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

llama : refactor sampling v2 (#9294)

Browse source 
- Add `struct llama_sampler` and `struct llama_sampler_i`
- Add `llama_sampler_` API
- Add `llama_sampler_chain_` API for chaining multiple samplers
- Remove `LLAMA_API_INTERNAL`
- Add `llama_perf_` API and remove old `llama_print_timings` and `llama_reset_timings`
This commit is contained in:
Georgi Gerganov 2024-09-07 15:16:19 +03:00 • committed by GitHub
parent 947538acb8
commit df270ef745
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
48 changed files with 3497 additions and 2914 deletions
Download patch file Download diff file Expand all files Collapse all files

13
examples/perplexity/perplexity.cpp
View file

@ -76,7 +76,7 @@ static void write_logfile(
fprintf(logfile, "ppl_value: %f\n", results.ppl_value);
yaml_dump_vector_float(logfile, "probs", results.probs);
llama_dump_timing_info_yaml(logfile, ctx);
llama_perf_dump_yaml(logfile, ctx);
fclose(logfile);
}
@ -2007,13 +2007,7 @@ int main(int argc, char ** argv) {
print_build_info();
if (params.seed == LLAMA_DEFAULT_SEED) {
params.seed = time(NULL);
}
fprintf(stderr, "%s: seed = %u\n", __func__, params.seed);
std::mt19937 rng(params.seed);
LOG_TEE("%s: seed = %u\n", __func__, params.sparams.seed);
llama_backend_init();
llama_numa_init(params.numa);
@ -2054,7 +2048,8 @@ int main(int argc, char ** argv) {
results = perplexity(ctx, params, n_ctx);
}
llama_print_timings(ctx);
LOG_TEE("\n");
llama_perf_print(ctx, LLAMA_PERF_TYPE_CONTEXT);
write_logfile(ctx, params, model, results);
llama_free(ctx);