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

Server: fix seed for multiple slots (#6835)

Browse source 
* Server: add tests for consistent results

* sampling: separate rng per sampling context
This commit is contained in:
Johannes Gäßler 2024-04-24 11:08:36 +02:00 • committed by GitHub
parent c0d1b3e03e
commit 28103f4832
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
11 changed files with 145 additions and 30 deletions
Download patch file Download diff file Expand all files Collapse all files

9
llama.h
View file

@ -987,7 +987,7 @@ extern "C" {
struct llama_context * ctx,
llama_token_data_array * candidates);
/// @details Randomly selects a token from the candidates based on their probabilities.
/// @details Randomly selects a token from the candidates based on their probabilities using the RNG of ctx.
LLAMA_API llama_token llama_sample_token(
struct llama_context * ctx,
llama_token_data_array * candidates);
@ -1074,8 +1074,9 @@ extern "C" {
// Internal API to be implemented by llama.cpp and used by tests/benchmarks only
#ifdef LLAMA_API_INTERNAL
#include <vector>
#include <random>
#include <string>
#include <vector>
struct ggml_tensor;
@ -1112,6 +1113,10 @@ std::pair<std::vector<uint32_t>, llama_partial_utf8> decode_utf8(
const std::string & src,
llama_partial_utf8 partial_start);
// Randomly selects a token from the candidates based on their probabilities using given std::mt19937.
// This is a temporary workaround in order to fix race conditions when sampling with multiple sequences.
llama_token llama_sample_token_with_rng(struct llama_context * ctx, llama_token_data_array * candidates, std::mt19937 & rng);
#endif // LLAMA_API_INTERNAL
#endif // LLAMA_H