2023-05-01 16:23:47 +00:00
# include "build-info.h"
2023-03-10 18:40:58 +00:00
2023-05-20 08:06:11 +00:00
# include "llama.h"
2023-03-10 18:40:58 +00:00
# include <cstdio>
2023-06-10 07:59:17 +00:00
# include <cstring>
2023-06-13 10:23:23 +00:00
# include <vector>
2023-03-10 18:40:58 +00:00
# include <string>
2023-03-21 17:21:50 +00:00
2023-06-13 10:23:23 +00:00
struct quant_option {
std : : string name ;
llama_ftype ftype ;
std : : string desc ;
2023-04-26 16:43:27 +00:00
} ;
2023-06-13 10:23:23 +00:00
static const std : : vector < struct quant_option > QUANT_OPTIONS = {
{
" Q4_0 " ,
LLAMA_FTYPE_MOSTLY_Q4_0 ,
" 3.50G, +0.2499 ppl @ 7B - small, very high quality loss - legacy, prefer using Q3_K_M " ,
} ,
{
" Q4_1 " ,
LLAMA_FTYPE_MOSTLY_Q4_1 ,
" 3.90G, +0.1846 ppl @ 7B - small, substantial quality loss - legacy, prefer using Q3_K_L " ,
} ,
{
" Q5_0 " ,
LLAMA_FTYPE_MOSTLY_Q5_0 ,
" 4.30G, +0.0796 ppl @ 7B - medium, balanced quality - legacy, prefer using Q4_K_M " ,
} ,
{
" Q5_1 " ,
LLAMA_FTYPE_MOSTLY_Q5_1 ,
" 4.70G, +0.0415 ppl @ 7B - medium, low quality loss - legacy, prefer using Q5_K_M " ,
} ,
# ifdef GGML_USE_K_QUANTS
{
" Q2_K " ,
LLAMA_FTYPE_MOSTLY_Q2_K ,
" 2.67G, +0.8698 ppl @ 7B - smallest, extreme quality loss - not recommended " ,
} ,
{
" Q3_K " ,
LLAMA_FTYPE_MOSTLY_Q3_K_M ,
" alias for Q3_K_M "
} ,
{
" Q3_K_S " ,
LLAMA_FTYPE_MOSTLY_Q3_K_S ,
" 2.75G, +0.5505 ppl @ 7B - very small, very high quality loss " ,
} ,
{
" Q3_K_M " ,
LLAMA_FTYPE_MOSTLY_Q3_K_M ,
" 3.06G, +0.2437 ppl @ 7B - very small, very high quality loss " ,
} ,
{
" Q3_K_L " ,
LLAMA_FTYPE_MOSTLY_Q3_K_L ,
" 3.35G, +0.1803 ppl @ 7B - small, substantial quality loss " ,
} ,
{
" Q4_K " ,
LLAMA_FTYPE_MOSTLY_Q4_K_M ,
" alias for Q4_K_M " ,
} ,
{
" Q4_K_S " ,
LLAMA_FTYPE_MOSTLY_Q4_K_S ,
" 3.56G, +0.1149 ppl @ 7B - small, significant quality loss " ,
} ,
{
" Q4_K_M " ,
LLAMA_FTYPE_MOSTLY_Q4_K_M ,
" 3.80G, +0.0535 ppl @ 7B - medium, balanced quality - *recommended* " ,
} ,
{
" Q5_K " ,
LLAMA_FTYPE_MOSTLY_Q5_K_M ,
" alias for Q5_K_M " ,
} ,
{
" Q5_K_S " ,
LLAMA_FTYPE_MOSTLY_Q5_K_S ,
" 4.33G, +0.0353 ppl @ 7B - large, low quality loss - *recommended* " ,
} ,
{
" Q5_K_M " ,
LLAMA_FTYPE_MOSTLY_Q5_K_M ,
" 4.45G, +0.0142 ppl @ 7B - large, very low quality loss - *recommended* " ,
} ,
{
" Q6_K " ,
LLAMA_FTYPE_MOSTLY_Q6_K ,
" 5.15G, +0.0044 ppl @ 7B - very large, extremely low quality loss " ,
} ,
# endif
{
" Q8_0 " ,
LLAMA_FTYPE_MOSTLY_Q8_0 ,
" 6.70G, +0.0004 ppl @ 7B - very large, extremely low quality loss - not recommended " ,
} ,
{
" F16 " ,
LLAMA_FTYPE_MOSTLY_F16 ,
" 13.00G @ 7B - extremely large, virtually no quality loss - not recommended " ,
} ,
{
" F32 " ,
LLAMA_FTYPE_ALL_F32 ,
" 26.00G @ 7B - absolutely huge, lossless - not recommended " ,
} ,
} ;
