Skip to content

Samplers order parameters #4285

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 13 commits into from
Dec 5, 2023
Merged

Samplers order parameters #4285

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
13 commits
Select commit Hold shift + click to select a range
d4dc3d2
Samplers sequence order w parameter
MaggotHATE Dec 1, 2023
7601a49
Cleaned commented code
MaggotHATE Dec 1, 2023
d363df3
Fixed formatting
MaggotHATE Dec 1, 2023
bd08c8f
Rewrote with unordered_map
MaggotHATE Dec 2, 2023
8d2b460
Revert and rewrite, too many problems and safeguards would be needed
MaggotHATE Dec 2, 2023
e115b5a
Merge branch 'ggerganov:master' into master
MaggotHATE Dec 2, 2023
ff8adc1
Fixed code style
MaggotHATE Dec 2, 2023
e6dc166
Code style fixes according to review
MaggotHATE Dec 2, 2023
3fa6726
More readable samplers input string, fixed help
MaggotHATE Dec 3, 2023
ddda6dd
Style fix in sampler_queue
MaggotHATE Dec 3, 2023
c879b6d
Merge branch 'ggerganov:master' into master
MaggotHATE Dec 5, 2023
a6c3278
Formatting fixes
MaggotHATE Dec 5, 2023
0b87ef4
Fixing whitespaces
MaggotHATE Dec 5, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
156 changes: 72 additions & 84 deletions common/sampling.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,5 @@
#include "sampling.h"

#include <functional>

struct llama_sampling_context * llama_sampling_init(const struct llama_sampling_params & params) {
struct llama_sampling_context * result = new llama_sampling_context();

Expand Down Expand Up @@ -103,96 +101,89 @@ std::string llama_sampling_print(const llama_sampling_params & params) {

std::string llama_sampling_order_print(const llama_sampling_params & params) {
std::string result = "CFG -> Penalties ";

std::unordered_map<char, std::string> samplers_map_display {
{'k', "-> top_k "},
{'f', "-> tfs_z "},
{'y', "-> typical_p "},
{'p', "-> top_p "},
{'m', "-> min_p "},
{'t', "-> temp "}
};

if (params.mirostat == 0){
for (auto s : params.samplers_sequence){
result += samplers_map_display[s];
switch (s){
case 'k':{
result += "-> top_k ";
break;
}
case 'f':{
result += "-> tfs_z ";
break;
}
case 'y':{
result += "-> typical_p ";
break;
}
case 'p':{
result += "-> top_p ";
break;
}
case 'm':{
result += "-> min_p ";
break;
}
case 't':{
result += "-> temp ";
break;
}
default: break;
}
}
} else result += "-> mirostat ";

return result;
}

void sample_top_k(
const llama_sampling_params & params,
struct llama_context * ctx_main,
llama_token_data_array & cur_p,
size_t & min_keep){

const int n_vocab = llama_n_vocab(llama_get_model(ctx_main));
const int32_t top_k = params.top_k <= 0 ? n_vocab : params.top_k;
llama_sample_top_k (ctx_main, &cur_p, top_k, min_keep);
}

void sample_top_p(
const llama_sampling_params & params,
struct llama_context * ctx_main,
llama_token_data_array & cur_p,
size_t & min_keep){

const float top_p = params.top_p;
llama_sample_top_p (ctx_main, &cur_p, top_p, min_keep);
}

void sample_tfs_z(
const llama_sampling_params & params,
struct llama_context * ctx_main,
llama_token_data_array & cur_p,
size_t & min_keep){

const float tfs_z = params.tfs_z;
llama_sample_tail_free (ctx_main, &cur_p, tfs_z, min_keep);
}
// no reasons to expose this function in header
void sampler_queue(
struct llama_context * ctx_main,
const llama_sampling_params & params,
llama_token_data_array & cur_p,
size_t & min_keep) {
const int n_vocab = llama_n_vocab(llama_get_model(ctx_main));

void sample_typical_p(
const llama_sampling_params & params,
struct llama_context * ctx_main,
llama_token_data_array & cur_p,
size_t & min_keep){
const float temp = params.temp;
const int32_t top_k = params.top_k <= 0 ? n_vocab : params.top_k;
const float top_p = params.top_p;
const float min_p = params.min_p;
const float tfs_z = params.tfs_z;
const float typical_p = params.typical_p;
const std::string samplers_sequence = params.samplers_sequence;

