diff --git a/tests/clone-CPU-BF16.log b/tests/clone-CPU-BF16.log index 7a29efa..254108c 100644 --- a/tests/clone-CPU-BF16.log +++ b/tests/clone-CPU-BF16.log @@ -1,9 +1,9 @@ -[Qwen] qwentts.cpp bcac463 (2026-07-05) +[Qwen] qwentts.cpp 53c84fe (2026-07-20) ggml_cuda_init: found 1 CUDA devices (Total VRAM: 97247 MiB): Device 0: NVIDIA RTX PRO 6000 Blackwell Workstation Edition, compute capability 12.0, VMM: yes, VRAM: 97247 MiB load_backend: loaded CUDA backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cuda.so ggml_vulkan: Found 1 Vulkan devices: -ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 +ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | fp4: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 load_backend: loaded Vulkan backend from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cpu-zen4.so [Load] Talker backend: CPU (CPU threads: 16) @@ -23,9 +23,9 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [DAC] Loaded: 4 blocks (strides 8/5/4/3), 24 kHz mono out, weights 142.9 MB [WeightCtx] Loaded 2 tensors, 3.0 MB into backend [Pipeline] Ready: hop 1920 samples @ 24000 Hz mono, 16 codebooks @ 12.5 Hz -[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096 -> 896 MB -[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16 -> 0 MB -[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=off clamp_fp16=off +[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096, 1 sets -> 896 MB +[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16, 1 sets -> 0 MB +[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=off clamp_fp16=off max_batch=1 [BPE] Loaded from GGUF: 151676 vocab, 151291 merges, eos_id=151643 [BPE] Registered 5 arch special tokens (total specials=6) [WAV] Read buffer: 380672 samples, 22050 Hz, 1 ch, 16 bit @@ -33,7 +33,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Audio-Resample] Done: 380672 -> 414337 samples [WeightCtx] Loaded 76 tensors, 34.9 MB into backend [SpeakerEncoder] Loaded: enc_dim=2048 sr=24000 mel_dim=128 hidden=512 mfa=1536 asp_attn=128 se=128 scale=8 -[Pipeline] Speaker encoder lazy loaded in 5 ms +[Pipeline] Speaker encoder lazy loaded in 6 ms [Debug] mel-spk: [1618, 128] first4: -6.330339 -6.170074 -6.011195 -6.243161 [Debug] mel-hann: [1024] first4: 0.000000 0.000009 0.000038 0.000085 [Debug] mel-basis: [128, 513] first4: 0.000000 0.033550 0.008569 0.000000 @@ -51,7 +51,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [EncDownsample] Loaded: k=4 stride=2, 512 -> 512 channels, weights 2.0 MB [WeightCtx] Loaded 20 tensors, 34.0 MB into backend [EncQuantizer] Loaded: 16 codebooks (1 semantic + 15 acoustic), 2048 entries x 256 dim, hidden 512, weights 34.0 MB -[Pipeline] Codec encoder lazy loaded in 45 ms +[Pipeline] Codec encoder lazy loaded in 53 ms [Debug] audio-input: [412800] first4: -0.000546 -0.001162 -0.000871 0.000340 [Debug] seanet-out: [430, 512] first4: -0.112500 13.760122 -0.329032 2.993861 [Debug] enc-transformer-out: [430, 512] first4: 0.526661 3.208130 -0.164395 0.605249 @@ -76,60 +76,31 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Debug] talker-hidden-prefill-l27: [225, 2048] first4: 7.054070 -34.409252 -7.069129 28.624571 [Debug] talker-hidden-prefill-final: [225, 2048] first4: 0.240086 -1.222942 -0.223565 0.888025 [Debug] talker-logits-prefill: [3072] first4: 1.339535 -7.642556 -6.988994 -10.529062 +[Batch] Admit slot=0 T_ctx=225 prefill=391.8 ms build=2.3 ms (stall for 0 active slots) [Sample] step=0 c0=27 u=-1.0000000000 subseq=0 -[Sample-CP] g=0 c=1408 u=-1.0000000000 subseq=1 -[Sample-CP] g=1 c=1824 u=-1.0000000000 subseq=2 -[Sample-CP] g=2 c=400 u=-1.0000000000 subseq=3 -[Sample-CP] g=3 c=442 u=-1.0000000000 subseq=4 -[Sample-CP] g=4 c=1506 u=-1.0000000000 subseq=5 -[Sample-CP] g=5 c=1085 u=-1.0000000000 subseq=6 -[Sample-CP] g=6 c=470 u=-1.0000000000 subseq=7 -[Sample-CP] g=7 c=1713 u=-1.0000000000 subseq=8 -[Sample-CP] g=8 c=843 u=-1.0000000000 subseq=9 -[Sample-CP] g=9 c=391 u=-1.0000000000 subseq=10 -[Sample-CP] g=10 c=829 u=-1.0000000000 subseq=11 -[Sample-CP] g=11 c=411 u=-1.0000000000 subseq=12 -[Sample-CP] g=12 c=866 u=-1.0000000000 subseq=13 -[Sample-CP] g=13 c=122 u=-1.0000000000 subseq=14 -[Sample-CP] g=14 c=365 u=-1.0000000000 subseq=15 [Debug] codes-step0: [16] first4: 27.000000 1408.000000 1824.000000 400.000000 [Debug] next-emb-step0: [2048] first4: 0.073552 -0.034240 -0.008814 0.042007 [Debug] talker-hidden-step1: [2048] first4: 0.576087 2.576824 -0.018687 0.235103 [Sample] step=1 c0=1902 u=-1.0000000000 subseq=16 -[Sample-CP] g=0 c=1924 u=-1.0000000000 subseq=17 -[Sample-CP] g=1 c=2025 u=-1.0000000000 subseq=18 -[Sample-CP] g=2 c=324 u=-1.0000000000 subseq=19 -[Sample-CP] g=3 c=350 u=-1.0000000000 subseq=20 -[Sample-CP] g=4 c=985 u=-1.0000000000 subseq=21 -[Sample-CP] g=5 c=853 u=-1.0000000000 subseq=22 -[Sample-CP] g=6 c=329 u=-1.0000000000 subseq=23 -[Sample-CP] g=7 c=564 u=-1.0000000000 subseq=24 -[Sample-CP] g=8 c=692 u=-1.0000000000 subseq=25 -[Sample-CP] g=9 c=1702 u=-1.0000000000 subseq=26 -[Sample-CP] g=10 c=437 u=-1.0000000000 subseq=27 -[Sample-CP] g=11 c=729 u=-1.0000000000 subseq=28 -[Sample-CP] g=12 c=344 u=-1.0000000000 subseq=29 -[Sample-CP] g=13 c=85 u=-1.0000000000 subseq=30 -[Sample-CP] g=14 c=1435 u=-1.0000000000 subseq=31 -[Pipeline] Generated 8 frames -[Pipeline] Generated 16 frames -[Pipeline] Generated 24 frames -[Pipeline] Generated 32 frames -[Pipeline] Generated 40 frames -[Pipeline] Generated 48 frames -[Pipeline] Generated 56 frames -[Pipeline] Generated 64 frames +[Pipeline] Generated 8 frames (slot 0) +[Pipeline] Generated 16 frames (slot 0) +[Pipeline] Generated 24 frames (slot 0) +[Pipeline] Generated 32 frames (slot 0) +[Pipeline] Generated 40 frames (slot 0) +[Pipeline] Generated 48 frames (slot 0) +[Pipeline] Generated 56 frames (slot 0) +[Pipeline] Generated 64 frames (slot 0) [Pipeline] Generation done : 64 frames [Debug] codes-full: [64, 16] first4: 27.000000 1408.000000 1824.000000 400.000000 [Debug] output-audio: [122880] first4: 0.000529 0.000353 -0.000865 -0.000453 [Perf] PromptBuild 2.3 ms -[Perf] Prefill 380.6 ms (T_ctx prefill) -[Perf] TTFA 408.5 ms (first frame codes) -[Perf] TalkerDecode 2916.5 ms (64 frames, 45.57 ms/frame) -[Perf] CodePredictor 1464.6 ms (22.88 ms/frame) +[Perf] Prefill 391.8 ms (T_ctx prefill) +[Perf] TTFA 1549.7 ms (first frame codes) +[Perf] TalkerDecode 2864.7 ms (64 frames, 44.76 ms/frame) +[Perf] CodePredictor 1446.2 ms (22.60 ms/frame) [Perf] HostCompose 0.1 ms (c0 sample + next emb) -[Perf] CodecDecode 1034.6 ms -[Perf] Total 5800.0 ms (64 frames, 68.46 ms/frame AR, audio 5.12 s, RTF 1.133) +[Perf] CodecDecode 1112.5 ms +[Perf] Total 6949.6 ms (64 frames, 67.36 ms/frame AR, audio 5.12 s, RTF 1.357) [WAV] Wrote cpp/clone/clone-cpp.wav: 122880 samples, 24000 Hz, mono S16 [Pipeline] Wrote 122880 samples (5.12 s) -> cpp/clone/clone-cpp.wav [Input] Prompt: 100 chars: qwentts.cpp is a minimal C++17 and GGML port of Qwen3-TTS fo... @@ -151,15 +122,15 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Cossim] PromptIDs exact: 100.00% (38 values) [Cossim] RefCodes exact: 86.77% (3440 values) [Cossim] Embed cos: 0.978030 max: 1.8825e-01 mean: 6.5145e-03 -[Cossim] TrailingText cos: 1.000000 max: 3.5048e-05 mean: 2.4958e-06 -[Cossim] TTSPadEmbed cos: 1.000000 max: 3.5048e-05 mean: 2.4958e-06 +[Cossim] TrailingText cos: 1.000000 max: 3.5048e-05 mean: 2.4957e-06 +[Cossim] TTSPadEmbed cos: 1.000000 max: 3.5048e-05 mean: 2.4957e-06 [Cossim] L0 cos: 0.989591 max: 7.7401e-01 mean: 8.9719e-03 -[Cossim] L7 cos: 0.999993 max: 1.2846e+01 mean: 2.3289e-02 -[Cossim] L14 cos: 0.999976 max: 1.2850e+01 mean: 5.2409e-02 +[Cossim] L7 cos: 0.999993 max: 1.2845e+01 mean: 2.3289e-02 +[Cossim] L14 cos: 0.999976 max: 1.2849e+01 mean: 5.2409e-02 [Cossim] L21 cos: 0.999696 max: 1.7400e+01 mean: 2.0189e-01 [Cossim] L27 cos: 0.998836 max: 4.6276e+01 mean: 4.6497e-01 [Cossim] Final cos: 0.993939 max: 1.9954e+01 mean: 1.5107e-01 -[Cossim] Logits cos: 0.999804 max: 5.9451e-01 mean: 7.9720e-02 +[Cossim] Logits cos: 0.999804 max: 5.9450e-01 mean: 7.9720e-02 [Cossim] NextEmbStep0 cos: 0.967765 max: 6.3429e-02 mean: 1.2632e-02 [Cossim] TalkerHiddenStep1 cos: 0.998217 max: 8.0423e-01 mean: 1.3036e-01 [Cossim] MelHann cos: 1.000000 max: 1.7881e-07 mean: 3.2043e-08 diff --git a/tests/clone-CPU-F32.log b/tests/clone-CPU-F32.log index cbe4b07..cd1a47f 100644 --- a/tests/clone-CPU-F32.log +++ b/tests/clone-CPU-F32.log @@ -1,9 +1,9 @@ -[Qwen] qwentts.cpp bcac463 (2026-07-05) +[Qwen] qwentts.cpp 53c84fe (2026-07-20) ggml_cuda_init: found 1 CUDA devices (Total VRAM: 97247 MiB): Device 0: NVIDIA RTX PRO 6000 Blackwell Workstation Edition, compute capability 12.0, VMM: yes, VRAM: 97247 MiB load_backend: loaded CUDA backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cuda.so ggml_vulkan: Found 1 Vulkan devices: -ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 +ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | fp4: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 load_backend: loaded Vulkan backend from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cpu-zen4.so [Load] Talker backend: CPU (CPU threads: 16) @@ -23,9 +23,9 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [DAC] Loaded: 4 blocks (strides 8/5/4/3), 24 kHz mono out, weights 142.9 MB [WeightCtx] Loaded 2 tensors, 3.0 MB into backend [Pipeline] Ready: hop 1920 samples @ 24000 Hz mono, 16 codebooks @ 12.5 Hz -[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096 -> 896 MB -[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16 -> 0 MB -[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=off clamp_fp16=off +[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096, 1 sets -> 896 MB +[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16, 1 sets -> 0 MB +[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=off clamp_fp16=off max_batch=1 [BPE] Loaded from GGUF: 151676 vocab, 151291 merges, eos_id=151643 [BPE] Registered 5 arch special tokens (total specials=6) [WAV] Read buffer: 380672 samples, 22050 Hz, 1 ch, 16 bit @@ -33,7 +33,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Audio-Resample] Done: 380672 -> 414337 samples [WeightCtx] Loaded 76 tensors, 45.8 MB into backend [SpeakerEncoder] Loaded: enc_dim=2048 sr=24000 mel_dim=128 hidden=512 mfa=1536 asp_attn=128 se=128 scale=8 -[Pipeline] Speaker encoder lazy loaded in 8 ms +[Pipeline] Speaker encoder lazy loaded in 7 ms [Debug] mel-spk: [1618, 128] first4: -6.330339 -6.170074 -6.011195 -6.243161 [Debug] mel-hann: [1024] first4: 0.000000 0.000009 0.000038 0.000085 [Debug] mel-basis: [128, 513] first4: 0.000000 0.033550 0.008569 0.000000 @@ -51,7 +51,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [EncDownsample] Loaded: k=4 stride=2, 512 -> 512 channels, weights 2.0 MB [WeightCtx] Loaded 20 tensors, 34.0 MB into backend [EncQuantizer] Loaded: 16 codebooks (1 semantic + 15 acoustic), 2048 entries x 256 dim, hidden 512, weights 34.0 MB -[Pipeline] Codec encoder lazy loaded in 58 ms +[Pipeline] Codec encoder lazy loaded in 46 ms [Debug] audio-input: [412800] first4: -0.000546 -0.001162 -0.000871 0.000340 [Debug] seanet-out: [430, 512] first4: -0.120292 13.742489 -0.348104 3.009387 [Debug] enc-transformer-out: [430, 512] first4: 0.526076 3.125370 -0.189241 0.630764 @@ -76,60 +76,31 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Debug] talker-hidden-prefill-l27: [225, 2048] first4: 7.133089 -34.474190 -7.108908 28.674923 [Debug] talker-hidden-prefill-final: [225, 2048] first4: 0.242892 -1.225840 -0.224931 0.890014 [Debug] talker-logits-prefill: [3072] first4: 1.330122 -7.768507 -6.935926 -10.338600 +[Batch] Admit slot=0 T_ctx=225 prefill=701.2 ms build=4.3 ms (stall for 0 active slots) [Sample] step=0 c0=27 u=-1.0000000000 subseq=0 -[Sample-CP] g=0 c=1408 u=-1.0000000000 subseq=1 -[Sample-CP] g=1 c=1824 u=-1.0000000000 subseq=2 -[Sample-CP] g=2 c=400 u=-1.0000000000 subseq=3 -[Sample-CP] g=3 c=442 u=-1.0000000000 subseq=4 -[Sample-CP] g=4 c=1506 u=-1.0000000000 subseq=5 -[Sample-CP] g=5 c=1085 u=-1.0000000000 subseq=6 -[Sample-CP] g=6 c=470 u=-1.0000000000 subseq=7 -[Sample-CP] g=7 c=1713 u=-1.0000000000 subseq=8 -[Sample-CP] g=8 c=843 u=-1.0000000000 subseq=9 -[Sample-CP] g=9 c=391 u=-1.0000000000 subseq=10 -[Sample-CP] g=10 c=829 u=-1.0000000000 subseq=11 -[Sample-CP] g=11 c=411 u=-1.0000000000 subseq=12 -[Sample-CP] g=12 c=866 u=-1.0000000000 subseq=13 -[Sample-CP] g=13 c=122 u=-1.0000000000 subseq=14 -[Sample-CP] g=14 c=365 u=-1.0000000000 subseq=15 [Debug] codes-step0: [16] first4: 27.000000 1408.000000 1824.000000 400.000000 [Debug] next-emb-step0: [2048] first4: 0.073561 -0.034233 -0.008814 0.042004 [Debug] talker-hidden-step1: [2048] first4: 0.584086 2.683544 -0.065920 0.135428 [Sample] step=1 c0=1902 u=-1.0000000000 subseq=16 -[Sample-CP] g=0 c=1924 u=-1.0000000000 subseq=17 -[Sample-CP] g=1 c=2025 u=-1.0000000000 subseq=18 -[Sample-CP] g=2 c=324 u=-1.0000000000 subseq=19 -[Sample-CP] g=3 c=350 u=-1.0000000000 subseq=20 -[Sample-CP] g=4 c=985 u=-1.0000000000 subseq=21 -[Sample-CP] g=5 c=853 u=-1.0000000000 subseq=22 -[Sample-CP] g=6 c=329 u=-1.0000000000 subseq=23 -[Sample-CP] g=7 c=564 u=-1.0000000000 subseq=24 -[Sample-CP] g=8 c=692 u=-1.0000000000 subseq=25 -[Sample-CP] g=9 c=1702 u=-1.0000000000 subseq=26 -[Sample-CP] g=10 c=437 u=-1.0000000000 subseq=27 -[Sample-CP] g=11 c=729 u=-1.0000000000 subseq=28 -[Sample-CP] g=12 c=344 u=-1.0000000000 subseq=29 -[Sample-CP] g=13 c=85 u=-1.0000000000 subseq=30 -[Sample-CP] g=14 c=1435 u=-1.0000000000 subseq=31 -[Pipeline] Generated 8 frames -[Pipeline] Generated 16 frames -[Pipeline] Generated 24 frames -[Pipeline] Generated 32 frames -[Pipeline] Generated 40 frames -[Pipeline] Generated 48 frames -[Pipeline] Generated 56 frames -[Pipeline] Generated 64 frames +[Pipeline] Generated 8 frames (slot 0) +[Pipeline] Generated 16 frames (slot 0) +[Pipeline] Generated 24 