Skip to content

Commit 8561b00

Browse files
ggerganovggerganov
committed
examples : fix build + compile warnings (close ggml-org#1256)
1 parent a3578e8 commit 8561b00

File tree

7 files changed

+29
-24
lines changed

7 files changed

+29
-24
lines changed

‎examples/common.cpp

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -792,7 +792,7 @@ bool sam_params_parse(int argc, char ** argv, sam_params & params) {
792792
return true;
793793
}
794794

795-
void sam_print_usage(int argc, char ** argv, const sam_params & params) {
795+
void sam_print_usage(int /*argc*/, char ** argv, const sam_params & params) {
796796
fprintf(stderr, "usage: %s [options]\n", argv[0]);
797797
fprintf(stderr, "\n");
798798
fprintf(stderr, "options:\n");

‎examples/lsp/lsp.cpp

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -324,12 +324,12 @@ json register_commandset(struct whisper_context * ctx, json jparams, std::vector
324324
commandset_list.push_back(cs);
325325
return json{{"index",index}};
326326
}
327-
json seek(struct whisper_context * ctx, audio_async &audio, json params) {
327+
json seek(struct whisper_context * /*ctx*/, audio_async & /*audio*/, json /*params*/) {
328328
// whisper_state has the pertinent offsets, but there also seem to be a large
329329
// number of scratch buffers that would prevent rewinding context in a manner similar to llama
330330
// I'll give this a another pass once everything else is implemented,
331331
// but for now, it's unsupported
332-
throw json{
332+
throw json {
333333
{"code", -32601},
334334
{"message", "Seeking is not yet supported."}
335335
};
@@ -412,7 +412,7 @@ void process_loop(struct whisper_context * ctx, audio_async &audio, const whispe
412412
jobqueue.pop_front();
413413
// send response
414414
std::string data = resp.dump(-1, ' ', false, json::error_handler_t::replace);
415-
fprintf(stdout, "Content-Length: %d\r\n\r\n%s\n", data.length()+1, data.c_str());
415+
fprintf(stdout, "Content-Length: %d\r\n\r\n%s\n", (int)data.length()+1, data.c_str());
416416
std::cout.flush();
417417

418418
}

‎examples/main/main.cpp

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -260,7 +260,7 @@ std::string estimate_diarization_speaker(std::vector<std::vector<float>> pcmf32s
260260

261261
return speaker;
262262
}
263-
void whisper_print_progress_callback(struct whisper_context * ctx, struct whisper_state * /*state*/, int progress, void * user_data) {
263+
void whisper_print_progress_callback(struct whisper_context * /*ctx*/, struct whisper_state * /*state*/, int progress, void * user_data) {
264264
int progress_step = ((whisper_print_user_data *) user_data)->params->progress_step;
265265
int * progress_prev = &(((whisper_print_user_data *) user_data)->progress_prev);
266266
if (progress >= *progress_prev + progress_step) {
@@ -492,7 +492,7 @@ bool output_csv(struct whisper_context * ctx, const char * fname, const whisper_
492492
return true;
493493
}
494494

495-
bool output_score(struct whisper_context * ctx, const char * fname, const whisper_params & params, std::vector<std::vector<float>> pcmf32s) {
495+
bool output_score(struct whisper_context * ctx, const char * fname, const whisper_params & /*params*/, std::vector<std::vector<float>> /*pcmf32s*/) {
496496
std::ofstream fout(fname);
497497
fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname);
498498

‎examples/talk-llama/llama.cpp

Lines changed: 12 additions & 13 deletions
Original file line numberDiff line numberDiff line change
@@ -1164,7 +1164,7 @@ static bool llama_eval_internal(
11641164
const llama_token * tokens,
11651165
const int n_tokens,
11661166
const int n_past,
1167-
const int n_threads) {
1167+
int n_threads) {
11681168