for (auto s : samplers_sequence){
switch (s){
case 'k':{
llama_sample_top_k (ctx_main, &cur_p, top_k, min_keep);
break;
}
case 'f':{
llama_sample_tail_free(ctx_main, &cur_p, tfs_z, min_keep);
break;
}
case 'y':{
llama_sample_typical (ctx_main, &cur_p, typical_p, min_keep);
break;
}
case 'p':{
llama_sample_top_p (ctx_main, &cur_p, top_p, min_keep);
break;
}
case 'm':{
llama_sample_min_p (ctx_main, &cur_p, min_p, min_keep);
break;
}
case 't':{
llama_sample_temp (ctx_main, &cur_p, temp);
break;
}
default: break;
}
}

const float typical_p = params.typical_p;
llama_sample_typical (ctx_main, &cur_p, typical_p, min_keep);
}

void sample_min_p(
const llama_sampling_params & params,
struct llama_context * ctx_main,
llama_token_data_array & cur_p,
size_t & min_keep){

const float min_p = params.min_p;
llama_sample_min_p (ctx_main, &cur_p, min_p, min_keep);
}

void sample_temp(
const llama_sampling_params & params,
struct llama_context * ctx_main,
llama_token_data_array & cur_p,
size_t & min_keep){

const float temp = params.temp;
llama_sample_temp (ctx_main, &cur_p, temp);
}

std::unordered_map<char, std::function<void(const llama_sampling_params &, struct llama_context *, llama_token_data_array&, size_t&)>> samplers_map
{
{'k', sample_top_k},
{'f', sample_tfs_z},
{'y', sample_typical_p},
{'p', sample_top_p},
{'m', sample_min_p},
{'t', sample_temp}
};

llama_token llama_sampling_sample(
struct llama_sampling_context * ctx_sampling,
struct llama_context * ctx_main,
Expand All @@ -211,7 +202,6 @@ llama_token llama_sampling_sample(
const float mirostat_tau = params.mirostat_tau;
const float mirostat_eta = params.mirostat_eta;
const bool penalize_nl = params.penalize_nl;
const std::string samplers_sequence = params.samplers_sequence;

auto & prev = ctx_sampling->prev;
auto & cur = ctx_sampling->cur;
Expand Down Expand Up @@ -278,9 +268,7 @@ llama_token llama_sampling_sample(
// temperature sampling
size_t min_keep = std::max(1, params.n_probs);

for (auto s : samplers_sequence){
samplers_map[s](params, ctx_main, cur_p, min_keep);
}
sampler_queue(ctx_main, params, cur_p, min_keep);

id = llama_sample_token(ctx_main, &cur_p);

Expand Down
36 changes: 0 additions & 36 deletions common/sampling.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -84,42 +84,6 @@ std::string llama_sampling_print(const llama_sampling_params & params);
// Print sampling order into a string
std::string llama_sampling_order_print(const llama_sampling_params & params);

void sample_top_k(
const llama_sampling_params & params,
struct llama_context * ctx_main,
llama_token_data_array & cur_p,
size_t & min_keep);

void sample_top_p(
const llama_sampling_params & params,
struct llama_context * ctx_main,
llama_token_data_array & cur_p,
size_t & min_keep);

void sample_tfs_z(
const llama_sampling_params & params,
struct llama_context * ctx_main,
llama_token_data_array & cur_p,
size_t & min_keep);

void sample_typical_p(
const llama_sampling_params & params,
struct llama_context * ctx_main,
llama_token_data_array & cur_p,
size_t & min_keep);

void sample_min_p(
const llama_sampling_params & params,
struct llama_context * ctx_main,
llama_token_data_array & cur_p,
size_t & min_keep);

void sample_temp(
const llama_sampling_params & params,
struct llama_context * ctx_main,
llama_token_data_array & cur_p,
size_t & min_keep);

// this is a common sampling function used across the examples for convenience
// it can serve as a starting point for implementing your own sampling function
// Note: When using multiple sequences, it is the caller's responsibility to call
Expand Down