MNT: Add unit tests for coarse model

.gitignore

@@ -7,6 +7,7 @@ encodec
 main
 tests/test-tokenizer
 tests/test-text-encoder
+tests/test-coarse-encoder
 
 *.o
 *.plist

Makefile

@@ -2,7 +2,7 @@
 BUILD_TARGETS = main
 
 # Binaries only useful for tests
-TEST_TARGETS = tests/test-tokenizer tests/test-text-encoder
+TEST_TARGETS = tests/test-tokenizer tests/test-text-encoder tests/test-coarse-encoder
 
 default: $(BUILD_TARGETS)
 
@@ -318,3 +318,6 @@ tests/test-tokenizer: tests/test-tokenizer.cpp ggml.o bark.o encodec.o $(OBJS)
 
 tests/test-text-encoder: tests/test-text-encoder.cpp ggml.o bark.o encodec.o $(OBJS)
  $(CXX) $(CXXFLAGS) $(filter-out %.txt,$^) -o $@ $(LDFLAGS)
+
+tests/test-coarse-encoder: tests/test-coarse-encoder.cpp ggml.o bark.o encodec.o $(OBJS)
+ $(CXX) $(CXXFLAGS) $(filter-out %.txt,$^) -o $@ $(LDFLAGS)

bark.cpp

@@ -1251,15 +1251,15 @@ std::vector<bark_vocab::id> bark_forward_text_encoder(
 
  input.clear();
 
- bark_vocab::id sampled_id = gpt_sample(logits, temp, rng, &eos_p);
+ bark_vocab::id next = gpt_sample(logits, temp, rng, &eos_p);
 
- if (early_stop && ((sampled_id == SEMANTIC_VOCAB_SIZE) || (eos_p > min_eos_p)))
+ if (early_stop && ((next == SEMANTIC_VOCAB_SIZE) || (eos_p > min_eos_p)))
  break;
 
- input.push_back(sampled_id);
- out.push_back(sampled_id);
+ input.push_back(next);
+ out.push_back(next);
 
- printf("%d ", sampled_id);
+ printf("%d ", next);
  fflush(stdout);
  }
 
@@ -1268,6 +1268,107 @@ std::vector<bark_vocab::id> bark_forward_text_encoder(
  return out;
 }
 
+std::vector<std::vector<bark_vocab::id>> bark_forward_coarse_encoder(
+ const std::vector<bark_vocab::id> & tokens,
+ const gpt_model model,
+ const int n_threads,
+ const float temp,
+ const bool early_stop,
+ const float min_eos_p,
+ const int max_coarse_history,
+ const int sliding_window_size) {
+ std::vector<std::vector<bark_vocab::id>> out_coarse(N_COARSE_CODEBOOKS);
+ std::vector<bark_vocab::id> out;
+
+ std::mt19937 rng(0);
+
+ float semantic_to_coarse_ratio = COARSE_RATE_HZ / SEMANTIC_RATE_HZ * N_COARSE_CODEBOOKS;
+ int max_semantic_history = floorf(max_coarse_history / semantic_to_coarse_ratio);
+
+ int n_steps = floorf(tokens.size() * semantic_to_coarse_ratio / N_COARSE_CODEBOOKS) * N_COARSE_CODEBOOKS;
+ int step_ix = 0;
+
+ BARK_ASSERT(n_steps > 0);
+ BARK_ASSERT(n_steps % N_COARSE_CODEBOOKS == 0);
+
+ int n_window_steps = ceilf(static_cast<float>(n_steps) / sliding_window_size);
+
+ std::vector<bark_vocab::id> input = tokens;
+ std::vector<float> logits;
+
+ // dry run to estimate mem_per_token
+ size_t mem_per_token = 0;
+ gpt_eval(model, n_threads, 0, false, { 0, 1, 2, 3 }, logits, mem_per_token);
+
+ for(int i = 0; i < n_window_steps; i++) {
+ int semantic_ix = roundf(n_steps / semantic_to_coarse_ratio);
+
+ std::vector<bark_vocab::id> input_in(
+ input.begin() + std::max(semantic_ix-max_semantic_history, 0),
+ input.end()
+ );
+ size_t original_size = input_in.size();
+ input_in.resize(256);
+
+ // padding from the right side
+ for (int ix = original_size; ix < 256; ix++)
+ input_in[ix] = COARSE_SEMANTIC_PAD_TOKEN;
+
+ input_in.push_back(COARSE_INFER_TOKEN);
+
+ // concatenate input_in and input_coarse
+ input_in.insert(
+ input_in.end(),
+ std::make_move_iterator(out.end() - std::min(max_coarse_history, (int) out.size())),
+ std::make_move_iterator(out.end())
+ );
+
+ int n_past = 0;
+ mem_per_token *= 1.1; // context length is growing, mem_per_token must grow as well
+
+ for (int j = 0; j < sliding_window_size; j++) {
+ if (step_ix >= n_steps)
+ continue;
+
+ gpt_eval(model, n_threads, n_past, false, input_in, logits, mem_per_token);
+
+ n_past += input_in.size();
+ input_in.clear();
+
+ bool is_major = step_ix % N_COARSE_CODEBOOKS == 0;
+ int start_ix = SEMANTIC_VOCAB_SIZE + (1 - is_major) * CODEBOOK_SIZE;
+ int end_ix = SEMANTIC_VOCAB_SIZE + (2 - is_major) * CODEBOOK_SIZE;
+ std::vector<float> relevant_logits(logits.begin() + start_ix, logits.begin() + end_ix);
+
+ bark_vocab::id next = gpt_sample(relevant_logits, temp, rng, NULL);
+ next += start_ix;
+
+ input_in.push_back(next);
+ out.push_back(next);
+
+ printf("%d ", next);
+ fflush(stdout);
+
+ step_ix += 1;
+ }
+ }
+
+ BARK_ASSERT((int) out.size() == n_steps);
+ BARK_ASSERT(out.size() % N_COARSE_CODEBOOKS == 0);
+
+ for (int i = 0; i < (int) out.size(); i++) {
+ if (i % 2 == 0)
+ out_coarse[0].push_back(out[i] - SEMANTIC_VOCAB_SIZE);
+ else
+ out_coarse[1].push_back(out[i] - SEMANTIC_VOCAB_SIZE - CODEBOOK_SIZE);
+ }
+
+ printf("\n\ncoarse sequence length: %zu\n\n", out.size());
+
+ return out_coarse;
+}
+
+
 bool bark_generate_audio(
  bark_model model,
  const bark_vocab& vocab,
@@ -1334,92 +1435,8 @@ bool bark_generate_audio(
  tokens, model.text_model, n_threads, temp, early_stop, min_eos_p);
 