frames (slot 0) +[Pipeline] Generated 32 frames (slot 0) +[Pipeline] Generated 40 frames (slot 0) +[Pipeline] Generated 48 frames (slot 0) +[Pipeline] Generated 56 frames (slot 0) +[Pipeline] Generated 64 frames (slot 0) [Pipeline] Generation done : 64 frames [Debug] codes-full: [64, 16] first4: 27.000000 1408.000000 1824.000000 400.000000 [Debug] output-audio: [122880] first4: 0.000450 0.000327 -0.000890 -0.000492 -[Perf] PromptBuild 3.9 ms -[Perf] Prefill 632.5 ms (T_ctx prefill) -[Perf] TTFA 700.2 ms (first frame codes) -[Perf] TalkerDecode 5539.4 ms (64 frames, 86.55 ms/frame) -[Perf] CodePredictor 3887.4 ms (60.74 ms/frame) +[Perf] PromptBuild 4.3 ms +[Perf] Prefill 701.2 ms (T_ctx prefill) +[Perf] TTFA 1963.7 ms (first frame codes) +[Perf] TalkerDecode 5589.4 ms (64 frames, 87.33 ms/frame) +[Perf] CodePredictor 3961.5 ms (61.90 ms/frame) [Perf] HostCompose 0.1 ms (c0 sample + next emb) -[Perf] CodecDecode 1033.3 ms -[Perf] Total 11098.2 ms (64 frames, 147.30 ms/frame AR, audio 5.12 s, RTF 2.168) +[Perf] CodecDecode 1026.8 ms +[Perf] Total 12474.5 ms (64 frames, 149.23 ms/frame AR, audio 5.12 s, RTF 2.436) [WAV] Wrote cpp/clone/clone-cpp.wav: 122880 samples, 24000 Hz, mono S16 [Pipeline] Wrote 122880 samples (5.12 s) -> cpp/clone/clone-cpp.wav [Input] Prompt: 100 chars: qwentts.cpp is a minimal C++17 and GGML port of Qwen3-TTS fo... @@ -151,17 +122,17 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Cossim] PromptIDs exact: 100.00% (38 values) [Cossim] RefCodes exact: 96.72% (3440 values) [Cossim] Embed cos: 0.994609 max: 1.5292e-01 mean: 1.6562e-03 -[Cossim] TrailingText cos: 1.000000 max: 2.9802e-08 mean: 5.6171e-10 -[Cossim] TTSPadEmbed cos: 1.000000 max: 2.9802e-08 mean: 5.6171e-10 +[Cossim] TrailingText cos: 1.000000 max: 2.9802e-08 mean: 6.7572e-10 +[Cossim] TTSPadEmbed cos: 1.000000 max: 2.9802e-08 mean: 6.7572e-10 [Cossim] L0 cos: 0.997321 max: 7.2147e-01 mean: 2.5377e-03 [Cossim] L7 cos: 0.999998 max: 1.2205e+00 mean: 7.8975e-03 [Cossim] L14 cos: 0.999994 max: 2.2472e+00 mean: 2.0513e-02 -[Cossim] L21 cos: 0.999920 max: 1.3671e+01 mean: 8.9237e-02 -[Cossim] L27 cos: 0.999724 max: 2.5151e+01 mean: 2.0175e-01 -[Cossim] Final cos: 0.998553 max: 1.1352e+01 mean: 6.5864e-02 -[Cossim] Logits cos: 0.999966 max: 2.0576e-01 mean: 3.0347e-02 +[Cossim] L21 cos: 0.999920 max: 1.3672e+01 mean: 8.9237e-02 +[Cossim] L27 cos: 0.999724 max: 2.5152e+01 mean: 2.0175e-01 +[Cossim] Final cos: 0.998553 max: 1.1353e+01 mean: 6.5865e-02 +[Cossim] Logits cos: 0.999966 max: 2.0577e-01 mean: 3.0348e-02 [Cossim] NextEmbStep0 cos: 0.967766 max: 6.3431e-02 mean: 1.2632e-02 -[Cossim] TalkerHiddenStep1 cos: 0.999113 max: 4.3387e-01 mean: 9.2054e-02 +[Cossim] TalkerHiddenStep1 cos: 0.999113 max: 4.3387e-01 mean: 9.2055e-02 [Cossim] MelHann cos: 1.000000 max: 1.7881e-07 mean: 3.2043e-08 [Cossim] MelBasis cos: 1.000000 max: 1.7881e-07 mean: 3.9379e-10 [Cossim] MelMag cos: 1.000000 max: 3.1307e-05 mean: 5.6396e-07 diff --git a/tests/clone-CPU-Q4_K_M.log b/tests/clone-CPU-Q4_K_M.log index 089aba1..957063f 100644 --- a/tests/clone-CPU-Q4_K_M.log +++ b/tests/clone-CPU-Q4_K_M.log @@ -1,9 +1,9 @@ -[Qwen] qwentts.cpp bcac463 (2026-07-05) +[Qwen] qwentts.cpp 53c84fe (2026-07-20) ggml_cuda_init: found 1 CUDA devices (Total VRAM: 97247 MiB): Device 0: NVIDIA RTX PRO 6000 Blackwell Workstation Edition, compute capability 12.0, VMM: yes, VRAM: 97247 MiB load_backend: loaded CUDA backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cuda.so ggml_vulkan: Found 1 Vulkan devices: -ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 +ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | fp4: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 load_backend: loaded Vulkan backend from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cpu-zen4.so [Load] Talker backend: CPU (CPU threads: 16) @@ -23,9 +23,9 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [DAC] Loaded: 4 blocks (strides 8/5/4/3), 24 kHz mono out, weights 142.9 MB [WeightCtx] Loaded 2 tensors, 3.0 MB into backend [Pipeline] Ready: hop 1920 samples @ 24000 Hz mono, 16 codebooks @ 12.5 Hz -[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096 -> 896 MB -[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16 -> 0 MB -[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=off clamp_fp16=off +[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096, 1 sets -> 896 MB +[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16, 1 sets -> 0 MB +[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=off clamp_fp16=off max_batch=1 [BPE] Loaded from GGUF: 151676 vocab, 151291 merges, eos_id=151643 [BPE] Registered 5 arch special tokens (total specials=6) [WAV] Read buffer: 380672 samples, 22050 Hz, 1 ch, 16 bit @@ -51,7 +51,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [EncDownsample] Loaded: k=4 stride=2, 512 -> 512 channels, weights 2.0 MB [WeightCtx] Loaded 20 tensors, 34.0 MB into backend [EncQuantizer] Loaded: 16 codebooks (1 semantic + 15 acoustic), 2048 entries x 256 dim, hidden 512, weights 34.0 MB -[Pipeline] Codec encoder lazy loaded in 7 ms +[Pipeline] Codec encoder lazy loaded in 8 ms [Debug] audio-input: [412800] first4: -0.000546 -0.001162 -0.000871 0.000340 [Debug] seanet-out: [430, 512] first4: -0.120292 13.742489 -0.348104 3.009387 [Debug] enc-transformer-out: [430, 512] first4: 0.867046 3.016766 0.059672 0.395116 @@ -76,60 +76,31 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Debug] talker-hidden-prefill-l27: [225, 2048] first4: 4.580647 -33.465454 -6.676281 23.526917 [Debug] talker-hidden-prefill-final: [225, 2048] first4: 0.155397 -1.185542 -0.210456 0.727513 [Debug] talker-logits-prefill: [3072] first4: -0.397744 -4.641821 -7.453447 -10.679588 +[Batch] Admit slot=0 T_ctx=225 prefill=429.9 ms build=2.9 ms (stall for 0 active slots) [Sample] step=0 c0=27 u=-1.0000000000 subseq=0 -[Sample-CP] g=0 c=1408 u=-1.0000000000 subseq=1 -[Sample-CP] g=1 c=1824 u=-1.0000000000 subseq=2 -[Sample-CP] g=2 c=417 u=-1.0000000000 subseq=3 -[Sample-CP] g=3 c=442 u=-1.0000000000 subseq=4 -[Sample-CP] g=4 c=985 u=-1.0000000000 subseq=5 -[Sample-CP] g=5 c=1155 u=-1.0000000000 subseq=6 -[Sample-CP] g=6 c=470 u=-1.0000000000 subseq=7 -[Sample-CP] g=7 c=1311 u=-1.0000000000 subseq=8 -[Sample-CP] g=8 c=1551 u=-1.0000000000 subseq=9 -[Sample-CP] g=9 c=1952 u=-1.0000000000 subseq=10 -[Sample-CP] g=10 c=1492 u=-1.0000000000 subseq=11 -[Sample-CP] g=11 c=272 u=-1.0000000000 subseq=12 -[Sample-CP] g=12 c=124 u=-1.0000000000 subseq=13 -[Sample-CP] g=13 c=408 u=-1.0000000000 subseq=14 -[Sample-CP] g=14 c=549 u=-1.0000000000 subseq=15 [Debug] codes-step0: [16] first4: 27.000000 1408.000000 1824.000000 417.000000 [Debug] next-emb-step0: [2048] first4: 0.035528 0.008631 -0.031989 0.034729 [Debug] talker-hidden-step1: [2048] first4: 0.529886 2.483560 0.540937 0.745365 [Sample] step=1 c0=1902 u=-1.0000000000 subseq=16 -[Sample-CP] g=0 c=1924 u=-1.0000000000 subseq=17 -[Sample-CP] g=1 c=2025 u=-1.0000000000 subseq=18 -[Sample-CP] g=2 c=324 u=-1.0000000000 subseq=19 -[Sample-CP] g=3 c=405 u=-1.0000000000 subseq=20 -[Sample-CP] g=4 c=390 u=-1.0000000000 subseq=21 -[Sample-CP] g=5 c=447 u=-1.0000000000 subseq=22 -[Sample-CP] g=6 c=1051 u=-1.0000000000 subseq=23 -[Sample-CP] g=7 c=564 u=-1.0000000000 subseq=24 -[Sample-CP] g=8 c=1004 u=-1.0000000000 subseq=25 -[Sample-CP] g=9 c=1702 u=-1.0000000000 subseq=26 -[Sample-CP] g=10 c=437 u=-1.0000000000 subseq=27 -[Sample-CP] g=11 c=729 u=-1.0000000000 subseq=28 -[Sample-CP] g=12 c=142 u=-1.0000000000 subseq=29 -[Sample-CP] g=13 c=328 u=-1.0000000000 subseq=30 -[Sample-CP] g=14 c=1341 u=-1.0000000000 subseq=31 -[Pipeline] Generated 8 frames -[Pipeline] Generated 16 frames -[Pipeline] Generated 24 frames -[Pipeline] Generated 32 frames -[Pipeline] Generated 40 frames -[Pipeline] Generated 48 frames -[Pipeline] Generated 56 frames -[Pipeline] Generated 64 frames +[Pipeline] Generated 8 frames (slot 0) +[Pipeline] Generated 16 frames (slot 0) +[Pipeline] Generated 24 frames (slot 0) +[Pipeline] Generated 32 frames (slot 0) +[Pipeline] Generated 40 frames (slot 0) +[Pipeline] Generated 48 frames (slot 0) +[Pipeline] Generated 56 frames (slot 0) +[Pipeline] Generated 64 frames (slot 0) [Pipeline] Generation done : 64 frames [Debug] codes-full: [64, 16] first4: 27.000000 1408.000000 1824.000000 417.000000 [Debug] output-audio: [122880] first4: -0.000160 -0.000284 -0.001051 -0.000894 -[Perf] PromptBuild 3.3 ms -[Perf] Prefill 412.6 ms (T_ctx prefill) -[Perf] TTFA 424.0 ms (first frame codes) -[Perf] TalkerDecode 990.1 ms (64 frames, 15.47 ms/frame) -[Perf] CodePredictor 413.0 ms (6.45 ms/frame) +[Perf] PromptBuild 2.9 ms +[Perf] Prefill 429.9 ms (T_ctx prefill) +[Perf] TTFA 1546.8 ms (first frame codes) +[Perf] TalkerDecode 1013.1 ms (64 frames, 15.83 ms/frame) +[Perf] CodePredictor 404.5 ms (6.32 ms/frame) [Perf] HostCompose 0.1 ms (c0 sample + next emb) -[Perf] CodecDecode 1015.1 ms -[Perf] Total 2835.6 ms (64 frames, 21.93 ms/frame AR, audio 5.12 s, RTF 0.554) +[Perf] CodecDecode 980.9 ms +[Perf] Total 3938.8 ms (64 frames, 22.15 ms/frame AR, audio 5.12 s, RTF 0.769) [WAV] Wrote cpp/clone/clone-cpp.wav: 122880 samples, 24000 Hz, mono S16 [Pipeline] Wrote 122880 samples (5.12 s) -> cpp/clone/clone-cpp.wav [Input] Prompt: 100 chars: qwentts.cpp is a minimal C++17 and GGML port of Qwen3-TTS fo... diff --git a/tests/clone-CPU-Q8_0.log b/tests/clone-CPU-Q8_0.log index a871f9c..baf597f 100644 --- a/tests/clone-CPU-Q8_0.log +++ b/tests/clone-CPU-Q8_0.log @@ -1,9 +1,9 @@ -[Qwen] qwentts.cpp bcac463 (2026-07-05) +[Qwen] qwentts.cpp 53c84fe (2026-07-20) ggml_cuda_init: found 1 CUDA devices (Total VRAM: 97247 MiB): Device 0: NVIDIA RTX PRO 6000 Blackwell Workstation Edition, compute capability 12.0, VMM: yes, VRAM: 97247 MiB load_backend: loaded CUDA backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cuda.so ggml_vulkan: Found 1 Vulkan devices: -ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 +ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | fp4: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 load_backend: loaded Vulkan backend from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cpu-zen4.so [Load] Talker backend: CPU (CPU threads: 16) @@ -23,9 +23,9 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [DAC] Loaded: 4 blocks (strides 8/5/4/3), 24 kHz mono out, weights 142.9 MB [WeightCtx] Loaded 2 tensors, 3.0 MB into backend [Pipeline] Ready: hop 1920 samples @ 24000 Hz mono, 16 codebooks @ 12.5 Hz -[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096 -> 896 MB -[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16 -> 0 MB -[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=off clamp_fp16=off +[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096, 1 sets -> 896 MB +[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16, 1 sets -> 0 MB +[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=off clamp_fp16=off max_batch=1 [BPE] Loaded from GGUF: 151676 vocab, 151291 merges, eos_id=151643 [BPE] Registered 5 arch special tokens (total specials=6) [WAV] Read buffer: 380672 samples, 22050 Hz, 1 ch, 16 bit @@ -33,7 +33,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Audio-Resample] Done: 380672 -> 414337 samples [WeightCtx] Loaded 76 tensors, 34.9 MB into backend [SpeakerEncoder] Loaded: enc_dim=2048 sr=24000 mel_dim=128 hidden=512 mfa=1536 asp_attn=128 se=128 scale=8 -[Pipeline] Speaker encoder lazy loaded in 3 ms +[Pipeline] Speaker encoder lazy loaded in 4 ms [Debug] mel-spk: [1618, 128] first4: -6.330339 -6.170074 -6.011195 -6.243161 [Debug] mel-hann: [1024] first4: 0.000000 0.000009 0.000038 0.000085 [Debug] mel-basis: [128, 513] first4: 0.000000 0.033550 0.008569 0.000000 @@ -51,7 +51,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [EncDownsample] Loaded: k=4 stride=2, 512 -> 512 channels, weights 2.0 MB [WeightCtx] Loaded 20 tensors, 34.0 MB into backend [EncQuantizer] Loaded: 16 codebooks (1 semantic + 15 acoustic), 2048 entries x 256 dim, hidden 512, weights 34.0 MB -[Pipeline] Codec encoder lazy loaded in 8 ms +[Pipeline] Codec encoder lazy loaded in 9 ms [Debug] audio-input: [412800] first4: -0.000546 -0.001162 -0.000871 0.000340 [Debug] seanet-out: [430, 512] first4: -0.120292 13.742489 -0.348104 3.009387 [Debug] enc-transformer-out: [430, 512] first4: 0.716872 2.959070 -0.221450 0.852651 @@ -76,60 +76,31 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Debug] talker-hidden-prefill-l27: [225, 2048] first4: 6.696979 -34.327644 -7.302014 28.627213 [Debug] talker-hidden-prefill-final: [225, 2048] first4: 0.227865 -1.219680 -0.230862 0.887843 [Debug] talker-logits-prefill: [3072] first4: 1.309391 -7.886597 -6.855894 -10.555799 +[Batch] Admit slot=0 T_ctx=225 prefill=545.3 ms build=2.6 ms (stall for 0 active slots) [Sample] step=0 c0=27 u=-1.0000000000 subseq=0 -[Sample-CP] g=0 c=1408 u=-1.0000000000 subseq=1 -[Sample-CP] g=1 c=1824 u=-1.0000000000 subseq=2 -[Sample-CP] g=2 c=400 u=-1.0000000000 subseq=3 -[Sample-CP] g=3 c=240 u=-1.0000000000 subseq=4 -[Sample-CP] g=4 c=2010 u=-1.0000000000 subseq=5 -[Sample-CP] g=5 c=196 u=-1.0000000000 subseq=6 -[Sample-CP] g=6 c=580 u=-1.0000000000 subseq=7 -[Sample-CP] g=7 c=1713 u=-1.0000000000 subseq=8 -[Sample-CP] g=8 c=1221 u=-1.0000000000 subseq=9 -[Sample-CP] g=9 c=1796 u=-1.0000000000 subseq=10 -[Sample-CP] g=10 c=1228 u=-1.0000000000 subseq=11 -[Sample-CP] g=11 c=2015 u=-1.0000000000 subseq=12 -[Sample-CP] g=12 c=143 u=-1.0000000000 subseq=13 -[Sample-CP] g=13 c=1062 u=-1.0000000000 subseq=14 -[Sample-CP] g=14 c=1094 u=-1.0000000000 subseq=15 [Debug] codes-step0: [16] first4: 27.000000 1408.000000 1824.000000 400.000000 [Debug] next-emb-step0: [2048] first4: 0.010612 -0.008921 0.074940 0.036301 [Debug] talker-hidden-step1: [2048] first4: 0.118400 2.726040 -0.843629 0.103909 [Sample] step=1 c0=1902 u=-1.0000000000 subseq=16 -[Sample-CP] g=0 c=1924 