bool try_parse_ftype ( const std : : string & ftype_str_in , llama_ftype & ftype , std : : string & ftype_str_out ) {
std : : string ftype_str ;
for ( auto ch : ftype_str_in ) {
ftype_str . push_back ( std : : toupper ( ch ) ) ;
}
for ( auto & it : QUANT_OPTIONS ) {
if ( it . name = = ftype_str ) {
ftype = it . ftype ;
ftype_str_out = it . name ;
return true ;
}
2023-05-04 22:58:56 +00:00
}
try {
int ftype_int = std : : stoi ( ftype_str ) ;
2023-06-13 10:23:23 +00:00
for ( auto & it : QUANT_OPTIONS ) {
if ( it . ftype = = ftype_int ) {
ftype = it . ftype ;
ftype_str_out = it . name ;
2023-05-04 22:58:56 +00:00
return true ;
}
}
}
catch ( . . . ) {
// stoi failed
}
return false ;
}
2023-03-10 18:40:58 +00:00
// usage:
2023-06-13 10:23:23 +00:00
// ./quantize [--allow-requantize] [--leave-output-tensor] models/llama/ggml-model.bin [models/llama/ggml-model-quant.bin] type [nthreads]
2023-03-10 18:40:58 +00:00
//
2023-06-10 07:59:17 +00:00
void usage ( const char * executable ) {
2023-06-13 10:23:23 +00:00
fprintf ( stderr , " usage: %s [--help] [--allow-requantize] [--leave-output-tensor] model-f32.bin [model-quant.bin] type [nthreads] \n \n " , executable ) ;
2023-06-10 07:59:17 +00:00
fprintf ( stderr , " --allow-requantize: Allows requantizing tensors that have already been quantized. Warning: This can severely reduce quality compared to quantizing from 16bit or 32bit \n " ) ;
fprintf ( stderr , " --leave-output-tensor: Will leave output.weight un(re)quantized. Increases model size but may also increase quality, especially when requantizing \n " ) ;
2023-06-13 10:23:23 +00:00
fprintf ( stderr , " \n Allowed quantization types: \n " ) ;
for ( auto & it : QUANT_OPTIONS ) {
printf ( " %2d or %-6s : %s \n " , it . ftype , it . name . c_str ( ) , it . desc . c_str ( ) ) ;
2023-06-10 07:59:17 +00:00
}
exit ( 1 ) ;
}
2023-03-10 18:40:58 +00:00
int main ( int argc , char * * argv ) {
2023-05-04 22:58:56 +00:00
if ( argc < 3 ) {
2023-06-10 07:59:17 +00:00
usage ( argv [ 0 ] ) ;
}
llama_model_quantize_params params = llama_model_quantize_default_params ( ) ;
int arg_idx = 1 ;
for ( ; arg_idx < argc & & strncmp ( argv [ arg_idx ] , " -- " , 2 ) = = 0 ; arg_idx + + ) {
if ( strcmp ( argv [ arg_idx ] , " --leave-output-tensor " ) = = 0 ) {
params . quantize_output_tensor = false ;
} else if ( strcmp ( argv [ arg_idx ] , " --allow-requantize " ) = = 0 ) {
params . allow_requantize = true ;
} else {
usage ( argv [ 0 ] ) ;
2023-04-26 16:43:27 +00:00
}
2023-06-10 07:59:17 +00:00
}
if ( argc - arg_idx < 3 ) {
usage ( argv [ 0 ] ) ;
2023-03-10 18:40:58 +00:00
}
2023-06-26 17:57:59 +00:00
llama_init_backend ( false ) ;
2023-03-11 15:40:14 +00:00
2023-05-04 22:58:56 +00:00
// parse command line arguments
2023-06-10 07:59:17 +00:00
const std : : string fname_inp = argv [ arg_idx ] ;
arg_idx + + ;
