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

server: fix comment about max n_probs

Browse source
This commit is contained in:
Wang Haoran(Robin) 2023-06-22 08:57:35 -07:00
parent ba210e4bc7
commit ccf254bd44
2 changed files with 2 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
examples/common.h
View file

@ -31,7 +31,7 @@ struct gpt_params {
int32_t main_gpu = 0; // the GPU that is used for scratch and small tensors
float tensor_split[LLAMA_MAX_DEVICES] = {0}; // how split tensors should be distributed across GPUs
bool low_vram = 0; // if true, reduce VRAM usage at the cost of performance
int32_t n_probs = 0; // if greater than 1, output the probabilities of top n_probs tokens. Max 5
int32_t n_probs = 0; // if greater than 0, output the probabilities of top n_probs tokens.
// sampling parameters
std::unordered_map<llama_token, float> logit_bias; // logit bias for specific tokens

1
examples/server/server.cpp
View file

@ -354,6 +354,7 @@ struct llama_server_context {
result.tok = llama_sample_token(ctx, &candidates_p);
}
}
// Add maximum of 5 most probable tokens to the result
for (size_t i = 0; i < std::min(candidates_p.size, std::min((size_t) n_probs, size_t(5))); ++i) {
result.probs.push_back({candidates_p.data[i].id, candidates_p.data[i].p});
}