u=-1.0000000000 subseq=17 -[Sample-CP] g=1 c=2025 u=-1.0000000000 subseq=18 -[Sample-CP] g=2 c=324 u=-1.0000000000 subseq=19 -[Sample-CP] g=3 c=350 u=-1.0000000000 subseq=20 -[Sample-CP] g=4 c=985 u=-1.0000000000 subseq=21 -[Sample-CP] g=5 c=853 u=-1.0000000000 subseq=22 -[Sample-CP] g=6 c=329 u=-1.0000000000 subseq=23 -[Sample-CP] g=7 c=564 u=-1.0000000000 subseq=24 -[Sample-CP] g=8 c=692 u=-1.0000000000 subseq=25 -[Sample-CP] g=9 c=1702 u=-1.0000000000 subseq=26 -[Sample-CP] g=10 c=437 u=-1.0000000000 subseq=27 -[Sample-CP] g=11 c=729 u=-1.0000000000 subseq=28 -[Sample-CP] g=12 c=344 u=-1.0000000000 subseq=29 -[Sample-CP] g=13 c=85 u=-1.0000000000 subseq=30 -[Sample-CP] g=14 c=1668 u=-1.0000000000 subseq=31 -[Pipeline] Generated 8 frames -[Pipeline] Generated 16 frames -[Pipeline] Generated 24 frames -[Pipeline] Generated 32 frames -[Pipeline] Generated 40 frames -[Pipeline] Generated 48 frames -[Pipeline] Generated 56 frames -[Pipeline] Generated 64 frames +[Pipeline] Generated 8 frames (slot 0) +[Pipeline] Generated 16 frames (slot 0) +[Pipeline] Generated 24 frames (slot 0) +[Pipeline] Generated 32 frames (slot 0) +[Pipeline] Generated 40 frames (slot 0) +[Pipeline] Generated 48 frames (slot 0) +[Pipeline] Generated 56 frames (slot 0) +[Pipeline] Generated 64 frames (slot 0) [Pipeline] Generation done : 64 frames [Debug] codes-full: [64, 16] first4: 27.000000 1408.000000 1824.000000 400.000000 [Debug] output-audio: [122880] first4: 0.000527 0.000324 -0.000983 -0.000676 -[Perf] PromptBuild 2.9 ms -[Perf] Prefill 612.7 ms (T_ctx prefill) -[Perf] TTFA 628.6 ms (first frame codes) -[Perf] TalkerDecode 1690.1 ms (64 frames, 26.41 ms/frame) -[Perf] CodePredictor 658.2 ms (10.28 ms/frame) +[Perf] PromptBuild 2.6 ms +[Perf] Prefill 545.3 ms (T_ctx prefill) +[Perf] TTFA 1680.2 ms (first frame codes) +[Perf] TalkerDecode 1713.7 ms (64 frames, 26.78 ms/frame) +[Perf] CodePredictor 707.1 ms (11.05 ms/frame) [Perf] HostCompose 0.1 ms (c0 sample + next emb) -[Perf] CodecDecode 1021.4 ms -[Perf] Total 3996.0 ms (64 frames, 36.69 ms/frame AR, audio 5.12 s, RTF 0.780) +[Perf] CodecDecode 1025.4 ms +[Perf] Total 5114.5 ms (64 frames, 37.83 ms/frame AR, audio 5.12 s, RTF 0.999) [WAV] Wrote cpp/clone/clone-cpp.wav: 122880 samples, 24000 Hz, mono S16 [Pipeline] Wrote 122880 samples (5.12 s) -> cpp/clone/clone-cpp.wav [Input] Prompt: 100 chars: qwentts.cpp is a minimal C++17 and GGML port of Qwen3-TTS fo... @@ -154,14 +125,14 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Cossim] TrailingText cos: 0.999999 max: 1.9968e-04 mean: 4.5182e-05 [Cossim] TTSPadEmbed cos: 0.999999 max: 1.9968e-04 mean: 4.5182e-05 [Cossim] L0 cos: 0.992177 max: 7.7295e-01 mean: 7.3151e-03 -[Cossim] L7 cos: 0.999982 max: 5.4102e+00 mean: 2.6624e-02 -[Cossim] L14 cos: 0.999969 max: 5.4209e+00 mean: 5.2229e-02 -[Cossim] L21 cos: 0.999712 max: 1.7887e+01 mean: 1.9992e-01 +[Cossim] L7 cos: 0.999982 max: 5.4097e+00 mean: 2.6624e-02 +[Cossim] L14 cos: 0.999969 max: 5.4204e+00 mean: 5.2229e-02 +[Cossim] L21 cos: 0.999712 max: 1.7888e+01 mean: 1.9992e-01 [Cossim] L27 cos: 0.998655 max: 1.1809e+02 mean: 4.8277e-01 [Cossim] Final cos: 0.994339 max: 1.6804e+01 mean: 1.4894e-01 -[Cossim] Logits cos: 0.999855 max: 6.7472e-01 mean: 6.7599e-02 +[Cossim] Logits cos: 0.999855 max: 6.7471e-01 mean: 6.7599e-02 [Cossim] NextEmbStep0 cos: 0.829035 max: 1.5489e-01 mean: 2.9164e-02 -[Cossim] TalkerHiddenStep1 cos: 0.993414 max: 1.8271e+00 mean: 2.4874e-01 +[Cossim] TalkerHiddenStep1 cos: 0.993414 max: 1.8272e+00 mean: 2.4874e-01 [Cossim] MelHann cos: 1.000000 max: 1.7881e-07 mean: 3.2043e-08 [Cossim] MelBasis cos: 1.000000 max: 1.7881e-07 mean: 3.9379e-10 [Cossim] MelMag cos: 1.000000 max: 3.1307e-05 mean: 5.6396e-07 @@ -180,5 +151,5 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Cossim] SpkASP cos: 1.000000 max: 2.6596e-04 mean: 2.3921e-05 [Cossim] SpeakerEmb cos: 1.000000 max: 2.3174e-04 mean: 1.6094e-05 [Cossim] CodesFull exact: 2.68% (1008 values) -[Cossim] Audio cos: -0.012062 -[Cossim] WAV stft_cos: 0.175184 samples: 120960 +[Cossim] Audio cos: -0.002547 +[Cossim] WAV stft_cos: 0.174666 samples: 120960 diff --git a/tests/clone-CUDA0-BF16.log b/tests/clone-CUDA0-BF16.log index 53a4a9d..b98ae8a 100644 --- a/tests/clone-CUDA0-BF16.log +++ b/tests/clone-CUDA0-BF16.log @@ -1,9 +1,9 @@ -[Qwen] qwentts.cpp bcac463 (2026-07-05) +[Qwen] qwentts.cpp 53c84fe (2026-07-20) ggml_cuda_init: found 1 CUDA devices (Total VRAM: 97247 MiB): Device 0: NVIDIA RTX PRO 6000 Blackwell Workstation Edition, compute capability 12.0, VMM: yes, VRAM: 97247 MiB load_backend: loaded CUDA backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cuda.so ggml_vulkan: Found 1 Vulkan devices: -ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 +ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | fp4: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 load_backend: loaded Vulkan backend from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cpu-zen4.so [Load] Talker backend: CUDA0 (CPU threads: 16) @@ -23,9 +23,9 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [DAC] Loaded: 4 blocks (strides 8/5/4/3), 24 kHz mono out, weights 142.9 MB [WeightCtx] Loaded 2 tensors, 3.0 MB into backend [Pipeline] Ready: hop 1920 samples @ 24000 Hz mono, 16 codebooks @ 12.5 Hz -[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096 -> 896 MB -[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16 -> 0 MB -[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=on clamp_fp16=off +[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096, 1 sets -> 896 MB +[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16, 1 sets -> 0 MB +[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=on clamp_fp16=off max_batch=1 [BPE] Loaded from GGUF: 151676 vocab, 151291 merges, eos_id=151643 [BPE] Registered 5 arch special tokens (total specials=6) [WAV] Read buffer: 380672 samples, 22050 Hz, 1 ch, 16 bit @@ -33,15 +33,15 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Audio-Resample] Done: 380672 -> 414337 samples [WeightCtx] Loaded 76 tensors, 34.9 MB into backend [SpeakerEncoder] Loaded: enc_dim=2048 sr=24000 mel_dim=128 hidden=512 mfa=1536 asp_attn=128 se=128 scale=8 -[Pipeline] Speaker encoder lazy loaded in 4 ms +[Pipeline] Speaker encoder lazy loaded in 18 ms [Debug] mel-spk: [1618, 128] first4: -6.330339 -6.170074 -6.011195 -6.243161 [Debug] mel-hann: [1024] first4: 0.000000 0.000009 0.000038 0.000085 [Debug] mel-basis: [128, 513] first4: 0.000000 0.033550 0.008569 0.000000 [Debug] mel-mag: [1618, 513] first4: 0.047604 0.042322 0.042187 0.066630 -[Debug] spk-frontend: [1618, 512] first4: 0.000000 0.101562 0.858398 0.000000 -[Debug] spk-block3: [1618, 512] first4: 1.958879 0.101562 1.239357 0.246141 -[Debug] spk-mfa: [1618, 1536] first4: 0.000000 0.919922 0.609741 0.000000 -[Debug] spk-asp: [1, 3072] first4: 0.067580 0.118872 0.240696 0.344030 +[Debug] spk-frontend: [1618, 512] first4: 0.000000 0.097904 0.854992 0.000000 +[Debug] spk-block3: [1618, 512] first4: 1.955988 0.097904 1.233304 0.246811 +[Debug] spk-mfa: [1618, 1536] first4: 0.000000 0.922734 0.614915 0.000000 +[Debug] spk-asp: [1, 3072] first4: 0.067723 0.119136 0.240880 0.344137 [SpkExtract] Extracted 2048-dim embedding (414337 samples, padded 415105) [WeightCtx] Loaded 28 tensors, 24.1 MB into backend [SEANet] Loaded: 4 stages (ratios 4/5/6/8), 64 -> 512 channels, weights 24.1 MB @@ -51,11 +51,11 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [EncDownsample] Loaded: k=4 stride=2, 512 -> 512 channels, weights 2.0 MB [WeightCtx] Loaded 20 tensors, 34.0 MB into backend [EncQuantizer] Loaded: 16 codebooks (1 semantic + 15 acoustic), 2048 entries x 256 dim, hidden 512, weights 34.0 MB -[Pipeline] Codec encoder lazy loaded in 42 ms +[Pipeline] Codec encoder lazy loaded in 116 ms [Debug] audio-input: [412800] first4: -0.000546 -0.001162 -0.000871 0.000340 [Debug] seanet-out: [430, 512] first4: -0.111631 13.770183 -0.302467 2.984188 -[Debug] enc-transformer-out: [430, 512] first4: 0.531195 3.238480 -0.146394 0.612285 -[Debug] codec-pre-fsq: [215, 512] first4: -3.886719 -7.015625 -0.871094 2.914062 +[Debug] enc-transformer-out: [430, 512] first4: 0.536580 3.214924 -0.152684 0.613806 +[Debug] codec-pre-fsq: [215, 512] first4: -3.876953 -7.023438 -0.895508 2.943359 [Debug] seanet-init: [412800, 64] first4: -0.015825 -0.054180 -0.117431 0.038554 [Debug] seanet-resnet0: [412800, 64] first4: 0.057569 -0.220525 -0.137337 -0.251168 [Debug] seanet-stage0: [103200, 128] first4: 0.233175 -0.323887 0.247025 0.205248 @@ -67,87 +67,41 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Debug] talker-input-embed: [225, 2048] first4: 0.022162 -0.009036 0.008171 -0.020198 [Debug] trailing-text-hidden: [1, 2048] first4: -0.005340 0.008744 -0.004225 0.001144 [Debug] tts-pad-embed: [2048] first4: -0.005340 0.008744 -0.004225 0.001144 -[Debug] spk-emb: [2048] first4: 0.098059 -0.023872 -0.011379 0.006934 +[Debug] spk-emb: [2048] first4: 0.098317 -0.023761 -0.011860 0.006760 [Debug] ref-codes: [16, 215] first4: 1221.000000 2042.000000 509.000000 1522.000000 -[Debug] talker-hidden-prefill-l0: [225, 2048] first4: -0.461236 -0.069827 0.016106 -0.039241 -[Debug] talker-hidden-prefill-l7: [225, 2048] first4: -1.527145 -0.010904 0.330830 1.785726 -[Debug] talker-hidden-prefill-l14: [225, 2048] first4: -1.298392 0.186105 0.431904 1.752741 -[Debug] talker-hidden-prefill-l21: [225, 2048] first4: -0.771681 0.167680 0.780671 1.118511 -[Debug] talker-hidden-prefill-l27: [225, 2048] first4: 7.162132 -34.503532 -7.128798 28.739269 -[Debug] talker-hidden-prefill-final: [225, 2048] first4: 0.243477 -1.224848 -0.225186 0.890532 -[Debug] talker-logits-prefill: [3072] first4: 0.871094 -7.875000 -6.593750 -10.375000 +[Debug] talker-hidden-prefill-l0: [225, 2048] first4: -0.461261 -0.069988 0.015911 -0.039094 +[Debug] talker-hidden-prefill-l7: [225, 2048] first4: -1.523625 -0.015407 0.347368 1.776475 +[Debug] talker-hidden-prefill-l14: [225, 2048] first4: -1.295189 0.181396 0.448390 1.743353 +[Debug] talker-hidden-prefill-l21: [225, 2048] first4: -0.769886 0.163224 0.799067 1.112840 +[Debug] talker-hidden-prefill-l27: [225, 2048] first4: 7.045298 -34.398609 -7.069448 28.624956 +[Debug] talker-hidden-prefill-final: [225, 2048] first4: 0.239789 -1.222574 -0.223577 0.888043 +[Debug] talker-logits-prefill: [3072] first4: 0.944606 -7.870981 -6.684478 -10.343767 +[Batch] Admit slot=0 T_ctx=225 prefill=22.4 ms build=1.4 ms (stall for 0 active slots) [Sample] step=0 c0=27 u=-1.0000000000 subseq=0 -[Sample-CP] g=0 c=1408 u=-1.0000000000 subseq=1 -[Sample-CP] g=1 c=1824 u=-1.0000000000 subseq=2 -[Sample-CP] g=2 c=400 u=-1.0000000000 subseq=3 -[Sample-CP] g=3 c=240 u=-1.0000000000 subseq=4 -[Sample-CP] g=4 c=2010 u=-1.0000000000 subseq=5 -[Sample-CP] g=5 c=196 u=-1.0000000000 subseq=6 -[Sample-CP] g=6 c=580 u=-1.0000000000 subseq=7 -[Sample-CP] g=7 c=1713 u=-1.0000000000 subseq=8 -[Sample-CP] g=8 c=1221 u=-1.0000000000 subseq=9 -[Sample-CP] g=9 c=1796 u=-1.0000000000 subseq=10 -[Sample-CP] g=10 c=1228 u=-1.0000000000 subseq=11 -[Sample-CP] g=11 c=2015 u=-1.0000000000 subseq=12 -[Sample-CP] g=12 c=143 u=-1.0000000000 subseq=13 -[Sample-CP] g=13 c=1062 u=-1.0000000000 subseq=14 -[Sample-CP] g=14 c=1094 u=-1.0000000000 subseq=15 [Debug] codes-step0: [16] first4: 27.000000 1408.000000 1824.000000 400.000000 -[Debug] next-emb-step0: [2048] first4: 0.010331 -0.009322 0.074922 0.036956 -[Debug] talker-hidden-step1: [2048] first4: 0.017435 2.931466 -0.766813 0.054457 +[Debug] next-emb-step0: [2048] first4: 0.013383 -0.012534 0.088655 0.033408 +[Debug] talker-hidden-step1: [2048] first4: 0.132555 2.816141 -0.591259 -0.000489 [Sample] step=1 c0=1902 u=-1.0000000000 subseq=16 ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=0 c=1924 u=-1.0000000000 subseq=17 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=1 c=2025 u=-1.0000000000 subseq=18 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=2 c=324 u=-1.0000000000 subseq=19 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=3 c=350 u=-1.0000000000 subseq=20 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=4 c=985 u=-1.0000000000 subseq=21 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=5 c=853 u=-1.0000000000 subseq=22 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=6 c=329 u=-1.0000000000 subseq=23 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=7 c=564 u=-1.0000000000 subseq=24 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=8 c=692 u=-1.0000000000 subseq=25 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=9 c=1702 u=-1.0000000000 subseq=26 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=10 c=437 u=-1.0000000000 subseq=27 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=11 c=729 u=-1.0000000000 subseq=28 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=12 c=344 u=-1.0000000000 subseq=29 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=13 c=85 u=-1.0000000000 subseq=30 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=14 c=1668 u=-1.0000000000 subseq=31 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Pipeline] Generated 8 frames -[Pipeline] Generated 16 frames -[Pipeline] Generated 24 frames -[Pipeline] Generated 32 frames -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Pipeline] Generated 40 frames -[Pipeline] Generated 48 frames -[Pipeline] Generated 56 frames -[Pipeline] Generated 64 frames +[Pipeline] Generated 8 frames (slot 0) +[Pipeline] Generated 16 frames (slot 0) +[Pipeline] Generated 24 frames (slot 0) +[Pipeline] Generated 32 frames (slot 0) +[Pipeline] Generated 40 frames (slot 0) +[Pipeline] Generated 48 frames (slot 0) +[Pipeline] Generated 56 frames (slot 0) +[Pipeline] Generated 64 frames (slot 0) [Pipeline] Generation done : 64 frames [Debug] codes-full: [64, 16] first4: 27.000000 1408.000000 1824.000000 400.000000 -[Debug] output-audio: [122880] first4: 