11691169
// enforce that the first token is BOS
11701170
if (n_past == 0 && tokens[0] != llama_token_bos()) {
@@ -1190,6 +1190,8 @@ static bool llama_eval_internal(
11901190
const int n_vocab = hparams.n_vocab;
11911191
const int n_rot = hparams.n_embd/hparams.n_head;
11921192

1193+
const float eps = 5e-6f; // TODO: take from hparams
1194+
11931195
auto & mem_per_token = lctx.mem_per_token;
11941196
auto & buf_compute = lctx.buf_compute;
11951197

@@ -1204,7 +1206,7 @@ static bool llama_eval_internal(
12041206
// for big prompts, if BLAS is enabled, it is better to use only one thread
12051207
// otherwise, the threads are spin-lock waiting for the BLAS calls and are degrading the performance
12061208
ggml_cgraph gf = {};
1207-
gf.n_threads = N >= 32 && ggml_cpu_has_blas() && !ggml_cpu_has_gpublas() ? 1 : n_threads;
1209+
n_threads = N >= 32 && ggml_cpu_has_blas() && !ggml_cpu_has_gpublas() ? 1 : n_threads;
12081210

12091211
struct ggml_tensor * embd = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, N);
12101212
ggml_set_name(embd, "embd");
@@ -1221,7 +1223,7 @@ static bool llama_eval_internal(
12211223

12221224
// norm
12231225
{
1224-
cur = ggml_rms_norm(ctx0, inpL);
1226+
cur = ggml_rms_norm(ctx0, inpL, eps);
12251227

12261228
// cur = cur*attention_norm(broadcasted)
12271229
cur = ggml_mul(ctx0, cur, model.layers[il].attention_norm);
@@ -1329,7 +1331,7 @@ static bool llama_eval_internal(
13291331
{
13301332
// norm
13311333
{
1332-
cur = ggml_rms_norm(ctx0, inpFF);
1334+
cur = ggml_rms_norm(ctx0, inpFF, eps);
13331335

13341336
// cur = cur*ffn_norm(broadcasted)
13351337
cur = ggml_mul(ctx0, cur, model.layers[il].ffn_norm);
@@ -1367,7 +1369,7 @@ static bool llama_eval_internal(
13671369
// norm
13681370
{
13691371

1370-
inpL = ggml_rms_norm(ctx0, inpL);
1372+
inpL = ggml_rms_norm(ctx0, inpL, eps);
13711373

13721374
// inpL = inpL*norm(broadcasted)
13731375
inpL = ggml_mul(ctx0, inpL, model.norm);
@@ -1384,8 +1386,8 @@ static bool llama_eval_internal(
13841386
//inpL = ggml_soft_max_inplace(ctx0, inpL);
13851387

13861388
// run the computation
1387-
ggml_build_forward_expand(&gf, inpL);
1388-
ggml_graph_compute (ctx0, &gf);
1389+
ggml_build_forward_expand (&gf, inpL);
1390+
ggml_graph_compute_with_ctx(ctx0, &gf, n_threads);
13891391

13901392
#ifdef GGML_PERF
13911393
// print timing information per ggml operation (for debugging purposes)
@@ -2488,8 +2490,7 @@ int llama_apply_lora_from_file_internal(struct llama_context * ctx, const char *
24882490
}
24892491

24902492
struct ggml_cgraph gf = ggml_build_forward(r);
2491-
gf.n_threads = n_threads;
2492-
ggml_graph_compute(lora_ctx, &gf);
2493+
ggml_graph_compute_with_ctx(lora_ctx, &gf, n_threads);
24932494

24942495
// we won't need these tensors again, reset the context to save memory
24952496
ggml_free(lora_ctx);
@@ -2635,7 +2636,6 @@ size_t llama_copy_state_data(struct llama_context * ctx, uint8_t * dst) {
26352636

26362637
ggml_context * cpy_ctx = ggml_init({ sizeof(buffer), buffer, /* no_alloc */ true });
26372638
ggml_cgraph gf{};
2638-
gf.n_threads = 1;
26392639

