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

server: fix reported top tokens for temperature 0 (#7203)

Browse source
This commit is contained in:
Johannes Gäßler 2024-05-11 10:11:28 +02:00 • committed by GitHub
parent b83cc3f5b3
commit 5ae3426b0b
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 7 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

6
common/sampling.cpp
View file

@ -35,7 +35,7 @@ struct llama_sampling_context * llama_sampling_init(const struct llama_sampling_
result->prev.resize(params.n_prev);
result->n_considered = 0;
result->n_valid = 0;
llama_sampling_set_rng_seed(result, params.seed);
@ -66,7 +66,7 @@ void llama_sampling_reset(llama_sampling_context * ctx) {
std::fill(ctx->prev.begin(), ctx->prev.end(), 0);
ctx->cur.clear();
ctx->n_considered = 0;
ctx->n_valid = 0;
}
void llama_sampling_set_rng_seed(struct llama_sampling_context * ctx, uint32_t seed) {
@ -256,7 +256,7 @@ static llama_token llama_sampling_sample_impl(
}
}
ctx_sampling->n_considered = cur_p.size;
ctx_sampling->n_valid = temp == 0.0f ? 0 : cur_p.size;
return id;
}

2
common/sampling.h
View file

@ -81,7 +81,7 @@ struct llama_sampling_context {
// TODO: replace with ring-buffer
std::vector<llama_token> prev;
std::vector<llama_token_data> cur;
size_t n_considered;
size_t n_valid; // Number of correct top tokens with correct probabilities.
std::mt19937 rng;
};