0.000465 0.000259 -0.001065 -0.000736 -[Perf] PromptBuild 1.3 ms -[Perf] Prefill 13.5 ms (T_ctx prefill) -[Perf] TTFA 23.3 ms (first frame codes) -[Perf] TalkerDecode 189.6 ms (64 frames, 2.96 ms/frame) -[Perf] CodePredictor 304.1 ms (4.75 ms/frame) +[Debug] output-audio: [122880] first4: 0.000498 0.000316 -0.001016 -0.000733 +[Perf] PromptBuild 1.4 ms +[Perf] Prefill 22.4 ms (T_ctx prefill) +[Perf] TTFA 913.7 ms (first frame codes) +[Perf] TalkerDecode 187.0 ms (64 frames, 2.92 ms/frame) +[Perf] CodePredictor 259.9 ms (4.06 ms/frame) [Perf] HostCompose 0.1 ms (c0 sample + next emb) -[Perf] CodecDecode 14.7 ms -[Perf] Total 524.3 ms (64 frames, 7.72 ms/frame AR, audio 5.12 s, RTF 0.102) +[Perf] CodecDecode 14.9 ms +[Perf] Total 1368.8 ms (64 frames, 6.98 ms/frame AR, audio 5.12 s, RTF 0.267) [WAV] Wrote cpp/clone/clone-cpp.wav: 122880 samples, 24000 Hz, mono S16 [Pipeline] Wrote 122880 samples (5.12 s) -> cpp/clone/clone-cpp.wav [Input] Prompt: 100 chars: qwentts.cpp is a minimal C++17 and GGML port of Qwen3-TTS fo... @@ -167,19 +121,19 @@ ggml_backend_cuda_graph_compute: CUDA graph warmup complete [GGML] Cmd: ../build/qwen-tts --model ../models/qwen-talker-1.7b-base-BF16.gguf --codec ../models/qwen-tokenizer-12hz-BF16.gguf --seed 42 --ref-wav ../examples/freeman.wav --ref-text ../examples/freeman.txt --lang english --max-new 64 --dump cpp/clone -o cpp/clone/clone-cpp.wav --greedy [GGML] Audio: 122880 samples 24000 Hz 5.12s -> cpp/clone/clone-cpp.wav [Cossim] PromptIDs exact: 100.00% (38 values) -[Cossim] RefCodes exact: 78.52% (3440 values) -[Cossim] Embed cos: 0.965578 max: 1.9438e-01 mean: 9.8620e-03 +[Cossim] RefCodes exact: 79.01% (3440 values) +[Cossim] Embed cos: 0.965841 max: 1.9438e-01 mean: 9.9500e-03 [Cossim] TrailingText cos: 1.000000 max: 3.5107e-05 mean: 2.5004e-06 [Cossim] TTSPadEmbed cos: 1.000000 max: 3.5107e-05 mean: 2.5004e-06 -[Cossim] L0 cos: 0.984329 max: 8.0274e-01 mean: 1.3008e-02 -[Cossim] L7 cos: 0.999981 max: 4.7525e+01 mean: 3.1978e-02 -[Cossim] L14 cos: 0.999956 max: 4.7533e+01 mean: 7.0164e-02 -[Cossim] L21 cos: 0.999567 max: 4.7568e+01 mean: 2.6764e-01 -[Cossim] L27 cos: 0.998445 max: 4.7995e+01 mean: 6.0528e-01 -[Cossim] Final cos: 0.991977 max: 1.5200e+01 mean: 1.9679e-01 -[Cossim] Logits cos: 0.999456 max: 1.0356e+00 mean: 1.3007e-01 -[Cossim] NextEmbStep0 cos: 0.829094 max: 1.5468e-01 mean: 2.9154e-02 -[Cossim] TalkerHiddenStep1 cos: 0.992002 max: 2.2514e+00 mean: 2.7397e-01 +[Cossim] L0 cos: 0.984359 max: 7.7032e-01 mean: 1.3103e-02 +[Cossim] L7 cos: 0.999990 max: 1.2875e+01 mean: 3.2110e-02 +[Cossim] L14 cos: 0.999965 max: 1.2878e+01 mean: 6.9459e-02 +[Cossim] L21 cos: 0.999584 max: 1.6917e+01 mean: 2.6241e-01 +[Cossim] L27 cos: 0.998499 max: 4.1020e+01 mean: 5.9180e-01 +[Cossim] Final cos: 0.992242 max: 1.4704e+01 mean: 1.9250e-01 +[Cossim] Logits cos: 0.999487 max: 9.8101e-01 mean: 1.2576e-01 +[Cossim] NextEmbStep0 cos: 0.832386 max: 1.5622e-01 mean: 2.8598e-02 +[Cossim] TalkerHiddenStep1 cos: 0.993512 max: 2.0032e+00 mean: 2.4247e-01 [Cossim] MelHann cos: 1.000000 max: 1.7881e-07 mean: 3.2043e-08 [Cossim] MelBasis cos: 1.000000 max: 1.7881e-07 mean: 3.9379e-10 [Cossim] MelMag cos: 1.000000 max: 3.1307e-05 mean: 5.6396e-07 @@ -190,13 +144,13 @@ ggml_backend_cuda_graph_compute: CUDA graph warmup complete [Cossim] SeanetStage1 cos: 0.999862 max: 7.2425e-01 mean: 5.3037e-03 [Cossim] SeanetStage3 cos: 0.999994 max: 2.1089e+00 mean: 4.7947e-02 [Cossim] SeanetOut cos: 0.999953 max: 3.2791e-01 mean: 2.0392e-02 -[Cossim] EncTransformer cos: 0.999956 max: 1.1787e+00 mean: 1.1794e-02 -[Cossim] CodecPreFSQ cos: 0.999960 max: 4.7962e-01 mean: 2.3320e-02 -[Cossim] SpkFrontend cos: 0.999995 max: 3.5096e-02 mean: 1.0924e-03 -[Cossim] SpkBlock3 cos: 0.999987 max: 1.3367e-01 mean: 4.9918e-03 -[Cossim] SpkMFA cos: 0.999973 max: 7.5789e-02 mean: 1.1867e-03 -[Cossim] SpkASP cos: 0.999999 max: 3.9836e-03 mean: 2.4063e-04 -[Cossim] SpeakerEmb cos: 1.000000 max: 1.7091e-03 mean: 2.1396e-04 -[Cossim] CodesFull exact: 5.85% (1008 values) -[Cossim] Audio cos: 0.030355 -[Cossim] WAV stft_cos: 0.297198 samples: 120960 +[Cossim] EncTransformer cos: 0.999962 max: 7.5433e-01 mean: 1.0610e-02 +[Cossim] CodecPreFSQ cos: 0.999962 max: 4.7767e-01 mean: 2.2384e-02 +[Cossim] SpkFrontend cos: 1.000000 max: 2.1511e-04 mean: 1.4495e-06 +[Cossim] SpkBlock3 cos: 1.000000 max: 3.2926e-04 mean: 4.3974e-06 +[Cossim] SpkMFA cos: 1.000000 max: 3.8069e-04 mean: 1.3195e-06 +[Cossim] SpkASP cos: 1.000000 max: 7.9870e-06 mean: 2.2269e-07 +[Cossim] SpeakerEmb cos: 1.000000 max: 1.4305e-05 mean: 2.5463e-07 +[Cossim] CodesFull exact: 2.08% (1008 values) +[Cossim] Audio cos: 0.022055 +[Cossim] WAV stft_cos: 0.185286 samples: 120960 diff --git a/tests/clone-CUDA0-F32.log b/tests/clone-CUDA0-F32.log index 13b8a0b..c755797 100644 --- a/tests/clone-CUDA0-F32.log +++ b/tests/clone-CUDA0-F32.log @@ -1,9 +1,9 @@ -[Qwen] qwentts.cpp bcac463 (2026-07-05) +[Qwen] qwentts.cpp 53c84fe (2026-07-20) ggml_cuda_init: found 1 CUDA devices (Total VRAM: 97247 MiB): Device 0: NVIDIA RTX PRO 6000 Blackwell Workstation Edition, compute capability 12.0, VMM: yes, VRAM: 97247 MiB load_backend: loaded CUDA backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cuda.so ggml_vulkan: Found 1 Vulkan devices: -ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 +ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | fp4: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 load_backend: loaded Vulkan backend from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cpu-zen4.so [Load] Talker backend: CUDA0 (CPU threads: 16) @@ -23,9 +23,9 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [DAC] Loaded: 4 blocks (strides 8/5/4/3), 24 kHz mono out, weights 142.9 MB [WeightCtx] Loaded 2 tensors, 3.0 MB into backend [Pipeline] Ready: hop 1920 samples @ 24000 Hz mono, 16 codebooks @ 12.5 Hz -[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096 -> 896 MB -[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16 -> 0 MB -[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=on clamp_fp16=off +[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096, 1 sets -> 896 MB +[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16, 1 sets -> 0 MB +[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=on clamp_fp16=off max_batch=1 [BPE] Loaded from GGUF: 151676 vocab, 151291 merges, eos_id=151643 [BPE] Registered 5 arch special tokens (total specials=6) [WAV] Read buffer: 380672 samples, 22050 Hz, 1 ch, 16 bit @@ -33,7 +33,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Audio-Resample] Done: 380672 -> 414337 samples [WeightCtx] Loaded 76 tensors, 45.8 MB into backend [SpeakerEncoder] Loaded: enc_dim=2048 sr=24000 mel_dim=128 hidden=512 mfa=1536 asp_attn=128 se=128 scale=8 -[Pipeline] Speaker encoder lazy loaded in 4 ms +[Pipeline] Speaker encoder lazy loaded in 17 ms [Debug] mel-spk: [1618, 128] first4: -6.330339 -6.170074 -6.011195 -6.243161 [Debug] mel-hann: [1024] first4: 0.000000 0.000009 0.000038 0.000085 [Debug] mel-basis: [128, 513] first4: 0.000000 0.033550 0.008569 0.000000 @@ -51,7 +51,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [EncDownsample] Loaded: k=4 stride=2, 512 -> 512 channels, weights 2.0 MB [WeightCtx] Loaded 20 tensors, 34.0 MB into backend [EncQuantizer] Loaded: 16 codebooks (1 semantic + 15 acoustic), 2048 entries x 256 dim, hidden 512, weights 34.0 MB -[Pipeline] Codec encoder lazy loaded in 31 ms +[Pipeline] Codec encoder lazy loaded in 90 ms [Debug] audio-input: [412800] first4: -0.000546 -0.001162 -0.000871 0.000340 [Debug] seanet-out: [430, 512] first4: -0.134336 13.731120 -0.351051 3.013485 [Debug] enc-transformer-out: [430, 512] first4: 0.525800 3.116711 -0.187038 0.631811 @@ -76,78 +76,32 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Debug] talker-hidden-prefill-l27: [225, 2048] first4: 7.131224 -34.471714 -7.108000 28.674042 [Debug] talker-hidden-prefill-final: [225, 2048] first4: 0.242833 -1.225771 -0.224906 0.890001 [Debug] talker-logits-prefill: [3072] first4: 1.171040 -7.796190 -6.822366 -10.286754 +[Batch] Admit slot=0 T_ctx=225 prefill=45.0 ms build=47.6 ms (stall for 0 active slots) [Sample] step=0 c0=27 u=-1.0000000000 subseq=0 -[Sample-CP] g=0 c=1408 u=-1.0000000000 subseq=1 -[Sample-CP] g=1 c=1824 u=-1.0000000000 subseq=2 -[Sample-CP] g=2 c=400 u=-1.0000000000 subseq=3 -[Sample-CP] g=3 c=240 u=-1.0000000000 subseq=4 -[Sample-CP] g=4 c=2010 u=-1.0000000000 subseq=5 -[Sample-CP] g=5 c=196 u=-1.0000000000 subseq=6 -[Sample-CP] g=6 c=1984 u=-1.0000000000 subseq=7 -[Sample-CP] g=7 c=816 u=-1.0000000000 subseq=8 -[Sample-CP] g=8 c=684 u=-1.0000000000 subseq=9 -[Sample-CP] g=9 c=1702 u=-1.0000000000 subseq=10 -[Sample-CP] g=10 c=273 u=-1.0000000000 subseq=11 -[Sample-CP] g=11 c=798 u=-1.0000000000 subseq=12 -[Sample-CP] g=12 c=593 u=-1.0000000000 subseq=13 -[Sample-CP] g=13 c=85 u=-1.0000000000 subseq=14 -[Sample-CP] g=14 c=541 u=-1.0000000000 subseq=15 [Debug] codes-step0: [16] first4: 27.000000 1408.000000 1824.000000 400.000000 [Debug] next-emb-step0: [2048] first4: 0.003672 0.000908 0.039506 0.094594 [Debug] talker-hidden-step1: [2048] first4: 0.754625 2.407892 -0.056593 0.158147 [Sample] step=1 c0=1902 u=-1.0000000000 subseq=16 ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=0 c=1924 u=-1.0000000000 subseq=17 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=1 c=2025 u=-1.0000000000 subseq=18 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=2 c=324 u=-1.0000000000 subseq=19 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=3 c=1501 u=-1.0000000000 subseq=20 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=4 c=1252 u=-1.0000000000 subseq=21 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=5 c=1831 u=-1.0000000000 subseq=22 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=6 c=1544 u=-1.0000000000 subseq=23 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=7 c=1058 u=-1.0000000000 subseq=24 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=8 c=1673 u=-1.0000000000 subseq=25 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=9 c=541 u=-1.0000000000 subseq=26 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=10 c=1162 u=-1.0000000000 subseq=27 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=11 c=729 u=-1.0000000000 subseq=28 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=12 c=781 u=-1.0000000000 subseq=29 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=13 c=85 u=-1.0000000000 subseq=30 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=14 c=1435 u=-1.0000000000 subseq=31 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Pipeline] Generated 8 frames -[Pipeline] Generated 16 frames -[Pipeline] Generated 24 frames -[Pipeline] Generated 32 frames -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Pipeline] Generated 40 frames -[Pipeline] Generated 48 frames -[Pipeline] Generated 56 frames -[Pipeline] Generated 64 frames +[Pipeline] Generated 8 frames (slot 0) +[Pipeline] Generated 16 frames (slot 0) +[Pipeline] Generated 24 frames (slot 0) +[Pipeline] Generated 32 frames (slot 0) +[Pipeline] Generated 40 frames (slot 0) +[Pipeline] Generated 48 frames (slot 0) +[Pipeline] Generated 56 frames (slot 0) +[Pipeline] Generated 64 frames (slot 0) [Pipeline] Generation done : 64 frames [Debug] codes-full: [64, 16] first4: 27.000000 1408.000000 1824.000000 400.000000 [Debug] output-audio: [122880] first4: 0.000739 0.000593 -0.000747 -0.000435 -[Perf] PromptBuild 1.3 ms -[Perf] Prefill 27.4 ms (T_ctx prefill) -[Perf] TTFA 37.4 ms (first frame codes) -[Perf] TalkerDecode 304.7 ms (64 frames, 4.76 ms/frame) -[Perf] CodePredictor 389.8 ms (6.09 ms/frame) +[Perf] PromptBuild 47.6 ms +[Perf] Prefill 45.0 ms (T_ctx prefill) +[Perf] TTFA 1027.4 ms (first frame codes) +[Perf] TalkerDecode 608.3 ms (64 frames, 9.50 ms/frame) +[Perf] CodePredictor 1654.1 ms (25.84 ms/frame) [Perf] HostCompose 0.1 ms (c0 sample + next emb) -[Perf] CodecDecode 14.8 ms -[Perf] Total 739.3 ms (64 frames, 10.85 ms/frame AR, audio 5.12 s, RTF 0.144) +[Perf] CodecDecode 24.8 ms +[Perf] Total 3261.0 ms (64 frames, 35.35 ms/frame AR, audio 5.12 s, RTF 0.637) [WAV] Wrote cpp/clone/clone-cpp.wav: 122880 samples, 24000 Hz, mono S16 [Pipeline] Wrote 122880 samples (5.12 s) -> cpp/clone/clone-cpp.wav [Input] Prompt: 100 chars: qwentts.cpp is a minimal C++17 and GGML port of Qwen3-TTS fo... @@ -169,15 +123,15 @@ ggml_backend_cuda_graph_compute: CUDA graph warmup complete [Cossim] PromptIDs exact: 100.00% (38 values) [Cossim] RefCodes exact: 82.41% (3440 values) [Cossim] Embed cos: 0.972420 max: 1.7531e-01 mean: 8.3100e-03 -[Cossim] TrailingText cos: 1.000000 max: 2.9802e-08 mean: 5.1686e-10 -[Cossim] TTSPadEmbed cos: 1.000000 max: 2.9802e-08 mean: 5.1686e-10 +[Cossim] TrailingText cos: 1.000000 max: 2.9802e-08 mean: 6.8263e-10 +[Cossim] TTSPadEmbed cos: 1.000000 max: 2.9802e-08 mean: 6.8263e-10 [Cossim] L0 cos: 0.986912 max: 6.6445e-01 mean: 1.1103e-02 [Cossim] L7 cos: 0.999992 max: 1.2570e+00 mean: 2.7510e-02 [Cossim] L14 cos: 0.999973 max: 2.1782e+00 mean: 5.9937e-02 -[Cossim] L21 cos: 0.999719 max: 8.3606e+00 mean: 2.2268e-01 -[Cossim] L27 cos: 0.998971 max: 2.3989e+01 mean: 5.1174e-01 -[Cossim] Final cos: 0.994704 max: 7.3140e+00 mean: 1.6649e-01 -[Cossim] Logits cos: 0.999743 max: 5.7110e-01 mean: 8.8929e-02 +[Cossim] L21 cos: 0.999719 max: 8.3600e+00 mean: 2.2268e-01 +[Cossim] L27 cos: 0.998971 max: 2.3987e+01 mean: 5.1173e-01 +[Cossim] Final cos: 0.994704 max: 7.3135e+00 mean: 1.6649e-01 +[Cossim] Logits cos: 0.999743 max: 5.7110e-01 mean: 8.8930e-02 [Cossim] NextEmbStep0 cos: 0.829216 max: 1.2523e-01 mean: 2.9129e-02 [Cossim] TalkerHiddenStep1 cos: 0.993854 max: 1.4329e+00 mean: 2.4202e-01 [Cossim] MelHann cos: 1.000000 max: 1.7881e-07 mean: 3.2043e-08 diff --git a/tests/clone-CUDA0-Q4_K_M.log b/tests/clone-CUDA0-Q4_K_M.log index aeadde3..be4444b 100644 --- a/tests/clone-CUDA0-Q4_K_M.log +++ b/tests/clone-CUDA0-Q4_K_M.log @@ -1,9 +1,9 @@ -[Qwen] qwentts.cpp bcac463 (2026-07-05) +[Qwen] qwentts.cpp 53c84fe (2026-07-20) ggml_cuda_init: found 1 CUDA devices (Total VRAM: 97247 MiB): Device 0: NVIDIA RTX PRO 6000 Blackwell Workstation Edition, compute capability 12.0, VMM: yes, VRAM: 97247 MiB load_backend: loaded CUDA backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cuda.so ggml_vulkan: Found 1 Vulkan devices: -ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 +ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | fp4: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 load_backend: loaded Vulkan backend from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cpu-zen4.so [Load] Talker backend: CUDA0 (CPU threads: 16) @@ -23,9 +23,9 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [DAC] Loaded: 4 blocks (strides 8/5/4/3), 24 kHz mono out, weights 142.9 MB [WeightCtx] Loaded 2 tensors, 3.0 MB into backend [Pipeline] Ready: hop 1920 samples @ 24000 Hz mono, 16 codebooks @ 12.5 Hz -[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096 -> 896 MB -[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16 -> 0 MB -[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=on clamp_fp16=off +[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096, 1 sets -> 896 MB +[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16, 1 sets -> 0 MB +[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=on clamp_fp16=off max_batch=1 [BPE] Loaded from GGUF: 151676 vocab, 151291 merges, eos_id=151643 [BPE] Registered 5 arch special tokens (total specials=6) [WAV] Read buffer: 380672 samples, 22050 Hz, 1 ch, 16 bit @@ -33,7 +33,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Audio-Resample] Done: 380672 -> 414337 samples [WeightCtx] Loaded 76 tensors, 34.9 MB into backend [SpeakerEncoder] Loaded: enc_dim=2048 sr=24000 mel_dim=128 hidden=512 mfa=1536 asp_attn=128 se=128 scale=8 -[Pipeline] Speaker encoder lazy loaded in 4 ms +[Pipeline] Speaker encoder lazy loaded in 27 ms [Debug] mel-spk: [1618, 128] first4: -6.330339 -6.170074 -6.011195 -6.243161 [Debug] mel-hann: [1024] first4: 0.000000 0.000009 0.000038 0.000085 [Debug] mel-basis: [128, 513] first4: 0.000000 0.033550 0.008569 0.000000 @@ -51,7 +51,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [EncDownsample] Loaded: k=4 stride=2, 512 -> 512 channels, weights 2.0 MB [WeightCtx] Loaded 20 tensors, 34.0 MB into backend [EncQuantizer] Loaded: 16 codebooks (1 semantic + 15 acoustic), 2048 entries x 256 dim, hidden 512, weights 34.0 MB -[Pipeline] Codec encoder lazy loaded in 8 ms +[Pipeline] Codec encoder lazy loaded in 43 ms [Debug] audio-input: [412800] first4: -0.000546 -0.001162 -0.000871 0.000340 [Debug] seanet-out: [430, 512] first4: -0.134336 13.731120 -0.351051 3.013485 [Debug] enc-transformer-out: [430, 512] first4: 1.140990 3.713432 0.131578 0.616251 @@ -76,120 +76,32 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Debug] talker-hidden-prefill-l27: [225, 2048] first4: 4.146309 -31.912390 -6.918371 22.418171 [Debug] talker-hidden-prefill-final: [225, 2048] first4: 0.140643 -1.130366 -0.218058 0.693131 [Debug] talker-logits-prefill: [3072] first4: -0.461190 -5.760653 -7.184410 -10.423358 +[Batch] Admit slot=0 T_ctx=225 prefill=24.0 ms build=6.9 ms (stall for 0 active slots) [Sample] step=0 c0=27 u=-1.0000000000 subseq=0 -[Sample-CP] g=0 c=1042 u=-1.0000000000 subseq=1 -[Sample-CP] g=1 c=390 u=-1.0000000000 subseq=2 -[Sample-CP] g=2 c=881 u=-1.0000000000 subseq=3 -[Sample-CP] g=3 c=405 u=-1.0000000000 subseq=4 -[Sample-CP] g=4 c=1347 u=-1.0000000000 subseq=5 -[Sample-CP] g=5 c=1640 u=-1.0000000000 subseq=6 -[Sample-CP] g=6 c=470 u=-1.0000000000 subseq=7 -[Sample-CP] g=7 c=488 u=-1.0000000000 subseq=8 -[Sample-CP] g=8 c=153 u=-1.0000000000 subseq=9 -[Sample-CP] g=9 c=1492 u=-1.0000000000 subseq=10 -[Sample-CP] g=10 c=1029 u=-1.0000000000 subseq=11 -[Sample-CP] g=11 c=863 u=-1.0000000000 subseq=12 -[Sample-CP] g=12 c=111 u=-1.0000000000 subseq=13 -[Sample-CP] g=13 c=262 u=-1.0000000000 subseq=14 -[Sample-CP] g=14 c=735 u=-1.0000000000 subseq=15 [Debug] codes-step0: [16] first4: 27.000000 1042.000000 390.000000 881.000000 [Debug] next-emb-step0: [2048] first4: -0.033968 -0.020822 -0.037756 -0.013151 [Debug] talker-hidden-step1: [2048] first4: 1.225613 1.784957 -1.039321 1.449351 [Sample] step=1 c0=99 u=-1.0000000000 subseq=16 ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=0 c=43 u=-1.0000000000 subseq=17 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=1 c=1859 u=-1.0000000000 subseq=18 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=2 c=22 u=-1.0000000000 subseq=19 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=3 c=239 u=-1.0000000000 subseq=20 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=4 c=915 u=-1.0000000000 subseq=21 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=5 c=2019 u=-1.0000000000 subseq=22 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=6 c=329 u=-1.0000000000 subseq=23 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=7 c=241 u=-1.0000000000 subseq=24 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=8 c=1004 u=-1.0000000000 subseq=25 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=9 c=826 u=-1.0000000000 subseq=26 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=10 c=1603 u=-1.0000000000 subseq=27 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=11 c=1160 u=-1.0000000000 subseq=28 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=12 c=1759 u=-1.0000000000 subseq=29 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=13 c=591 u=-1.0000000000 subseq=30 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=14 c=901 u=-1.0000000000 subseq=31 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Pipeline] Generated 8 frames -[Pipeline] Generated 16 frames -[Pipeline] Generated 24 frames -[Pipeline] Generated 32 frames -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Pipeline] Generated 40 frames -[Pipeline] Generated 48 frames -[Pipeline] Generated 56 frames -[Pipeline] Generated 64 frames +[Pipeline] Generated 8 frames (slot 0) +[Pipeline] Generated 16 frames (slot 0) +[Pipeline] Generated 24 frames (slot 0) +[Pipeline] Generated 32 frames (slot 0) +[Pipeline] Generated 40 frames (slot 0) +[Pipeline] Generated 48 frames (slot 0) +[Pipeline] Generated 56 frames (slot 0) +[Pipeline] Generated 64 frames (slot 0) [Pipeline] Generation done : 64 frames [Debug] codes-full: [64, 16] first4: 27.000000 1042.000000 390.000000 881.000000 [Debug] output-audio: [122880] first4: 0.000087 0.000212 -0.000383 -0.000346 -[Perf] PromptBuild 56.4 ms -[Perf] Prefill 14.6 ms (T_ctx prefill) -[Perf] TTFA 79.2 ms (first frame codes) -[Perf] TalkerDecode 224.9 ms (64 frames, 3.51 ms/frame) -[Perf] CodePredictor 338.0 ms (5.28 ms/frame) +[Perf] PromptBuild 6.9 ms +[Perf] Prefill 24.0 ms (T_ctx prefill) +[Perf] TTFA 844.4 ms (first frame codes) +[Perf] TalkerDecode 191.5 ms (64 frames, 2.99 ms/frame) +[Perf] CodePredictor 735.8 ms (11.50 ms/frame) [Perf] HostCompose 0.1 ms (c0 sample + next emb) -[Perf] CodecDecode 23.8 ms -[Perf] Total 658.8 ms (64 frames, 8.80 ms/frame AR, audio 5.12 s, RTF 0.129) +[Perf] CodecDecode 19.7 ms +[Perf] Total 1766.6 ms (64 frames, 14.49 ms/frame AR, audio 5.12 s, RTF 0.345) [WAV] Wrote cpp/clone/clone-cpp.wav: 122880 samples, 24000 Hz, mono S16 [Pipeline] Wrote 122880 samples (5.12 s) -> cpp/clone/clone-cpp.wav [Input] Prompt: 100 chars: qwentts.cpp is a minimal C++17 and GGML port of Qwen3-TTS fo... diff --git a/tests/clone-CUDA0-Q8_0.log b/tests/clone-CUDA0-Q8_0.log index 0ccb822..6932f94 100644 --- a/tests/clone-CUDA0-Q8_0.log +++ b/tests/clone-CUDA0-Q8_0.log @@ -1,9 +1,9 @@ -[Qwen] qwentts.cpp bcac463 (2026-07-05) +[Qwen] qwentts.cpp 53c84fe (2026-07-20) ggml_cuda_init: found 1 CUDA devices (Total VRAM: 97247 MiB): Device 0: NVIDIA RTX PRO 6000 Blackwell Workstation Edition, compute capability 12.0, VMM: yes, VRAM: 97247 MiB load_backend: loaded CUDA backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cuda.so ggml_vulkan: Found 1 Vulkan devices: -ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 +ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | fp4: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 load_backend: loaded Vulkan backend from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cpu-zen4.so [Load] Talker backend: CUDA0 (CPU threads: 16) @@ -23,9 +23,9 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [DAC] Loaded: 4 blocks (strides 8/5/4/3), 24 kHz mono out, weights 142.9 MB [WeightCtx] Loaded 2 tensors, 3.0 MB into backend [Pipeline] Ready: hop 1920 samples @ 24000 Hz mono, 16 codebooks @ 12.5 Hz -[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096 -> 896 MB -[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16 -> 0 MB -[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=on clamp_fp16=off +[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096, 1 sets -> 896 MB +[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16, 1 sets -> 0 MB +[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=on clamp_fp16=off max_batch=1 [BPE] Loaded from GGUF: 151676 vocab, 151291 merges, eos_id=151643 [BPE] Registered 5 arch special tokens (total specials=6) [WAV] Read buffer: 380672 samples, 22050 Hz, 1 ch, 16 bit @@ -33,7 +33,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Audio-Resample] Done: 380672 -> 414337 samples [WeightCtx] Loaded 76 tensors, 34.9 MB into backend [SpeakerEncoder] Loaded: enc_dim=2048 sr=24000 mel_dim=128 hidden=512 mfa=1536 asp_attn=128 se=128 scale=8 -[Pipeline] Speaker encoder lazy loaded in 3 ms +[Pipeline] Speaker encoder lazy loaded in 26 ms [Debug] mel-spk: [1618, 128] first4: -6.330339 -6.170074 -6.011195 -6.243161 [Debug] mel-hann: [1024] first4: 0.000000 0.000009 0.000038 0.000085 [Debug] mel-basis: [128, 513] first4: 0.000000 0.033550 0.008569 0.000000 @@ -51,7 +51,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [EncDownsample] Loaded: k=4 stride=2, 512 -> 512 channels, weights 2.0 MB [WeightCtx] Loaded 20 tensors, 34.0 MB into backend [EncQuantizer] Loaded: 16 codebooks (1 semantic + 15 acoustic), 2048 entries x 256 dim, hidden 512, weights 34.0 MB -[Pipeline] Codec encoder lazy loaded in 8 ms +[Pipeline] Codec encoder lazy loaded in 46 ms [Debug] audio-input: [412800] first4: -0.000546 -0.001162 -0.000871 0.000340 [Debug] seanet-out: [430, 512] first4: -0.134336 13.731120 -0.351051 3.013485 [Debug] enc-transformer-out: [430, 512] first4: 0.628775 2.909739 -0.209228 0.726215 @@ -76,78 +76,32 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Debug] talker-hidden-prefill-l27: [225, 2048] first4: 6.677233 -34.349281 -7.298450 28.605591 [Debug] talker-hidden-prefill-final: [225, 2048] first4: 0.227255 -1.220783 -0.230812 0.887415 [Debug] talker-logits-prefill: [3072] first4: 1.706040 -8.254342 -7.318320 -10.816376 +[Batch] Admit slot=0 T_ctx=225 prefill=15.5 ms build=1.6 ms (stall for 0 active slots) [Sample] step=0 c0=27 u=-1.0000000000 subseq=0 -[Sample-CP] g=0 c=1408 u=-1.0000000000 subseq=1 -[Sample-CP] g=1 c=1824 u=-1.0000000000 subseq=2 -[Sample-CP] g=2 c=400 u=-1.0000000000 subseq=3 -[Sample-CP] g=3 c=550 u=-1.0000000000 subseq=4 -[Sample-CP] g=4 c=1506 u=-1.0000000000 subseq=5 -[Sample-CP] g=5 c=970 u=-1.0000000000 subseq=6 -[Sample-CP] g=6 c=470 u=-1.0000000000 subseq=7 -[Sample-CP] g=7 c=1713 u=-1.0000000000 subseq=8 -[Sample-CP] g=8 c=1221 u=-1.0000000000 subseq=9 -[Sample-CP] g=9 c=645 u=-1.0000000000 subseq=10 -[Sample-CP] g=10 c=180 u=-1.0000000000 subseq=11 -[Sample-CP] g=11 c=1713 u=-1.0000000000 subseq=12 -[Sample-CP] g=12 c=217 u=-1.0000000000 subseq=13 -[Sample-CP] g=13 c=561 u=-1.0000000000 subseq=14 -[Sample-CP] g=14 c=1681 u=-1.0000000000 subseq=15 [Debug] codes-step0: [16] first4: 27.000000 1408.000000 1824.000000 400.000000 [Debug] next-emb-step0: [2048] first4: -0.034936 -0.024575 0.029866 0.045185 [Debug] talker-hidden-step1: [2048] first4: 0.931840 2.357023 0.427709 -0.264416 [Sample] step=1 c0=1902 u=-1.0000000000 subseq=16 ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=0 c=1924 u=-1.0000000000 subseq=17 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=1 c=2025 u=-1.0000000000 subseq=18 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=2 c=324 u=-1.0000000000 subseq=19 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=3 c=1501 u=-1.0000000000 subseq=20 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=4 c=407 u=-1.0000000000 subseq=21 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=5 c=1452 u=-1.0000000000 subseq=22 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=6 c=1544 u=-1.0000000000 subseq=23 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=7 c=1058 u=-1.0000000000 subseq=24 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=8 c=692 u=-1.0000000000 subseq=25 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=9 c=1702 u=-1.0000000000 subseq=26 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=10 c=362 u=-1.0000000000 subseq=27 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=11 c=729 u=-1.0000000000 subseq=28 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=12 c=781 u=-1.0000000000 subseq=29 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=13 c=82 u=-1.0000000000 subseq=30 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Sample-CP] g=14 c=1435 u=-1.0000000000 subseq=31 -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Pipeline] Generated 8 frames -[Pipeline] Generated 16 frames -[Pipeline] Generated 24 frames -[Pipeline] Generated 32 frames -ggml_backend_cuda_graph_compute: CUDA graph warmup reset -ggml_backend_cuda_graph_compute: CUDA graph warmup complete -[Pipeline] Generated 40 frames -[Pipeline] Generated 48 frames -[Pipeline] Generated 56 frames -[Pipeline] Generated 64 frames +[Pipeline] Generated 8 frames (slot 0) +[Pipeline] Generated 