  // coarse encoding (coarse model)
- std::vector<std::vector<bark_vocab::id>> out_coarse(N_COARSE_CODEBOOKS);
- {
- std::vector<bark_vocab::id> out;
- float semantic_to_coarse_ratio = COARSE_RATE_HZ / SEMANTIC_RATE_HZ * N_COARSE_CODEBOOKS;
- int max_semantic_history = floorf(max_coarse_history / semantic_to_coarse_ratio);
-
- int n_steps = floorf(out_semantic.size() * semantic_to_coarse_ratio / N_COARSE_CODEBOOKS) * N_COARSE_CODEBOOKS;
- int step_ix = 0;
-
- BARK_ASSERT(n_steps > 0);
- BARK_ASSERT(n_steps % N_COARSE_CODEBOOKS == 0);
-
- int n_window_steps = ceilf(static_cast<float>(n_steps) / sliding_window_size);
-
- std::vector<bark_vocab::id> input = out_semantic;
- std::vector<float> logits;
-
- // dry run to estimate mem_per_token
- size_t mem_per_token = 0;
- gpt_eval(model.coarse_model, n_threads, 0, false, { 0, 1, 2, 3 }, logits, mem_per_token);
-
- for(int i = 0; i < n_window_steps; i++) {
- int semantic_ix = roundf(n_steps / semantic_to_coarse_ratio);
-
- std::vector<bark_vocab::id> input_in(input.begin() + std::max(semantic_ix-max_semantic_history, 0), input.end());
- size_t original_size = input_in.size();
- input_in.resize(256);
-
- // padding from the right side
- for (int ix = original_size; ix < 256; ix++)
- input_in[ix] = COARSE_SEMANTIC_PAD_TOKEN;
-
- input_in.push_back(COARSE_INFER_TOKEN);
-
- // concatenate input_in and input_coarse
- input_in.insert(
- input_in.end(),
- std::make_move_iterator(out.end() - std::min(max_coarse_history, (int) out.size())),
- std::make_move_iterator(out.end())
- );
-
- int n_past = 0;
- mem_per_token *= 1.1; // context length is growing, mem_per_token must grow as well
-
- for (int j = 0; j < sliding_window_size; j++) {
- if (step_ix >= n_steps)
- continue;
-
- gpt_eval(model.coarse_model, n_threads, n_past, false, input_in, logits, mem_per_token);
-
- n_past += input_in.size();
- input_in.clear();
-
- bool is_major = step_ix % N_COARSE_CODEBOOKS == 0;
- int start_ix = SEMANTIC_VOCAB_SIZE + (1 - is_major) * CODEBOOK_SIZE;
- int end_ix = SEMANTIC_VOCAB_SIZE + (2 - is_major) * CODEBOOK_SIZE;
- std::vector<float> relevant_logits(logits.begin() + start_ix, logits.begin() + end_ix);
-
- bark_vocab::id sampled_id = gpt_sample(relevant_logits, temp, rng, NULL);
- sampled_id += start_ix;
-
- input_in.push_back(sampled_id);
- out.push_back(sampled_id);
-
- printf("%d ", sampled_id);
- fflush(stdout);
-
- step_ix += 1;
- }
- }
-
- BARK_ASSERT((int) out.size() == n_steps);
- BARK_ASSERT(out.size() % N_COARSE_CODEBOOKS == 0);
-
- // gen_coarse_audio_arr = gen_coarse_arr.reshape(-1, N_COARSE_CODEBOOKS).T - SEMANTIC_VOCAB_SIZE
- int mid = out.size() / 2;
- for(int ix = 0; ix < (int) out.size(); ix++) {
- if(ix < mid)
- out_coarse[0].push_back(out[ix] - SEMANTIC_VOCAB_SIZE);
- else
- // !! this only works when CODEBOOK_SIZE = 2
- out_coarse[1].push_back(out[ix] - SEMANTIC_VOCAB_SIZE - CODEBOOK_SIZE);
- }
-
- printf("\n\ncoarse sequence length: %zu\n\n", out.size());
- }
+ std::vector<std::vector<bark_vocab::id>> out_coarse = bark_forward_coarse_encoder(
+ out_semantic, model.coarse_model, n_threads, temp, early_stop, min_eos_p, max_coarse_history, sliding_window_size);
 
