thek0tyara/llama-cpp-turboquant
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 4

common : refactor common_sampler + grammar logic changes (#17937)

Browse source 
* common : refactor common_sampler + grammar logic changes

* tests : increase max_tokens to get needed response

* batched : fix uninitialized samplers
This commit is contained in:
Georgi Gerganov 2025-12-14 10:11:13 +02:00 • committed by GitHub
parent 3238b1400c
commit 254098a279
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
27 changed files with 372 additions and 293 deletions
Download patch file Download diff file Expand all files Collapse all files

2
common/arg.cpp
View file

@ -1415,7 +1415,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
params.sampling.top_k = value;
params.sampling.user_sampling_config |= common_params_sampling_config::COMMON_PARAMS_SAMPLING_CONFIG_TOP_K;
}
).set_sparam());
).set_sparam().set_env("LLAMA_ARG_TOP_K"));
add_opt(common_arg(
{"--top-p"}, "N",
string_format("top-p sampling (default: %.1f, 1.0 = disabled)", (double)params.sampling.top_p),