vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 17:14:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge branch 'upstream' into concedo_experimental

Browse source 
# Conflicts:
#	examples/tts/tts.cpp
This commit is contained in:
Concedo 2025-03-16 15:42:39 +08:00
commit 5d7c5e9e33
6 changed files with 79 additions and 41 deletions
Download patch file Download diff file Expand all files Collapse all files

2
common/arg.cpp
View file

@ -1890,7 +1890,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
[](common_params & params, const std::string & value) { [](common_params & params, const std::string & value) {
params.out_file = value; params.out_file = value;
} }
).set_examples({LLAMA_EXAMPLE_IMATRIX, LLAMA_EXAMPLE_CVECTOR_GENERATOR, LLAMA_EXAMPLE_EXPORT_LORA})); ).set_examples({LLAMA_EXAMPLE_IMATRIX, LLAMA_EXAMPLE_CVECTOR_GENERATOR, LLAMA_EXAMPLE_EXPORT_LORA, LLAMA_EXAMPLE_TTS}));
add_opt(common_arg( add_opt(common_arg(
{"-ofreq", "--output-frequency"}, "N", {"-ofreq", "--output-frequency"}, "N",
string_format("output the imatrix every N iterations (default: %d)", params.n_out_freq), string_format("output the imatrix every N iterations (default: %d)", params.n_out_freq),

23
examples/tts/tts.cpp
View file

@ -87,11 +87,11 @@ struct wav_header {
uint32_t data_size; uint32_t data_size;
}; };
static void save_wav16(const std::string & fname, const std::vector<float> & data, int sample_rate) { static bool save_wav16(const std::string & fname, const std::vector<float> & data, int sample_rate) {
std::ofstream file(fname, std::ios::binary); std::ofstream file(fname, std::ios::binary);
if (!file) { if (!file) {
LOG_ERR("%s: Failed to open file '%s' for writing", __func__, fname.c_str()); LOG_ERR("%s: Failed to open file '%s' for writing.\n", __func__, fname.c_str());
return; return false;
} }
wav_header header; wav_header header;
@ -108,7 +108,7 @@ static void save_wav16(const std::string & fname, const std::vector<float> & dat
file.write(reinterpret_cast<const char*>(&pcm_sample), sizeof(pcm_sample)); file.write(reinterpret_cast<const char*>(&pcm_sample), sizeof(pcm_sample));
} }
file.close(); return file.good();
} }
static void fill_hann_window(int length, bool periodic, float * output) { static void fill_hann_window(int length, bool periodic, float * output) {
@ -536,6 +536,7 @@ static std::string audio_data_from_speaker(json speaker, const outetts_version t
int main(int argc, char ** argv) { int main(int argc, char ** argv) {
common_params params; common_params params;
params.out_file = "output.wav";
params.prompt = ""; params.prompt = "";
params.n_predict = 4096; params.n_predict = 4096;
@ -1060,8 +1061,6 @@ lovely<|t_0.56|><|code_start|><|634|><|596|><|1766|><|1556|><|1306|><|1285|><|14
} }
#endif #endif
const std::string fname = "output.wav";
const int n_sr = 24000; // sampling rate const int n_sr = 24000; // sampling rate
// zero out first 0.25 seconds // zero out first 0.25 seconds
@ -1072,11 +1071,15 @@ lovely<|t_0.56|><|code_start|><|634|><|596|><|1766|><|1556|><|1306|><|1285|><|14
LOG_INF("%s: time for spectral ops: %.3f ms\n", __func__, (ggml_time_us() - t_spec_start) / 1000.0f); LOG_INF("%s: time for spectral ops: %.3f ms\n", __func__, (ggml_time_us() - t_spec_start) / 1000.0f);
LOG_INF("%s: total time: %.3f ms\n", __func__, (ggml_time_us() - t_main_start) / 1000.0f); LOG_INF("%s: total time: %.3f ms\n", __func__, (ggml_time_us() - t_main_start) / 1000.0f);
save_wav16(fname, audio, n_sr); int retval = 0;
LOG_INF("%s: audio written to file '%s'\n", __func__, fname.c_str()); if (save_wav16(params.out_file, audio, n_sr)) {
LOG_INF("%s: audio written to file '%s'\n", __func__, params.out_file.c_str());
} else {
retval = ENOENT;
}
llama_backend_free(); llama_backend_free();
return 0; return retval;
} }