  // fine encoding (fine model)
  std::vector<std::vector<bark_vocab::id>> out_fine;

bark.h

@@ -154,3 +154,13 @@ std::vector<bark_vocab::id> bark_forward_text_encoder(
  const float temp,
  const bool early_stop,
  const float min_eos_p);
+
+std::vector<std::vector<bark_vocab::id>> bark_forward_coarse_encoder(
+ const std::vector<bark_vocab::id> & tokens,
+ const gpt_model model,
+ const int n_threads,
+ const float temp,
+ const bool early_stop,
+ const float min_eos_p,
+ const int max_coarse_history,
+ const int sliding_window_size);

tests/CMakeLists.txt

@@ -8,3 +8,4 @@ endfunction()
 
 bark_add_test(test-tokenizer.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../ggml_weights/ggml_vocab.bin)
 bark_add_test(test-text-encoder.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../ggml_weights/ggml_weights_text.bin)
+bark_add_test(test-coarse-encoder.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../ggml_weights/ggml_weights_coarse.bin)

tests/test-coarse-encoder.cpp

@@ -0,0 +1,91 @@
+#include "bark.h"
+
+#include <cstdio>
+#include <string>
+#include <map>
+#include <random>
+#include <vector>
+
+static const std::map<std::vector<bark_vocab::id>, std::vector<std::vector<bark_vocab::id>>> & k_tests()
+{
+ static const std::vector<bark_vocab::id> seq1 = { 215, 1988, 3275, 1898, 1898, 1898, 9372, 9372, 222, 334, 8568, 8568, 7963, 222, 8568, 55, 7963, 1270, 55, 1283, 1283, 222, 1283, 1283, 1283, 55, 1283, 5960, 5960, 5960, 5960, 5960, 5960, 5960, 231, 5960, 5960, 5960, 5960, 5960, 5960, 5960, 5960, 5960, 5960, 5960, 5960, 5960, 340, 5960, 5960, 5960, 5960, 1374, 4193, 4193, 9323, 1374, 1374, 1374, 1374, 4193, 1374, 4193, 1374, 1374, 4193, 1374, 231, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 8328, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 9318, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374, 1374 };
+ static const std::vector<bark_vocab::id> seq2 = { 59, 28, 28, 107, 7409, 1999, 7695, 6486, 6486, 5836, 5836, 5836, 873, 2585, 92, 92, 59, 28, 28, 107, 315, 5623, 1025, 10, 173, 125, 7385, 147, 147, 3689, 302, 9600, 6876, 6876, 321, 41, 164, 1367, 739, 41, 10, 140, 140, 6202, 6051, 6051, 4071, 9804, 8583, 677, 3, 17, 113, 9414, 5419, 5419, 3831, 3663, 3663, 3663, 2224, 2224, 2224, 73, 9144, 9144, 1667, 1997, 1957, 1093, 825, 175, 175, 1087, 736, 1233, 230, 147, 147, 230, 230, 230, 230, 230, 528, 528, 528, 528, 528, 528, 528, 528, 528, 528, 528, 528, 528, 528, 528, 1613, 528, 1613, 1613, 1613, 1613, 1613, 1613, 1613, 1613, 1613, 1613, 1613, 2009, 2009 };