2023-05-04 22:58:56 +00:00
std : : string fname_out ;
std : : string ftype_str ;
2023-06-10 07:59:17 +00:00
if ( try_parse_ftype ( argv [ arg_idx ] , params . ftype , ftype_str ) ) {
2023-05-04 22:58:56 +00:00
std : : string fpath ;
const size_t pos = fname_inp . find_last_of ( ' / ' ) ;
if ( pos ! = std : : string : : npos ) {
fpath = fname_inp . substr ( 0 , pos + 1 ) ;
}
// export as [inp path]/ggml-model-[ftype].bin
fname_out = fpath + " ggml-model- " + ftype_str + " .bin " ;
arg_idx + + ;
}
else {
fname_out = argv [ arg_idx ] ;
arg_idx + + ;
2023-03-10 18:40:58 +00:00
2023-05-04 22:58:56 +00:00
if ( argc < = arg_idx ) {
fprintf ( stderr , " %s: missing ftype \n " , __func__ ) ;
return 1 ;
}
2023-06-10 07:59:17 +00:00
if ( ! try_parse_ftype ( argv [ arg_idx ] , params . ftype , ftype_str ) ) {
2023-05-04 22:58:56 +00:00
fprintf ( stderr , " %s: invalid ftype '%s' \n " , __func__ , argv [ 3 ] ) ;
return 1 ;
}
arg_idx + + ;
}
// parse nthreads
if ( argc > arg_idx ) {
try {
2023-06-10 07:59:17 +00:00
params . nthread = std : : stoi ( argv [ arg_idx ] ) ;
2023-05-04 22:58:56 +00:00
}
catch ( const std : : exception & e ) {
fprintf ( stderr , " %s: invalid nthread '%s' (%s) \n " , __func__ , argv [ arg_idx ] , e . what ( ) ) ;
2023-04-26 16:43:27 +00:00
return 1 ;
}
}
2023-05-01 16:23:47 +00:00
fprintf ( stderr , " %s: build = %d (%s) \n " , __func__ , BUILD_NUMBER , BUILD_COMMIT ) ;
2023-05-04 22:58:56 +00:00
fprintf ( stderr , " %s: quantizing '%s' to '%s' as %s " , __func__ , fname_inp . c_str ( ) , fname_out . c_str ( ) , ftype_str . c_str ( ) ) ;
2023-06-10 07:59:17 +00:00
if ( params . nthread > 0 ) {
fprintf ( stderr , " using %d threads " , params . nthread ) ;
2023-05-04 22:58:56 +00:00
}
fprintf ( stderr , " \n " ) ;
2023-03-10 18:40:58 +00:00
2023-05-20 08:06:11 +00:00
const int64_t t_main_start_us = llama_time_us ( ) ;
2023-03-10 18:40:58 +00:00
int64_t t_quantize_us = 0 ;
// load the model
{
2023-05-20 08:06:11 +00:00
const int64_t t_start_us = llama_time_us ( ) ;
2023-03-10 18:40:58 +00:00
2023-06-10 07:59:17 +00:00
if ( llama_model_quantize ( fname_inp . c_str ( ) , fname_out . c_str ( ) , & params ) ) {
2023-03-10 18:40:58 +00:00
fprintf ( stderr , " %s: failed to quantize model from '%s' \n " , __func__ , fname_inp . c_str ( ) ) ;
return 1 ;
}
2023-05-20 08:06:11 +00:00
t_quantize_us = llama_time_us ( ) - t_start_us ;
2023-03-10 18:40:58 +00:00
}
// report timing
{
2023-05-20 08:06:11 +00:00
const int64_t t_main_end_us = llama_time_us ( ) ;
2023-03-10 18:40:58 +00:00
printf ( " \n " ) ;
2023-03-28 16:48:20 +00:00
printf ( " %s: quantize time = %8.2f ms \n " , __func__ , t_quantize_us / 1000.0 ) ;
printf ( " %s: total time = %8.2f ms \n " , __func__ , ( t_main_end_us - t_main_start_us ) / 1000.0 ) ;
2023-03-10 18:40:58 +00:00
}
return 0 ;
}