87
ggml/src/ggml-sycl/common.hpp
View file

@ -474,6 +474,7 @@ static void k_bin_bcast(const src0_t * src0, const src1_t * src1, dst_t * dst,
int ne0, int ne1, int ne2, int ne3, int ne0, int ne1, int ne2, int ne3,
int ne10, int ne11, int ne12, int ne13, int ne10, int ne11, int ne12, int ne13,
/*int s0, */ int s1, int s2, int s3, /*int s0, */ int s1, int s2, int s3,
/*int s00,*/ int s01, int s02, int s03,
/*int s10,*/ int s11, int s12, int s13, /*int s10,*/ int s11, int s12, int s13,
const sycl::nd_item<3> &item_ct1) { const sycl::nd_item<3> &item_ct1) {
const int i0s = item_ct1.get_local_range(2) * item_ct1.get_group(2) + const int i0s = item_ct1.get_local_range(2) * item_ct1.get_group(2) +
@ -495,9 +496,9 @@ static void k_bin_bcast(const src0_t * src0, const src1_t * src1, dst_t * dst,
const int i12 = i2 % ne12; const int i12 = i2 % ne12;
const int i13 = i3 % ne13; const int i13 = i3 % ne13;
const size_t i_src0 = i3*s3 + i2*s2 + i1*s1; const size_t i_src0 = i3*s03 + i2*s02 + i1*s01;
const size_t i_src1 = i13*s13 + i12*s12 + i11*s11; const size_t i_src1 = i13*s13 + i12*s12 + i11*s11;
const size_t i_dst = i_src0; const size_t i_dst = i3*s3 + i2*s2 + i1*s1;
const src0_t * src0_row = src0 + i_src0; const src0_t * src0_row = src0 + i_src0;
const src1_t * src1_row = src1 + i_src1; const src1_t * src1_row = src1 + i_src1;
@ -515,6 +516,7 @@ static void k_bin_bcast_unravel(const src0_t * src0, const src1_t * src1, dst_t
int ne0, int ne1, int ne2, int ne3, int ne0, int ne1, int ne2, int ne3,
int ne10, int ne11, int ne12, int ne13, int ne10, int ne11, int ne12, int ne13,
/*int s0, */ int s1, int s2, int s3, /*int s0, */ int s1, int s2, int s3,
/*int s00,*/ int s01, int s02, int s03,
/*int s10,*/ int s11, int s12, int s13, /*int s10,*/ int s11, int s12, int s13,
const sycl::nd_item<3> &item_ct1) { const sycl::nd_item<3> &item_ct1) {
@ -534,9 +536,9 @@ static void k_bin_bcast_unravel(const src0_t * src0, const src1_t * src1, dst_t
const int i12 = i2 % ne12; const int i12 = i2 % ne12;
const int i13 = i3 % ne13; const int i13 = i3 % ne13;
const size_t i_src0 = i3*s3 + i2*s2 + i1*s1; const size_t i_src0 = i3*s03 + i2*s02 + i1*s01;
const size_t i_src1 = i13*s13 + i12*s12 + i11*s11; const size_t i_src1 = i13*s13 + i12*s12 + i11*s11;
const size_t i_dst = i_src0; const size_t i_dst = i3*s3 + i2*s2 + i1*s1;
const src0_t * src0_row = src0 + i_src0; const src0_t * src0_row = src0 + i_src0;
const src1_t * src1_row = src1 + i_src1; const src1_t * src1_row = src1 + i_src1;
@ -566,9 +568,11 @@ struct bin_bcast_sycl {
int nr[4] = { nr0, nr1, nr2, nr3 }; int nr[4] = { nr0, nr1, nr2, nr3 };
// collapse dimensions until first broadcast dimension // collapse dimensions until first broadcast dimension
int64_t cne0[] = {ne0, ne1, ne2, ne3}; int64_t cne[] = {ne0, ne1, ne2, ne3};
int64_t cne0[] = {ne00, ne01, ne02, ne03};
int64_t cne1[] = {ne10, ne11, ne12, ne13}; int64_t cne1[] = {ne10, ne11, ne12, ne13};
size_t cnb0[] = {nb0, nb1, nb2, nb3}; size_t cnb[] = {nb0, nb1, nb2, nb3};
size_t cnb0[] = {nb00, nb01, nb02, nb03};
size_t cnb1[] = {nb10, nb11, nb12, nb13}; size_t cnb1[] = {nb10, nb11, nb12, nb13};