+ static const std::vector<bark_vocab::id> seq3 = { 10, 10, 560, 10, 9602, 10, 10, 10, 302, 2363, 2919, 6860, 5127, 7134, 7134, 3934, 3934, 3352, 3352, 3507, 50, 10, 27, 27, 3320, 6107, 9891, 9891, 9891, 321, 41, 4287, 5667, 6152, 6152, 557, 1228, 12, 12, 200, 59, 28, 28, 28, 28, 1133, 9569, 5920, 1424, 1424, 51, 51, 682, 3820, 2107, 6059, 348, 210, 10, 10, 5, 2187, 7842, 988, 1728, 1728, 438, 366, 50, 27, 27, 181, 181, 7352, 9725, 4431, 6445, 2428, 41, 41, 41, 5119, 6557, 4212, 3963, 26, 26, 934, 1025, 1024, 173, 10, 41, 5467, 6684, 6684, 6684, 4958, 41, 298, 5982, 5982, 526, 3219, 122, 181, 10, 10, 884, 3446, 2599, 4478, 4478, 2549 };
+
+ static const std::vector<std::vector<bark_vocab::id>> ans1 = { {}, {} };
+ static const std::vector<std::vector<bark_vocab::id>> ans2 = { {}, {} };
+ static const std::vector<std::vector<bark_vocab::id>> ans3 = { {}, {} };
+
+ static std::map<std::vector<bark_vocab::id>, std::vector<std::vector<bark_vocab::id>>> _k_tests = {
+ // { seq1, ans1 }, // hello world
+ // { seq2, ans2 }, // this is an audio
+ { seq3, ans3 }, // You cannot, sir, take from me anything
+ };
+ return _k_tests;
+};
+
+int main(int argc, char** argv) {
+ if (argc < 2) {
+ fprintf(stderr, "Usage: %s <model-file>\n", argv[0]);
+ return 1;
+ }
+
+ const std::string fname = argv[1];
+
+ gpt_model model;
+
+ const int n_threads = 4;
+ const float min_eos_p = 0.2;
+ const float temp = 0.7;
+
+ const int max_coarse_history = 630;
+ const int sliding_window_size = 60;
+
+ printf("%s: reading bark coarse model\n", __func__);
+ if(!gpt_model_load(fname, model)) {
+ fprintf(stderr, "%s: invalid model file '%s'\n", __func__, fname.c_str());
+ return 1;
+ }
+
+ for (const auto & test_kv : k_tests()) {
+ std::vector<std::vector<bark_vocab::id>> res = bark_forward_coarse_encoder(
+ test_kv.first, model, n_threads, temp, true, min_eos_p, max_coarse_history, sliding_window_size);
+
+ bool correct = res.size() == test_kv.second.size();
+
+ for (int i = 0; i < (int) res.size() && correct; ++i) {
+ correct = res[i].size() == test_kv.second[i].size();
+ for (int j = 0; j < (int) res[i].size() && correct; j++) {
+ if (res[i][j] != test_kv.second[i][j]) {
+ correct = false;
+ }
+ }
+ }
+
+ if (!correct) {
+ fprintf(stderr, "%s : failed test \n", __func__);
+ fprintf(stderr, "%s : expected tokens (n=%zu): ", __func__, test_kv.second.size());
+ for (int i = 0; i < (int) test_kv.second.size(); i++) {
+ for (int j = 0; j < (int) test_kv.second[i].size(); j++) {
+ fprintf(stderr, "%d ", test_kv.second[i][j]);
+ }
+ fprintf(stderr, "\n");
+ }
+ fprintf(stderr, "\n");
+ fprintf(stderr, "%s : got tokens (n=%zu): ", __func__, res.size());
+ for (int i = 0; i < (int) res.size(); i++) {
+ for (int j = 0; j < (int) res[i].size(); j++) {
+ fprintf(stderr, "%d ", res[i][j]);
+ }
+ fprintf(stderr, "\n");
+ }
+ fprintf(stderr, "\n");
+
+ return 3;
+ }
+ }
+
+ fprintf(stderr, "%s : tests passed successfully.\n", __func__);
+
+ return 0;
+}