16 frames (slot 0) +[Pipeline] Generated 24 frames (slot 0) +[Pipeline] Generated 32 frames (slot 0) +[Pipeline] Generated 40 frames (slot 0) +[Pipeline] Generated 48 frames (slot 0) +[Pipeline] Generated 56 frames (slot 0) +[Pipeline] Generated 64 frames (slot 0) [Pipeline] Generation done : 64 frames [Debug] codes-full: [64, 16] first4: 27.000000 1408.000000 1824.000000 400.000000 [Debug] output-audio: [122880] first4: 0.000596 0.000259 -0.001224 -0.000817 -[Perf] PromptBuild 1.5 ms -[Perf] Prefill 15.2 ms (T_ctx prefill) -[Perf] TTFA 24.0 ms (first frame codes) -[Perf] TalkerDecode 152.3 ms (64 frames, 2.38 ms/frame) -[Perf] CodePredictor 246.1 ms (3.85 ms/frame) +[Perf] PromptBuild 1.6 ms +[Perf] Prefill 15.5 ms (T_ctx prefill) +[Perf] TTFA 784.9 ms (first frame codes) +[Perf] TalkerDecode 126.9 ms (64 frames, 1.98 ms/frame) +[Perf] CodePredictor 196.0 ms (3.06 ms/frame) [Perf] HostCompose 0.1 ms (c0 sample + next emb) -[Perf] CodecDecode 15.0 ms -[Perf] Total 431.5 ms (64 frames, 6.23 ms/frame AR, audio 5.12 s, RTF 0.084) +[Perf] CodecDecode 15.2 ms +[Perf] Total 1117.9 ms (64 frames, 5.05 ms/frame AR, audio 5.12 s, RTF 0.218) [WAV] Wrote cpp/clone/clone-cpp.wav: 122880 samples, 24000 Hz, mono S16 [Pipeline] Wrote 122880 samples (5.12 s) -> cpp/clone/clone-cpp.wav [Input] Prompt: 100 chars: qwentts.cpp is a minimal C++17 and GGML port of Qwen3-TTS fo... @@ -172,9 +126,9 @@ ggml_backend_cuda_graph_compute: CUDA graph warmup complete [Cossim] TrailingText cos: 0.999999 max: 1.9968e-04 mean: 4.5182e-05 [Cossim] TTSPadEmbed cos: 0.999999 max: 1.9968e-04 mean: 4.5182e-05 [Cossim] L0 cos: 0.984964 max: 7.6776e-01 mean: 1.2815e-02 -[Cossim] L7 cos: 0.999978 max: 3.6929e+00 mean: 3.8281e-02 -[Cossim] L14 cos: 0.999955 max: 3.7129e+00 mean: 7.4563e-02 -[Cossim] L21 cos: 0.999568 max: 1.7895e+01 mean: 2.7422e-01 +[Cossim] L7 cos: 0.999978 max: 3.6924e+00 mean: 3.8281e-02 +[Cossim] L14 cos: 0.999955 max: 3.7124e+00 mean: 7.4563e-02 +[Cossim] L21 cos: 0.999568 max: 1.7896e+01 mean: 2.7422e-01 [Cossim] L27 cos: 0.998158 max: 9.9299e+01 mean: 6.4856e-01 [Cossim] Final cos: 0.991511 max: 1.5600e+01 mean: 2.0381e-01 [Cossim] Logits cos: 0.998602 max: 1.4375e+00 mean: 2.2099e-01 diff --git a/tests/clone-Vulkan0-BF16.log b/tests/clone-Vulkan0-BF16.log index 8a6011a..2085bf2 100644 --- a/tests/clone-Vulkan0-BF16.log +++ b/tests/clone-Vulkan0-BF16.log @@ -1,9 +1,9 @@ -[Qwen] qwentts.cpp bcac463 (2026-07-05) +[Qwen] qwentts.cpp 53c84fe (2026-07-20) ggml_cuda_init: found 1 CUDA devices (Total VRAM: 97247 MiB): Device 0: NVIDIA RTX PRO 6000 Blackwell Workstation Edition, compute capability 12.0, VMM: yes, VRAM: 97247 MiB load_backend: loaded CUDA backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cuda.so ggml_vulkan: Found 1 Vulkan devices: -ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 +ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | fp4: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 load_backend: loaded Vulkan backend from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cpu-zen4.so [Load] Talker backend: Vulkan0 (CPU threads: 16) @@ -23,9 +23,9 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [DAC] Loaded: 4 blocks (strides 8/5/4/3), 24 kHz mono out, weights 142.9 MB [WeightCtx] Loaded 2 tensors, 3.0 MB into backend [Pipeline] Ready: hop 1920 samples @ 24000 Hz mono, 16 codebooks @ 12.5 Hz -[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096 -> 896 MB -[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16 -> 0 MB -[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=on clamp_fp16=off +[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096, 1 sets -> 896 MB +[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16, 1 sets -> 0 MB +[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=on clamp_fp16=off max_batch=1 [BPE] Loaded from GGUF: 151676 vocab, 151291 merges, eos_id=151643 [BPE] Registered 5 arch special tokens (total specials=6) [WAV] Read buffer: 380672 samples, 22050 Hz, 1 ch, 16 bit @@ -51,7 +51,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [EncDownsample] Loaded: k=4 stride=2, 512 -> 512 channels, weights 2.0 MB [WeightCtx] Loaded 20 tensors, 34.0 MB into backend [EncQuantizer] Loaded: 16 codebooks (1 semantic + 15 acoustic), 2048 entries x 256 dim, hidden 512, weights 34.0 MB -[Pipeline] Codec encoder lazy loaded in 45 ms +[Pipeline] Codec encoder lazy loaded in 46 ms [Debug] audio-input: [412800] first4: -0.000546 -0.001162 -0.000871 0.000340 [Debug] seanet-out: [430, 512] first4: -0.110593 13.765666 -0.317055 2.981106 [Debug] enc-transformer-out: [430, 512] first4: 0.536260 3.195749 -0.156796 0.610111 @@ -76,62 +76,51 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Debug] talker-hidden-prefill-l27: [225, 2048] first4: 7.051195 -34.399265 -7.070650 28.631254 [Debug] talker-hidden-prefill-final: [225, 2048] first4: 0.239991 -1.222603 -0.223616 0.888243 [Debug] talker-logits-prefill: [3072] first4: 0.882721 -7.881639 -6.796915 -10.414125 +[Batch] Admit slot=0 T_ctx=225 prefill=32.2 ms build=2.1 ms (stall for 0 active slots) [Sample] step=0 c0=27 u=-1.0000000000 subseq=0 -[Sample-CP] g=0 c=1408 u=-1.0000000000 subseq=1 -[Sample-CP] g=1 c=1824 u=-1.0000000000 subseq=2 -[Sample-CP] g=2 c=400 u=-1.0000000000 subseq=3 -[Sample-CP] g=3 c=240 u=-1.0000000000 subseq=4 -[Sample-CP] g=4 c=2010 u=-1.0000000000 subseq=5 -[Sample-CP] g=5 c=196 u=-1.0000000000 subseq=6 -[Sample-CP] g=6 c=1984 u=-1.0000000000 subseq=7 -[Sample-CP] g=7 c=816 u=-1.0000000000 subseq=8 -[Sample-CP] g=8 c=1625 u=-1.0000000000 subseq=9 -[Sample-CP] g=9 c=288 u=-1.0000000000 subseq=10 -[Sample-CP] g=10 c=277 u=-1.0000000000 subseq=11 -[Sample-CP] g=11 c=577 u=-1.0000000000 subseq=12 -[Sample-CP] g=12 c=429 u=-1.0000000000 subseq=13 -[Sample-CP] g=13 c=127 u=-1.0000000000 subseq=14 -[Sample-CP] g=14 c=348 u=-1.0000000000 subseq=15 [Debug] codes-step0: [16] first4: 27.000000 1408.000000 1824.000000 400.000000 [Debug] next-emb-step0: [2048] first4: -0.008599 -0.015697 0.093591 0.086640 -[Debug] talker-hidden-step1: [2048] first4: 0.466706 2.304709 -0.154099 0.228768 -[Sample] step=1 c0=1902 u=-1.0000000000 subseq=16 -[Sample-CP] g=0 c=1924 u=-1.0000000000 subseq=17 -[Sample-CP] g=1 c=2025 u=-1.0000000000 subseq=18 -[Sample-CP] g=2 c=324 u=-1.0000000000 subseq=19 -[Sample-CP] g=3 c=1210 u=-1.0000000000 subseq=20 -[Sample-CP] g=4 c=985 u=-1.0000000000 subseq=21 -[Sample-CP] g=5 c=506 u=-1.0000000000 subseq=22 -[Sample-CP] g=6 c=329 u=-1.0000000000 subseq=23 -[Sample-CP] g=7 c=564 u=-1.0000000000 subseq=24 -[Sample-CP] g=8 c=1004 u=-1.0000000000 subseq=25 -[Sample-CP] g=9 c=1702 u=-1.0000000000 subseq=26 -[Sample-CP] g=10 c=437 u=-1.0000000000 subseq=27 -[Sample-CP] g=11 c=729 u=-1.0000000000 subseq=28 -[Sample-CP] g=12 c=1484 u=-1.0000000000 subseq=29 -[Sample-CP] g=13 c=85 u=-1.0000000000 subseq=30 -[Sample-CP] g=14 c=1978 u=-1.0000000000 subseq=31 -[Pipeline] Generated 8 frames -[Pipeline] Generated 16 frames -[Pipeline] Generated 24 frames -[Pipeline] Generated 32 frames -[Pipeline] Generated 40 frames -[Pipeline] Generated 48 frames -[Pipeline] Generated 56 frames -[Pipeline] Generated 64 frames -[Pipeline] Generation done : 64 frames -[Debug] codes-full: [64, 16] first4: 27.000000 1408.000000 1824.000000 400.000000 -[Debug] output-audio: [122880] first4: 0.000515 0.000357 -0.000915 -0.000575 -[Perf] PromptBuild 2.0 ms -[Perf] Prefill 31.7 ms (T_ctx prefill) -[Perf] TTFA 41.7 ms (first frame codes) -[Perf] TalkerDecode 248.0 ms (64 frames, 3.88 ms/frame) -[Perf] CodePredictor 415.5 ms (6.49 ms/frame) -[Perf] HostCompose 0.2 ms (c0 sample + next emb) -[Perf] CodecDecode 30.8 ms -[Perf] Total 729.2 ms (64 frames, 10.37 ms/frame AR, audio 5.12 s, RTF 0.142) -[WAV] Wrote cpp/clone/clone-cpp.wav: 122880 samples, 24000 Hz, mono S16 -[Pipeline] Wrote 122880 samples (5.12 s) -> cpp/clone/clone-cpp.wav +/mnt/workspace/git/qwentts.cpp/ggml/src/ggml-vulkan/ggml-vulkan.cpp:11497: GGML_ASSERT(dst->op != GGML_OP_GET_ROWS || (a_offset == 0 && b_offset == 0 && d_offset == 0)) failed +[New LWP 121432] +[New LWP 121431] +[New LWP 121430] +[New LWP 121429] +[New LWP 121428] +[New LWP 121427] +[New LWP 121426] +[New LWP 121425] +[New LWP 121424] +[New LWP 121423] +[New LWP 121422] +[New LWP 121421] +[New LWP 121420] +[New LWP 121419] +[New LWP 121418] +[New LWP 121417] +[New LWP 121416] +[New LWP 121415] +[New LWP 121414] +[New LWP 121413] +[Thread debugging using libthread_db enabled] +Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1". +0x00007f5e576799ee in ?? () from /lib/x86_64-linux-gnu/libc.so.6 +#0 0x00007f5e576799ee in ?? () from /lib/x86_64-linux-gnu/libc.so.6 +#1 0x00007f5e5766e668 in ?? () from /lib/x86_64-linux-gnu/libc.so.6 +#2 0x00007f5e5766e6ad in ?? () from /lib/x86_64-linux-gnu/libc.so.6 +#3 0x00007f5e576d97c7 in wait4 () from /lib/x86_64-linux-gnu/libc.so.6 +#4 0x00007f5e57b89a93 in ggml_print_backtrace () from /mnt/workspace/git/qwentts.cpp/build/libggml-base.so.0 +#5 0x00007f5e57b89bfe in ggml_abort () from /mnt/workspace/git/qwentts.cpp/build/libggml-base.so.0 +#6 0x00007f5c0e2d4bf1 in void init_pushconst_tensor_offsets(ggml_backend_vk_context*, vk_op_binary_push_constants&, ggml_tensor const*, ggml_tensor const*, ggml_tensor const*, ggml_tensor const*, ggml_tensor*) () from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so +#7 0x00007f5c0e449201 in void ggml_vk_op_f32(ggml_backend_vk_context*, std::shared_ptr&, ggml_tensor const*, ggml_tensor const*, ggml_tensor const*, ggml_tensor const*, ggml_tensor*, ggml_op, vk_op_binary_push_constants&&) [clone .constprop.0] () from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so +#8 0x00007f5c0e464f48 in ggml_vk_build_graph(ggml_backend_vk_context*, ggml_cgraph*, int, ggml_tensor*, int, bool, bool, bool) [clone .isra.0] () from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so +#9 0x00007f5c0e4679bc in ggml_backend_vk_graph_compute(ggml_backend*, ggml_cgraph*) () from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so +#10 0x00007f5e57ba196e in ggml_backend_graph_compute () from /mnt/workspace/git/qwentts.cpp/build/libggml-base.so.0 +#11 0x000055bbb3e94365 in tts_engine_step(TtsEngine*, std::vector >*) () +#12 0x000055bbb3ea166a in pipeline_tts_synthesize(PipelineTTS*, BPETokenizer*, qt_tts_params const*, long, qt_audio*) () +#13 0x000055bbb3e7db80 in qt_synthesize () +#14 0x000055bbb3e74cdd in run(Args const&) () +#15 0x000055bbb3e72a83 in main () +[Inferior 1 (process 121412) detached] [Input] Prompt: 100 chars: qwentts.cpp is a minimal C++17 and GGML port of Qwen3-TTS fo... [Input] RefAudio: ../examples/freeman.wav [Input] RefText: 213 chars: If you go into different cultures, they have different conce... @@ -147,38 +136,3 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Python] Audio: 120960 samples 24000 Hz 5.04s -> python/clone/clone-python.wav [Quant] BF16 -> ../models/qwen-talker-1.7b-base-BF16.gguf + ../models/qwen-tokenizer-12hz-BF16.gguf [GGML] Cmd: ../build/qwen-tts --model ../models/qwen-talker-1.7b-base-BF16.gguf --codec ../models/qwen-tokenizer-12hz-BF16.gguf --seed 42 --ref-wav ../examples/freeman.wav --ref-text ../examples/freeman.txt --lang english --max-new 64 --dump cpp/clone -o cpp/clone/clone-cpp.wav --greedy -[GGML] Audio: 122880 samples 24000 Hz 5.12s -> cpp/clone/clone-cpp.wav -[Cossim] PromptIDs exact: 100.00% (38 values) -[Cossim] RefCodes exact: 76.83% (3440 values) -[Cossim] Embed cos: 0.962709 max: 1.9437e-01 mean: 1.0689e-02 -[Cossim] TrailingText cos: 1.000000 max: 2.9802e-08 mean: 5.4933e-10 -[Cossim] TTSPadEmbed cos: 1.000000 max: 2.9802e-08 mean: 5.4933e-10 -[Cossim] L0 cos: 0.983221 max: 7.7030e-01 mean: 1.3992e-02 -[Cossim] L7 cos: 0.999989 max: 1.2885e+01 mean: 3.3235e-02 -[Cossim] L14 cos: 0.999964 max: 1.2887e+01 mean: 7.1843e-02 -[Cossim] L21 cos: 0.999531 max: 1.6955e+01 mean: 2.8087e-01 -[Cossim] L27 cos: 0.998209 max: 4.0700e+01 mean: 6.4207e-01 -[Cossim] Final cos: 0.990658 max: 1.5838e+01 mean: 2.0917e-01 -[Cossim] Logits cos: 0.999301 max: 1.1222e+00 mean: 1.3438e-01 -[Cossim] NextEmbStep0 cos: 0.818255 max: 1.4225e-01 mean: 3.0455e-02 -[Cossim] TalkerHiddenStep1 cos: 0.993880 max: 1.7297e+00 mean: 2.3463e-01 -[Cossim] MelHann cos: 1.000000 max: 1.7881e-07 mean: 3.2043e-08 -[Cossim] MelBasis cos: 1.000000 max: 1.7881e-07 mean: 3.9379e-10 -[Cossim] MelMag cos: 1.000000 max: 3.1307e-05 mean: 5.6396e-07 -[Cossim] MelSpk cos: 1.000000 max: 2.5492e-03 mean: 9.3754e-06 -[Cossim] SeanetInit cos: 1.000000 max: 9.1669e-04 mean: 1.6221e-05 -[Cossim] SeanetResnet0 cos: 0.999961 max: 2.1108e-03 mean: 1.6910e-04 -[Cossim] SeanetStage0 cos: 0.999983 max: 4.4128e-02 mean: 7.5091e-04 -[Cossim] SeanetStage1 cos: 0.999862 max: 7.2962e-01 mean: 5.2996e-03 -[Cossim] SeanetStage3 cos: 0.999994 max: 2.1304e+00 mean: 4.7641e-02 -[Cossim] SeanetOut cos: 0.999953 max: 3.4988e-01 mean: 2.0273e-02 -[Cossim] EncTransformer cos: 0.999963 max: 8.5211e-01 mean: 1.0424e-02 -[Cossim] CodecPreFSQ cos: 0.999962 max: 4.6668e-01 mean: 2.2088e-02 -[Cossim] SpkFrontend cos: 0.999997 max: 2.5180e-02 mean: 9.4975e-04 -[Cossim] SpkBlock3 cos: 0.999991 max: 1.0191e-01 mean: 4.1519e-03 -[Cossim] SpkMFA cos: 0.999979 max: 7.0797e-02 mean: 1.0369e-03 -[Cossim] SpkASP cos: 1.000000 max: 2.3533e-03 mean: 1.4947e-04 -[Cossim] SpeakerEmb cos: 1.000000 max: 6.7258e-04 mean: 1.2809e-04 -[Cossim] CodesFull exact: 2.28% (1008 values) -[Cossim] Audio cos: 0.003079 -[Cossim] WAV stft_cos: 0.208123 samples: 120960 diff --git a/tests/clone-Vulkan0-F32.log b/tests/clone-Vulkan0-F32.log index d4056b3..214cab8 100644 --- a/tests/clone-Vulkan0-F32.log +++ b/tests/clone-Vulkan0-F32.log @@ -1,9 +1,9 @@ -[Qwen] qwentts.cpp bcac463 (2026-07-05) +[Qwen] qwentts.cpp 53c84fe (2026-07-20) ggml_cuda_init: found 1 CUDA devices (Total VRAM: 97247 MiB): Device 0: NVIDIA RTX PRO 6000 Blackwell Workstation Edition, compute capability 