auto collapse = [](int64_t cne[]) { auto collapse = [](int64_t cne[]) {
cne[0] *= cne[1]; cne[0] *= cne[1];
@ -583,32 +587,41 @@ struct bin_bcast_sycl {
cnb[3] *= cne[3]; cnb[3] *= cne[3];
}; };
for (int i = 0; i < 4; i++) { if (ggml_is_contiguous(src0) && ggml_is_contiguous(src1) && ggml_is_contiguous(dst)) {
if (nr[i] != 1) { for (int i = 0; i < 4; i++) {
break; if (nr[i] != 1) {
} break;
if (i > 0) { }
collapse_nb(cnb0, cne0); if (i > 0) {
collapse_nb(cnb1, cne1); collapse_nb(cnb, cne);
collapse(cne0); collapse_nb(cnb0, cne0);
collapse(cne1); collapse_nb(cnb1, cne1);
collapse(cne);
collapse(cne0);
collapse(cne1);
}
} }
} }
{ {
int64_t ne0 = cne0[0]; int64_t ne0 = cne[0];
int64_t ne1 = cne0[1]; int64_t ne1 = cne[1];
int64_t ne2 = cne0[2]; int64_t ne2 = cne[2];
int64_t ne3 = cne0[3]; int64_t ne3 = cne[3];
int64_t ne10 = cne1[0]; int64_t ne10 = cne1[0];
int64_t ne11 = cne1[1]; int64_t ne11 = cne1[1];
int64_t ne12 = cne1[2]; int64_t ne12 = cne1[2];
int64_t ne13 = cne1[3]; int64_t ne13 = cne1[3];
size_t nb0 = cnb0[0]; size_t nb0 = cnb[0];
size_t nb1 = cnb0[1]; size_t nb1 = cnb[1];
size_t nb2 = cnb0[2]; size_t nb2 = cnb[2];
size_t nb3 = cnb0[3]; size_t nb3 = cnb[3];
size_t nb00 = cnb0[0];
size_t nb01 = cnb0[1];
size_t nb02 = cnb0[2];
size_t nb03 = cnb0[3];
size_t nb10 = cnb1[0]; size_t nb10 = cnb1[0];
size_t nb11 = cnb1[1]; size_t nb11 = cnb1[1];
@ -625,6 +638,28 @@ struct bin_bcast_sycl {
size_t s12 = nb12 / sizeof(src1_t); size_t s12 = nb12 / sizeof(src1_t);
size_t s13 = nb13 / sizeof(src1_t); size_t s13 = nb13 / sizeof(src1_t);
size_t s00 = nb00 / sizeof(src0_t);
size_t s01 = nb01 / sizeof(src0_t);
size_t s02 = nb02 / sizeof(src0_t);
size_t s03 = nb03 / sizeof(src0_t);
GGML_UNUSED(s00);
GGML_ASSERT(nb0 % sizeof(dst_t) == 0);
GGML_ASSERT(nb1 % sizeof(dst_t) == 0);
GGML_ASSERT(nb2 % sizeof(dst_t) == 0);
GGML_ASSERT(nb3 % sizeof(dst_t) == 0);
GGML_ASSERT(nb00 % sizeof(src0_t) == 0);
GGML_ASSERT(nb01 % sizeof(src0_t) == 0);
GGML_ASSERT(nb02 % sizeof(src0_t) == 0);
GGML_ASSERT(nb03 % sizeof(src0_t) == 0);
GGML_ASSERT(nb10 % sizeof(src1_t) == 0);
GGML_ASSERT(nb11 % sizeof(src1_t) == 0);
GGML_ASSERT(nb12 % sizeof(src1_t) == 0);
GGML_ASSERT(nb13 % sizeof(src1_t) == 0);
GGML_ASSERT(s0 == 1); GGML_ASSERT(s0 == 1);
GGML_ASSERT(s10 == 1); GGML_ASSERT(s10 == 1);
@ -661,8 +696,8 @@ struct bin_bcast_sycl {
[=](sycl::nd_item<3> item_ct1) { [=](sycl::nd_item<3> item_ct1) {
k_bin_bcast_unravel<bin_op>( k_bin_bcast_unravel<bin_op>(
src0_dd, src1_dd, dst_dd, ne0, ne1, ne2, ne3, src0_dd, src1_dd, dst_dd, ne0, ne1, ne2, ne3,
ne10, ne11, ne12, ne13, s1, s2, s3, s11, s12, ne10, ne11, ne12, ne13, s1, s2, s3, s01, s02,
s13, item_ct1); s03, s11, s12, s13, item_ct1);
}); });
} }
} else { } else {
@ -680,7 +715,7 @@ struct bin_bcast_sycl {
[=](sycl::nd_item<3> item_ct1) { [=](sycl::nd_item<3> item_ct1) {
k_bin_bcast<bin_op>(src0_dd, src1_dd, dst_dd, ne0, ne1, k_bin_bcast<bin_op>(src0_dd, src1_dd, dst_dd, ne0, ne1,
ne2, ne3, ne10, ne11, ne12, ne13, ne2, ne3, ne10, ne11, ne12, ne13,
s1, s2, s3, s11, s12, s13, s1, s2, s3, s01, s02, s03, s11, s12, s13,
item_ct1); item_ct1);
}); });
} }

