocabulary, but
// it is much faster, since we avoid sorting all tokens and should give a good approximation
llama_sampler_chain_add(result->chain, llama_sampler_init_top_k(params.n_probs));
llama_sampler_chain_add(result->chain, llama_sampler_init_softmax());
}
llama_sampler_chain_add(result->chain, llama_sampler_init_greedy());
}