12.0, VMM: yes, VRAM: 97247 MiB load_backend: loaded CUDA backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cuda.so ggml_vulkan: Found 1 Vulkan devices: -ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 +ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | fp4: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 load_backend: loaded Vulkan backend from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cpu-zen4.so [Load] Talker backend: Vulkan0 (CPU threads: 16) @@ -23,9 +23,9 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [DAC] Loaded: 4 blocks (strides 8/5/4/3), 24 kHz mono out, weights 142.9 MB [WeightCtx] Loaded 2 tensors, 3.0 MB into backend [Pipeline] Ready: hop 1920 samples @ 24000 Hz mono, 16 codebooks @ 12.5 Hz -[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096 -> 896 MB -[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16 -> 0 MB -[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=on clamp_fp16=off +[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096, 1 sets -> 896 MB +[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16, 1 sets -> 0 MB +[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=on clamp_fp16=off max_batch=1 [BPE] Loaded from GGUF: 151676 vocab, 151291 merges, eos_id=151643 [BPE] Registered 5 arch special tokens (total specials=6) [WAV] Read buffer: 380672 samples, 22050 Hz, 1 ch, 16 bit @@ -51,7 +51,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [EncDownsample] Loaded: k=4 stride=2, 512 -> 512 channels, weights 2.0 MB [WeightCtx] Loaded 20 tensors, 34.0 MB into backend [EncQuantizer] Loaded: 16 codebooks (1 semantic + 15 acoustic), 2048 entries x 256 dim, hidden 512, weights 34.0 MB -[Pipeline] Codec encoder lazy loaded in 35 ms +[Pipeline] Codec encoder lazy loaded in 36 ms [Debug] audio-input: [412800] first4: -0.000546 -0.001162 -0.000871 0.000340 [Debug] seanet-out: [430, 512] first4: -0.138303 13.699992 -0.351357 3.025051 [Debug] enc-transformer-out: [430, 512] first4: 0.526569 3.116264 -0.195330 0.625080 @@ -76,62 +76,51 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Debug] talker-hidden-prefill-l27: [225, 2048] first4: 7.115197 -34.442551 -7.104288 28.617508 [Debug] talker-hidden-prefill-final: [225, 2048] first4: 0.242297 -1.224786 -0.224798 0.888284 [Debug] talker-logits-prefill: [3072] first4: 1.013672 -7.734375 -6.857910 -10.250000 +[Batch] Admit slot=0 T_ctx=225 prefill=1426.7 ms build=21.4 ms (stall for 0 active slots) [Sample] step=0 c0=27 u=-1.0000000000 subseq=0 -[Sample-CP] g=0 c=1408 u=-1.0000000000 subseq=1 -[Sample-CP] g=1 c=1824 u=-1.0000000000 subseq=2 -[Sample-CP] g=2 c=400 u=-1.0000000000 subseq=3 -[Sample-CP] g=3 c=550 u=-1.0000000000 subseq=4 -[Sample-CP] g=4 c=1506 u=-1.0000000000 subseq=5 -[Sample-CP] g=5 c=915 u=-1.0000000000 subseq=6 -[Sample-CP] g=6 c=470 u=-1.0000000000 subseq=7 -[Sample-CP] g=7 c=1713 u=-1.0000000000 subseq=8 -[Sample-CP] g=8 c=1221 u=-1.0000000000 subseq=9 -[Sample-CP] g=9 c=223 u=-1.0000000000 subseq=10 -[Sample-CP] g=10 c=1502 u=-1.0000000000 subseq=11 -[Sample-CP] g=11 c=1702 u=-1.0000000000 subseq=12 -[Sample-CP] g=12 c=1860 u=-1.0000000000 subseq=13 -[Sample-CP] g=13 c=1008 u=-1.0000000000 subseq=14 -[Sample-CP] g=14 c=127 u=-1.0000000000 subseq=15 [Debug] codes-step0: [16] first4: 27.000000 1408.000000 1824.000000 400.000000 [Debug] next-emb-step0: [2048] first4: 0.034006 -0.013787 0.039109 0.091488 -[Debug] talker-hidden-step1: [2048] first4: 0.742126 2.656314 -0.143770 -0.182433 -[Sample] step=1 c0=1902 u=-1.0000000000 subseq=16 -[Sample-CP] g=0 c=1924 u=-1.0000000000 subseq=17 -[Sample-CP] g=1 c=2025 u=-1.0000000000 subseq=18 -[Sample-CP] g=2 c=324 u=-1.0000000000 subseq=19 -[Sample-CP] g=3 c=350 u=-1.0000000000 subseq=20 -[Sample-CP] g=4 c=985 u=-1.0000000000 subseq=21 -[Sample-CP] g=5 c=853 u=-1.0000000000 subseq=22 -[Sample-CP] g=6 c=329 u=-1.0000000000 subseq=23 -[Sample-CP] g=7 c=564 u=-1.0000000000 subseq=24 -[Sample-CP] g=8 c=692 u=-1.0000000000 subseq=25 -[Sample-CP] g=9 c=1702 u=-1.0000000000 subseq=26 -[Sample-CP] g=10 c=437 u=-1.0000000000 subseq=27 -[Sample-CP] g=11 c=729 u=-1.0000000000 subseq=28 -[Sample-CP] g=12 c=344 u=-1.0000000000 subseq=29 -[Sample-CP] g=13 c=85 u=-1.0000000000 subseq=30 -[Sample-CP] g=14 c=1668 u=-1.0000000000 subseq=31 -[Pipeline] Generated 8 frames -[Pipeline] Generated 16 frames -[Pipeline] Generated 24 frames -[Pipeline] Generated 32 frames -[Pipeline] Generated 40 frames -[Pipeline] Generated 48 frames -[Pipeline] Generated 56 frames -[Pipeline] Generated 64 frames -[Pipeline] Generation done : 64 frames -[Debug] codes-full: [64, 16] first4: 27.000000 1408.000000 1824.000000 400.000000 -[Debug] output-audio: [122880] first4: 0.000673 0.000478 -0.000828 -0.000482 -[Perf] PromptBuild 9.1 ms -[Perf] Prefill 19.6 ms (T_ctx prefill) -[Perf] TTFA 37.9 ms (first frame codes) -[Perf] TalkerDecode 369.2 ms (64 frames, 5.77 ms/frame) -[Perf] CodePredictor 512.4 ms (8.01 ms/frame) -[Perf] HostCompose 0.1 ms (c0 sample + next emb) -[Perf] CodecDecode 28.9 ms -[Perf] Total 940.4 ms (64 frames, 13.78 ms/frame AR, audio 5.12 s, RTF 0.184) -[WAV] Wrote cpp/clone/clone-cpp.wav: 122880 samples, 24000 Hz, mono S16 -[Pipeline] Wrote 122880 samples (5.12 s) -> cpp/clone/clone-cpp.wav +/mnt/workspace/git/qwentts.cpp/ggml/src/ggml-vulkan/ggml-vulkan.cpp:11497: GGML_ASSERT(dst->op != GGML_OP_GET_ROWS || (a_offset == 0 && b_offset == 0 && d_offset == 0)) failed +[New LWP 121280] +[New LWP 121279] +[New LWP 121278] +[New LWP 121277] +[New LWP 121276] +[New LWP 121275] +[New LWP 121274] +[New LWP 121273] +[New LWP 121272] +[New LWP 121271] +[New LWP 121270] +[New LWP 121269] +[New LWP 121268] +[New LWP 121267] +[New LWP 121266] +[New LWP 121265] +[New LWP 121264] +[New LWP 121263] +[New LWP 121262] +[New LWP 121261] +[Thread debugging using libthread_db enabled] +Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1". +0x00007f5ef94a69ee in ?? () from /lib/x86_64-linux-gnu/libc.so.6 +#0 0x00007f5ef94a69ee in ?? () from /lib/x86_64-linux-gnu/libc.so.6 +#1 0x00007f5ef949b668 in ?? () from /lib/x86_64-linux-gnu/libc.so.6 +#2 0x00007f5ef949b6ad in ?? () from /lib/x86_64-linux-gnu/libc.so.6 +#3 0x00007f5ef95067c7 in wait4 () from /lib/x86_64-linux-gnu/libc.so.6 +#4 0x00007f5ef9a60a93 in ggml_print_backtrace () from /mnt/workspace/git/qwentts.cpp/build/libggml-base.so.0 +#5 0x00007f5ef9a60bfe in ggml_abort () from /mnt/workspace/git/qwentts.cpp/build/libggml-base.so.0 +#6 0x00007f5cb08d4bf1 in void init_pushconst_tensor_offsets(ggml_backend_vk_context*, vk_op_binary_push_constants&, ggml_tensor const*, ggml_tensor const*, ggml_tensor const*, ggml_tensor const*, ggml_tensor*) () from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so +#7 0x00007f5cb0a49201 in void ggml_vk_op_f32(ggml_backend_vk_context*, std::shared_ptr&, ggml_tensor const*, ggml_tensor const*, ggml_tensor const*, ggml_tensor const*, ggml_tensor*, ggml_op, vk_op_binary_push_constants&&) [clone .constprop.0] () from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so +#8 0x00007f5cb0a64f48 in ggml_vk_build_graph(ggml_backend_vk_context*, ggml_cgraph*, int, ggml_tensor*, int, bool, bool, bool) [clone .isra.0] () from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so +#9 0x00007f5cb0a679bc in ggml_backend_vk_graph_compute(ggml_backend*, ggml_cgraph*) () from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so +#10 0x00007f5ef9a7896e in ggml_backend_graph_compute () from /mnt/workspace/git/qwentts.cpp/build/libggml-base.so.0 +#11 0x000055af16f11365 in tts_engine_step(TtsEngine*, std::vector >*) () +#12 0x000055af16f1e66a in pipeline_tts_synthesize(PipelineTTS*, BPETokenizer*, qt_tts_params const*, long, qt_audio*) () +#13 0x000055af16efab80 in qt_synthesize () +#14 0x000055af16ef1cdd in run(Args const&) () +#15 0x000055af16eefa83 in main () +[Inferior 1 (process 121260) detached] [Input] Prompt: 100 chars: qwentts.cpp is a minimal C++17 and GGML port of Qwen3-TTS fo... [Input] RefAudio: ../examples/freeman.wav [Input] RefText: 213 chars: If you go into different cultures, they have different conce... @@ -147,38 +136,3 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Python] Audio: 120960 samples 24000 Hz 5.04s -> python/clone/clone-python.wav [Quant] F32 -> ../models/qwen-talker-1.7b-base-F32.gguf + ../models/qwen-tokenizer-12hz-F32.gguf [GGML] Cmd: ../build/qwen-tts --model ../models/qwen-talker-1.7b-base-F32.gguf --codec ../models/qwen-tokenizer-12hz-F32.gguf --seed 42 --ref-wav ../examples/freeman.wav --ref-text ../examples/freeman.txt --lang english --max-new 64 --dump cpp/clone -o cpp/clone/clone-cpp.wav --greedy -[GGML] Audio: 122880 samples 24000 Hz 5.12s -> cpp/clone/clone-cpp.wav -[Cossim] PromptIDs exact: 100.00% (38 values) -[Cossim] RefCodes exact: 82.41% (3440 values) -[Cossim] Embed cos: 0.972129 max: 1.9482e-01 mean: 8.1111e-03 -[Cossim] TrailingText cos: 1.000000 max: 2.9802e-08 mean: 5.4933e-10 -[Cossim] TTSPadEmbed cos: 1.000000 max: 2.9802e-08 mean: 5.4933e-10 -[Cossim] L0 cos: 0.987066 max: 6.8534e-01 mean: 1.0819e-02 -[Cossim] L7 cos: 0.999990 max: 1.4636e+01 mean: 2.7206e-02 -[Cossim] L14 cos: 0.999971 max: 1.4633e+01 mean: 5.9840e-02 -[Cossim] L21 cos: 0.999707 max: 1.4641e+01 mean: 2.2449e-01 -[Cossim] L27 cos: 0.998906 max: 2.5746e+01 mean: 5.2221e-01 -[Cossim] Final cos: 0.994366 max: 7.7972e+00 mean: 1.6928e-01 -[Cossim] Logits cos: 0.999568 max: 8.3676e-01 mean: 1.1911e-01 -[Cossim] NextEmbStep0 cos: 0.822712 max: 2.7881e-01 mean: 3.0097e-02 -[Cossim] TalkerHiddenStep1 cos: 0.990742 max: 1.7658e+00 mean: 2.9620e-01 -[Cossim] MelHann cos: 1.000000 max: 1.7881e-07 mean: 3.2043e-08 -[Cossim] MelBasis cos: 1.000000 max: 1.7881e-07 mean: 3.9379e-10 -[Cossim] MelMag cos: 1.000000 max: 3.1307e-05 mean: 5.6396e-07 -[Cossim] MelSpk cos: 1.000000 max: 2.5492e-03 mean: 9.3754e-06 -[Cossim] SeanetInit cos: 1.000000 max: 4.1014e-04 mean: 3.4146e-06 -[Cossim] SeanetResnet0 cos: 0.999987 max: 1.9526e-03 mean: 9.4296e-05 -[Cossim] SeanetStage0 cos: 0.999985 max: 4.2238e-02 mean: 6.5089e-04 -[Cossim] SeanetStage1 cos: 0.999866 max: 7.7454e-01 mean: 4.9543e-03 -[Cossim] SeanetStage3 cos: 0.999995 max: 1.4657e+00 mean: 3.8722e-02 -[Cossim] SeanetOut cos: 0.999963 max: 4.0791e-01 mean: 1.6984e-02 -[Cossim] EncTransformer cos: 0.999974 max: 3.3284e-01 mean: 7.6420e-03 -[Cossim] CodecPreFSQ cos: 0.999972 max: 4.5137e-01 mean: 1.8013e-02 -[Cossim] SpkFrontend cos: 1.000000 max: 2.6600e-03 mean: 1.1964e-04 -[Cossim] SpkBlock3 cos: 1.000000 max: 1.2138e-02 mean: 5.2076e-04 -[Cossim] SpkMFA cos: 1.000000 max: 8.9911e-03 mean: 1.3077e-04 -[Cossim] SpkASP cos: 1.000000 max: 2.2851e-04 mean: 1.7482e-05 -[Cossim] SpeakerEmb cos: 1.000000 max: 7.8417e-05 mean: 1.3521e-05 -[Cossim] CodesFull exact: 3.17% (1008 values) -[Cossim] Audio cos: 0.023761 -[Cossim] WAV stft_cos: 0.225096 samples: 120960 diff --git a/tests/clone-Vulkan0-Q4_K_M.log b/tests/clone-Vulkan0-Q4_K_M.log index fad7e5a..5719a8a 100644 --- a/tests/clone-Vulkan0-Q4_K_M.log +++ b/tests/clone-Vulkan0-Q4_K_M.log @@ -1,9 +1,9 @@ -[Qwen] qwentts.cpp bcac463 (2026-07-05) +[Qwen] qwentts.cpp 53c84fe (2026-07-20) ggml_cuda_init: found 1 CUDA devices (Total VRAM: 97247 MiB): Device 0: NVIDIA RTX PRO 6000 Blackwell Workstation Edition, compute capability 12.0, VMM: yes, VRAM: 97247 MiB load_backend: loaded CUDA backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cuda.so ggml_vulkan: Found 1 Vulkan devices: -ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 +ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | fp4: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 load_backend: loaded Vulkan backend from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cpu-zen4.so [Load] Talker backend: Vulkan0 (CPU threads: 16) @@ -23,9 +23,9 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [DAC] Loaded: 4 blocks (strides 8/5/4/3), 24 kHz mono out, weights 142.9 MB [WeightCtx] Loaded 2 tensors, 3.0 MB into backend [Pipeline] Ready: hop 1920 samples @ 24000 Hz mono, 16 codebooks @ 12.5 Hz -[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096 -> 896 MB -[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16 -> 0 MB -[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=on clamp_fp16=off +[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096, 1 sets -> 896 MB +[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16, 1 sets -> 0 MB +[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=on clamp_fp16=off max_batch=1 [BPE] Loaded from GGUF: 151676 vocab, 151291 merges, eos_id=151643 [BPE] Registered 5 arch special tokens (total specials=6) [WAV] Read buffer: 380672 samples, 22050 Hz, 1 ch, 16 bit @@ -33,7 +33,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Audio-Resample] Done: 380672 -> 414337 samples [WeightCtx] Loaded 76 tensors, 34.9 MB into backend [SpeakerEncoder] Loaded: enc_dim=2048 sr=24000 mel_dim=128 hidden=512 mfa=1536 asp_attn=128 se=128 scale=8 -[Pipeline] Speaker encoder lazy loaded in 4 ms +[Pipeline] Speaker encoder lazy loaded in 5 ms [Debug] mel-spk: [1618, 128] first4: -6.330339 -6.170074 -6.011195 -6.243161 [Debug] mel-hann: [1024] first4: 0.000000 0.000009 0.000038 0.000085 [Debug] mel-basis: [128, 513] first4: 0.000000 0.033550 0.008569 0.000000 @@ -76,62 +76,51 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Debug] talker-hidden-prefill-l27: [225, 2048] first4: 4.921398 -31.754539 -6.813190 21.707518 [Debug] talker-hidden-prefill-final: [225, 2048] first4: 0.166135 -1.119393 -0.213715 0.667947 [Debug] talker-logits-prefill: [3072] first4: -0.151123 -5.516602 -7.045898 -10.537109 +[Batch] Admit slot=0 T_ctx=225 prefill=1887.5 ms build=1227.7 ms (stall for 0 active slots) [Sample] step=0 c0=27 u=-1.0000000000 subseq=0 -[Sample-CP] g=0 c=1408 u=-1.0000000000 subseq=1 -[Sample-CP] g=1 c=1824 u=-1.0000000000 subseq=2 -[Sample-CP] g=2 c=417 u=-1.0000000000 subseq=3 -[Sample-CP] g=3 c=442 