26402640
ggml_tensor * kout3d = ggml_new_tensor_3d(cpy_ctx, kv_self.k->type, n_embd, kv_ntok, n_layer);
26412641
kout3d->data = out;
@@ -2655,7 +2655,7 @@ size_t llama_copy_state_data(struct llama_context * ctx, uint8_t * dst) {
26552655

26562656
ggml_build_forward_expand(&gf, ggml_cpy(cpy_ctx, k3d, kout3d));
26572657
ggml_build_forward_expand(&gf, ggml_cpy(cpy_ctx, v3d, vout3d));
2658-
ggml_graph_compute(cpy_ctx, &gf);
2658+
ggml_graph_compute_with_ctx(cpy_ctx, &gf, 1);
26592659

26602660
ggml_free(cpy_ctx);
26612661
}
@@ -2743,7 +2743,6 @@ size_t llama_set_state_data(struct llama_context * ctx, uint8_t * src) {
27432743

27442744
ggml_context * cpy_ctx = ggml_init({ sizeof(buffer), buffer, /* no_alloc */ true });
27452745
ggml_cgraph gf{};
2746-
gf.n_threads = 1;
27472746

27482747
ggml_tensor * kin3d = ggml_new_tensor_3d(cpy_ctx, kv_self.k->type, n_embd, kv_ntok, n_layer);
27492748
kin3d->data = (void *) inp;
@@ -2763,7 +2762,7 @@ size_t llama_set_state_data(struct llama_context * ctx, uint8_t * src) {
27632762

27642763
ggml_build_forward_expand(&gf, ggml_cpy(cpy_ctx, kin3d, k3d));
27652764
ggml_build_forward_expand(&gf, ggml_cpy(cpy_ctx, vin3d, v3d));
2766-
ggml_graph_compute(cpy_ctx, &gf);
2765+
ggml_graph_compute_with_ctx(cpy_ctx, &gf, 1);
27672766

27682767
ggml_free(cpy_ctx);
27692768
}

‎examples/talk-llama/talk-llama.cpp

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -649,7 +649,10 @@ int main(int argc, char ** argv) {
649649
}
650650

651651
text_to_speak = ::replace(text_to_speak, "\"", "");
652-
system((params.speak + " " + std::to_string(voice_id) + " \"" + text_to_speak + "\"").c_str());
652+
int ret = system((params.speak + " " + std::to_string(voice_id) + " \"" + text_to_speak + "\"").c_str());
653+
if (ret != 0) {
654+
fprintf(stderr, "%s: failed to speak\n", __func__);
655+
}
653656

654657
audio.clear();
655658

‎examples/talk/gpt-2.cpp

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -191,9 +191,9 @@ bool gpt2_model_load(const std::string & fname, gpt2_model & model, gpt_vocab &
191191
// create the ggml context
192192
{
193193
struct ggml_init_params params = {
194-
.mem_size = ctx_size,
195-
.mem_buffer = NULL,
196-
.no_alloc = false,
194+
/*.mem_size =*/ ctx_size,
195+
/*.mem_buffer =*/ NULL,
196+
/*.no_alloc =*/ false,
197197
};
198198

199199
model.ctx = ggml_init(params);

‎examples/talk/talk.cpp

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -349,7 +349,10 @@ int main(int argc, char ** argv) {
349349
gpt2_set_prompt(ctx_gpt, prompt_base.c_str());
350350

351351
text_to_speak = ::replace(text_to_speak, params.person + ": ", "");
352-
system((params.speak + " " + std::to_string(voice_id) + " \"" + text_to_speak + "\"").c_str());
352+
int ret = system((params.speak + " " + std::to_string(voice_id) + " \"" + text_to_speak + "\"").c_str());
353+
if (ret != 0) {
354+
fprintf(stderr, "%s: system() failed!\n", __func__);
355+
}
353356

354357
audio.clear();
355358

0 commit comments

Comments
 (0)