4
ggml/src/ggml-sycl/ggml-sycl.cpp
View file

@ -3113,8 +3113,8 @@ static void ggml_sycl_mul_mat_id(ggml_backend_sycl_context & ctx,
const int64_t i2 = i12; const int64_t i2 = i12;
src0_row.data = src0_original + i02*nb02; src0_row.data = src0_original + i02*nb02;
src1_row.data = src1_original + + i11*nb11 + i12*nb12; src1_row.data = src1_original + i11*nb11 + i12*nb12;
dst_row.data = dst_original + i1*nb1 + i2*nb2; dst_row.data = dst_original + i1*nb1 + i2*nb2;
ggml_sycl_mul_mat(ctx, &src0_row, &src1_row, &dst_row); ggml_sycl_mul_mat(ctx, &src0_row, &src1_row, &dst_row);
} }

2
klite.embd
View file

@ -9719,7 +9719,7 @@ Current version indicated by LITEVER below.
},()=>{ },()=>{
}); });
} }
else if(localflag && no_txt_model && !has_txt2img && !koboldcpp_has_vision && !koboldcpp_has_whisper && !koboldcpp_has_tts) else if(localflag && no_txt_model && !has_txt2img && !koboldcpp_has_vision && !koboldcpp_has_whisper && !koboldcpp_has_tts && !is_using_kcpp_with_admin())
{ {
msgboxYesNo("This KoboldCpp instance has no models loaded. You can still use the WebUI to edit or view existing stories.<br><br>Would you like to connect to an external API service?","No Models Loaded", msgboxYesNo("This KoboldCpp instance has no models loaded. You can still use the WebUI to edit or view existing stories.<br><br>Would you like to connect to an external API service?","No Models Loaded",
()=>{ ()=>{

2
koboldcpp.py
View file

@ -49,7 +49,7 @@ logit_bias_max = 512
dry_seq_break_max = 128 dry_seq_break_max = 128
# global vars # global vars
KcppVersion = "1.86.1" KcppVersion = "1.86.2"
showdebug = True showdebug = True
kcpp_instance = None #global running instance kcpp_instance = None #global running instance
global_memory = {"tunnel_url": "", "restart_target":"", "input_to_exit":False, "load_complete":False} global_memory = {"tunnel_url": "", "restart_target":"", "input_to_exit":False, "load_complete":False}