u=-1.0000000000 subseq=4 -[Sample-CP] g=4 c=1412 u=-1.0000000000 subseq=5 -[Sample-CP] g=5 c=823 u=-1.0000000000 subseq=6 -[Sample-CP] g=6 c=470 u=-1.0000000000 subseq=7 -[Sample-CP] g=7 c=996 u=-1.0000000000 subseq=8 -[Sample-CP] g=8 c=318 u=-1.0000000000 subseq=9 -[Sample-CP] g=9 c=223 u=-1.0000000000 subseq=10 -[Sample-CP] g=10 c=426 u=-1.0000000000 subseq=11 -[Sample-CP] g=11 c=625 u=-1.0000000000 subseq=12 -[Sample-CP] g=12 c=866 u=-1.0000000000 subseq=13 -[Sample-CP] g=13 c=1456 u=-1.0000000000 subseq=14 -[Sample-CP] g=14 c=1047 u=-1.0000000000 subseq=15 [Debug] codes-step0: [16] first4: 27.000000 1408.000000 1824.000000 417.000000 [Debug] next-emb-step0: [2048] first4: 0.005714 0.034105 -0.000204 0.073642 -[Debug] talker-hidden-step1: [2048] first4: 0.590231 2.653425 0.827661 0.831199 -[Sample] step=1 c0=1902 u=-1.0000000000 subseq=16 -[Sample-CP] g=0 c=1924 u=-1.0000000000 subseq=17 -[Sample-CP] g=1 c=2025 u=-1.0000000000 subseq=18 -[Sample-CP] g=2 c=324 u=-1.0000000000 subseq=19 -[Sample-CP] g=3 c=405 u=-1.0000000000 subseq=20 -[Sample-CP] g=4 c=390 u=-1.0000000000 subseq=21 -[Sample-CP] g=5 c=447 u=-1.0000000000 subseq=22 -[Sample-CP] g=6 c=1059 u=-1.0000000000 subseq=23 -[Sample-CP] g=7 c=564 u=-1.0000000000 subseq=24 -[Sample-CP] g=8 c=1004 u=-1.0000000000 subseq=25 -[Sample-CP] g=9 c=1702 u=-1.0000000000 subseq=26 -[Sample-CP] g=10 c=1887 u=-1.0000000000 subseq=27 -[Sample-CP] g=11 c=1713 u=-1.0000000000 subseq=28 -[Sample-CP] g=12 c=338 u=-1.0000000000 subseq=29 -[Sample-CP] g=13 c=715 u=-1.0000000000 subseq=30 -[Sample-CP] g=14 c=1842 u=-1.0000000000 subseq=31 -[Pipeline] Generated 8 frames -[Pipeline] Generated 16 frames -[Pipeline] Generated 24 frames -[Pipeline] Generated 32 frames -[Pipeline] Generated 40 frames -[Pipeline] Generated 48 frames -[Pipeline] Generated 56 frames -[Pipeline] Generated 64 frames -[Pipeline] Generation done : 64 frames -[Debug] codes-full: [64, 16] first4: 27.000000 1408.000000 1824.000000 417.000000 -[Debug] output-audio: [122880] first4: -0.000101 0.000003 -0.000779 -0.000457 -[Perf] PromptBuild 2.8 ms -[Perf] Prefill 16.8 ms (T_ctx prefill) -[Perf] TTFA 26.1 ms (first frame codes) -[Perf] TalkerDecode 181.9 ms (64 frames, 2.84 ms/frame) -[Perf] CodePredictor 328.5 ms (5.13 ms/frame) -[Perf] HostCompose 0.1 ms (c0 sample + next emb) -[Perf] CodecDecode 29.7 ms -[Perf] Total 561.0 ms (64 frames, 7.98 ms/frame AR, audio 5.12 s, RTF 0.110) -[WAV] Wrote cpp/clone/clone-cpp.wav: 122880 samples, 24000 Hz, mono S16 -[Pipeline] Wrote 122880 samples (5.12 s) -> cpp/clone/clone-cpp.wav +/mnt/workspace/git/qwentts.cpp/ggml/src/ggml-vulkan/ggml-vulkan.cpp:11497: GGML_ASSERT(dst->op != GGML_OP_GET_ROWS || (a_offset == 0 && b_offset == 0 && d_offset == 0)) failed +[New LWP 121736] +[New LWP 121735] +[New LWP 121734] +[New LWP 121733] +[New LWP 121732] +[New LWP 121731] +[New LWP 121730] +[New LWP 121729] +[New LWP 121728] +[New LWP 121727] +[New LWP 121726] +[New LWP 121725] +[New LWP 121724] +[New LWP 121723] +[New LWP 121722] +[New LWP 121721] +[New LWP 121720] +[New LWP 121719] +[New LWP 121718] +[New LWP 121717] +[Thread debugging using libthread_db enabled] +Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1". +0x00007f9b896a69ee in ?? () from /lib/x86_64-linux-gnu/libc.so.6 +#0 0x00007f9b896a69ee in ?? () from /lib/x86_64-linux-gnu/libc.so.6 +#1 0x00007f9b8969b668 in ?? () from /lib/x86_64-linux-gnu/libc.so.6 +#2 0x00007f9b8969b6ad in ?? () from /lib/x86_64-linux-gnu/libc.so.6 +#3 0x00007f9b897067c7 in wait4 () from /lib/x86_64-linux-gnu/libc.so.6 +#4 0x00007f9b89c75a93 in ggml_print_backtrace () from /mnt/workspace/git/qwentts.cpp/build/libggml-base.so.0 +#5 0x00007f9b89c75bfe in ggml_abort () from /mnt/workspace/git/qwentts.cpp/build/libggml-base.so.0 +#6 0x00007f99408d4bf1 in void init_pushconst_tensor_offsets(ggml_backend_vk_context*, vk_op_binary_push_constants&, ggml_tensor const*, ggml_tensor const*, ggml_tensor const*, ggml_tensor const*, ggml_tensor*) () from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so +#7 0x00007f9940a49201 in void ggml_vk_op_f32(ggml_backend_vk_context*, std::shared_ptr&, ggml_tensor const*, ggml_tensor const*, ggml_tensor const*, ggml_tensor const*, ggml_tensor*, ggml_op, vk_op_binary_push_constants&&) [clone .constprop.0] () from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so +#8 0x00007f9940a64f48 in ggml_vk_build_graph(ggml_backend_vk_context*, ggml_cgraph*, int, ggml_tensor*, int, bool, bool, bool) [clone .isra.0] () from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so +#9 0x00007f9940a679bc in ggml_backend_vk_graph_compute(ggml_backend*, ggml_cgraph*) () from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so +#10 0x00007f9b89c8d96e in ggml_backend_graph_compute () from /mnt/workspace/git/qwentts.cpp/build/libggml-base.so.0 +#11 0x000055f31eca3365 in tts_engine_step(TtsEngine*, std::vector >*) () +#12 0x000055f31ecb066a in pipeline_tts_synthesize(PipelineTTS*, BPETokenizer*, qt_tts_params const*, long, qt_audio*) () +#13 0x000055f31ec8cb80 in qt_synthesize () +#14 0x000055f31ec83cdd in run(Args const&) () +#15 0x000055f31ec81a83 in main () +[Inferior 1 (process 121716) detached] [Input] Prompt: 100 chars: qwentts.cpp is a minimal C++17 and GGML port of Qwen3-TTS fo... [Input] RefAudio: ../examples/freeman.wav [Input] RefText: 213 chars: If you go into different cultures, they have different conce... @@ -147,38 +136,3 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Python] Audio: 120960 samples 24000 Hz 5.04s -> python/clone/clone-python.wav [Quant] Q4_K_M -> ../models/qwen-talker-1.7b-base-Q4_K_M.gguf + ../models/qwen-tokenizer-12hz-Q4_K_M.gguf [GGML] Cmd: ../build/qwen-tts --model ../models/qwen-talker-1.7b-base-Q4_K_M.gguf --codec ../models/qwen-tokenizer-12hz-Q4_K_M.gguf --seed 42 --ref-wav ../examples/freeman.wav --ref-text ../examples/freeman.txt --lang english --max-new 64 --dump cpp/clone -o cpp/clone/clone-cpp.wav --greedy -[GGML] Audio: 122880 samples 24000 Hz 5.12s -> cpp/clone/clone-cpp.wav -[Cossim] PromptIDs exact: 100.00% (38 values) -[Cossim] RefCodes exact: 49.83% (3440 values) -[Cossim] Embed cos: 0.917128 max: 2.2245e-01 mean: 2.1853e-02 -[Cossim] TrailingText cos: 0.999967 max: 1.5749e-03 mean: 2.3645e-04 -[Cossim] TTSPadEmbed cos: 0.999967 max: 1.5749e-03 mean: 2.3645e-04 -[Cossim] L0 cos: 0.961486 max: 8.1847e-01 mean: 2.8364e-02 -[Cossim] L7 cos: 0.997433 max: 3.9308e+01 mean: 1.6337e-01 -[Cossim] L14 cos: 0.997359 max: 3.8675e+01 mean: 2.4737e-01 -[Cossim] L21 cos: 0.995877 max: 3.8922e+01 mean: 7.7256e-01 -[Cossim] L27 cos: 0.956688 max: 1.7483e+03 mean: 1.9751e+00 -[Cossim] Final cos: 0.955767 max: 3.3655e+01 mean: 5.4944e-01 -[Cossim] Logits cos: 0.992621 max: 2.9744e+00 mean: 4.5055e-01 -[Cossim] NextEmbStep0 cos: 0.819095 max: 1.4988e-01 mean: 3.0364e-02 -[Cossim] TalkerHiddenStep1 cos: 0.957568 max: 3.9459e+00 mean: 6.2838e-01 -[Cossim] MelHann cos: 1.000000 max: 1.7881e-07 mean: 3.2043e-08 -[Cossim] MelBasis cos: 1.000000 max: 1.7881e-07 mean: 3.9379e-10 -[Cossim] MelMag cos: 1.000000 max: 3.1307e-05 mean: 5.6396e-07 -[Cossim] MelSpk cos: 1.000000 max: 2.5492e-03 mean: 9.3754e-06 -[Cossim] SeanetInit cos: 1.000000 max: 4.1014e-04 mean: 3.4146e-06 -[Cossim] SeanetResnet0 cos: 0.999987 max: 1.9526e-03 mean: 9.4296e-05 -[Cossim] SeanetStage0 cos: 0.999985 max: 4.2238e-02 mean: 6.5089e-04 -[Cossim] SeanetStage1 cos: 0.999866 max: 7.7454e-01 mean: 4.9543e-03 -[Cossim] SeanetStage3 cos: 0.999995 max: 1.4657e+00 mean: 3.8722e-02 -[Cossim] SeanetOut cos: 0.999963 max: 4.0791e-01 mean: 1.6984e-02 -[Cossim] EncTransformer cos: 0.992979 max: 3.5659e+00 mean: 1.6375e-01 -[Cossim] CodecPreFSQ cos: 0.996263 max: 2.2477e+00 mean: 2.4022e-01 -[Cossim] SpkFrontend cos: 1.000000 max: 2.6600e-03 mean: 1.1964e-04 -[Cossim] SpkBlock3 cos: 1.000000 max: 1.2137e-02 mean: 5.2085e-04 -[Cossim] SpkMFA cos: 1.000000 max: 8.9911e-03 mean: 1.3079e-04 -[Cossim] SpkASP cos: 1.000000 max: 2.2919e-04 mean: 1.7381e-05 -[Cossim] SpeakerEmb cos: 1.000000 max: 8.0518e-05 mean: 1.3621e-05 -[Cossim] CodesFull exact: 2.68% (1008 values) -[Cossim] Audio cos: -0.013397 -[Cossim] WAV stft_cos: 0.197850 samples: 120960 diff --git a/tests/clone-Vulkan0-Q8_0.log b/tests/clone-Vulkan0-Q8_0.log index 02bcb15..f9bb141 100644 --- a/tests/clone-Vulkan0-Q8_0.log +++ b/tests/clone-Vulkan0-Q8_0.log @@ -1,9 +1,9 @@ -[Qwen] qwentts.cpp bcac463 (2026-07-05) +[Qwen] qwentts.cpp 53c84fe (2026-07-20) ggml_cuda_init: found 1 CUDA devices (Total VRAM: 97247 MiB): Device 0: NVIDIA RTX PRO 6000 Blackwell Workstation Edition, compute capability 12.0, VMM: yes, VRAM: 97247 MiB load_backend: loaded CUDA backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cuda.so ggml_vulkan: Found 1 Vulkan devices: -ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 +ggml_vulkan: 0 = NVIDIA RTX PRO 6000 Blackwell Workstation Edition (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | fp4: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2 load_backend: loaded Vulkan backend from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libggml-cpu-zen4.so [Load] Talker backend: Vulkan0 (CPU threads: 16) @@ -23,9 +23,9 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [DAC] Loaded: 4 blocks (strides 8/5/4/3), 24 kHz mono out, weights 142.9 MB [WeightCtx] Loaded 2 tensors, 3.0 MB into backend [Pipeline] Ready: hop 1920 samples @ 24000 Hz mono, 16 codebooks @ 12.5 Hz -[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096 -> 896 MB -[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16 -> 0 MB -[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=on clamp_fp16=off +[KVCache] Allocated: 28 layers, 8 KV heads, head_dim 128, max_seq_len 4096, 1 sets -> 896 MB +[KVCache] Allocated: 5 layers, 8 KV heads, head_dim 128, max_seq_len 16, 1 sets -> 0 MB +[Pipeline] Loaded: arch=1b7 variant=base tokenizer=qwen3_tts_tokenizer_12hz codebooks=16 speaker_encoder=deferred speakers=0 fa=on clamp_fp16=off max_batch=1 [BPE] Loaded from GGUF: 151676 vocab, 151291 merges, eos_id=151643 [BPE] Registered 5 arch special tokens (total specials=6) [WAV] Read buffer: 380672 samples, 22050 Hz, 1 ch, 16 bit @@ -33,7 +33,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Audio-Resample] Done: 380672 -> 414337 samples [WeightCtx] Loaded 76 tensors, 34.9 MB into backend [SpeakerEncoder] Loaded: enc_dim=2048 sr=24000 mel_dim=128 hidden=512 mfa=1536 asp_attn=128 se=128 scale=8 -[Pipeline] Speaker encoder lazy loaded in 4 ms +[Pipeline] Speaker encoder lazy loaded in 5 ms [Debug] mel-spk: [1618, 128] first4: -6.330339 -6.170074 -6.011195 -6.243161 [Debug] mel-hann: [1024] first4: 0.000000 0.000009 0.000038 0.000085 [Debug] mel-basis: [128, 513] first4: 0.000000 0.033550 0.008569 0.000000 @@ -51,7 +51,7 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [EncDownsample] Loaded: k=4 stride=2, 512 -> 512 channels, weights 2.0 MB [WeightCtx] Loaded 20 tensors, 34.0 MB into backend [EncQuantizer] Loaded: 16 codebooks (1 semantic + 15 acoustic), 2048 entries x 256 dim, hidden 512, weights 34.0 MB -[Pipeline] Codec encoder lazy loaded in 10 ms +[Pipeline] Codec encoder lazy loaded in 11 ms [Debug] audio-input: [412800] first4: -0.000546 -0.001162 -0.000871 0.000340 [Debug] seanet-out: [430, 512] first4: -0.138303 13.699992 -0.351357 3.025051 [Debug] enc-transformer-out: [430, 512] first4: 0.633121 3.145168 -0.143299 0.564666 @@ -76,62 +76,51 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Debug] talker-hidden-prefill-l27: [225, 2048] first4: 6.677526 -34.405407 -7.383842 28.511698 [Debug] talker-hidden-prefill-final: [225, 2048] first4: 0.227110 -1.221941 -0.233353 0.883898 [Debug] talker-logits-prefill: [3072] first4: 1.264160 -7.826172 -7.041016 -10.638672 +[Batch] Admit slot=0 T_ctx=225 prefill=2227.7 ms build=1539.1 ms (stall for 0 active slots) [Sample] step=0 c0=27 u=-1.0000000000 subseq=0 -[Sample-CP] g=0 c=1408 u=-1.0000000000 subseq=1 -[Sample-CP] g=1 c=1824 u=-1.0000000000 subseq=2 -[Sample-CP] g=2 c=400 u=-1.0000000000 subseq=3 -[Sample-CP] g=3 c=550 u=-1.0000000000 subseq=4 -[Sample-CP] g=4 c=1506 u=-1.0000000000 subseq=5 -[Sample-CP] g=5 c=1452 u=-1.0000000000 subseq=6 -[Sample-CP] g=6 c=470 u=-1.0000000000 subseq=7 -[Sample-CP] g=7 c=1713 u=-1.0000000000 subseq=8 -[Sample-CP] g=8 c=1221 u=-1.0000000000 subseq=9 -[Sample-CP] g=9 c=18 u=-1.0000000000 subseq=10 -[Sample-CP] g=10 c=1502 u=-1.0000000000 subseq=11 -[Sample-CP] g=11 c=1677 u=-1.0000000000 subseq=12 -[Sample-CP] g=12 c=411 u=-1.0000000000 subseq=13 -[Sample-CP] g=13 c=82 u=-1.0000000000 subseq=14 -[Sample-CP] g=14 c=516 u=-1.0000000000 subseq=15 [Debug] codes-step0: [16] first4: 27.000000 1408.000000 1824.000000 400.000000 [Debug] next-emb-step0: [2048] first4: 0.021546 -0.036059 0.003111 0.073181 -[Debug] talker-hidden-step1: [2048] first4: 0.372459 2.555521 -0.076819 0.171399 -[Sample] step=1 c0=1902 u=-1.0000000000 subseq=16 -[Sample-CP] g=0 c=1924 u=-1.0000000000 subseq=17 -[Sample-CP] g=1 c=2025 u=-1.0000000000 subseq=18 -[Sample-CP] g=2 c=324 u=-1.0000000000 subseq=19 -[Sample-CP] g=3 c=350 u=-1.0000000000 subseq=20 -[Sample-CP] g=4 c=985 u=-1.0000000000 subseq=21 -[Sample-CP] g=5 c=356 u=-1.0000000000 subseq=22 -[Sample-CP] g=6 c=1544 u=-1.0000000000 subseq=23 -[Sample-CP] g=7 c=1058 u=-1.0000000000 subseq=24 -[Sample-CP] g=8 c=1004 u=-1.0000000000 subseq=25 -[Sample-CP] g=9 c=1702 u=-1.0000000000 subseq=26 -[Sample-CP] g=10 c=437 u=-1.0000000000 subseq=27 -[Sample-CP] g=11 c=729 u=-1.0000000000 subseq=28 -[Sample-CP] g=12 c=781 u=-1.0000000000 subseq=29 -[Sample-CP] g=13 c=85 u=-1.0000000000 subseq=30 -[Sample-CP] g=14 c=1435 u=-1.0000000000 subseq=31 -[Pipeline] Generated 8 frames -[Pipeline] Generated 16 frames -[Pipeline] Generated 24 frames -[Pipeline] Generated 32 frames -[Pipeline] Generated 40 frames -[Pipeline] Generated 48 frames -[Pipeline] Generated 56 frames -[Pipeline] Generated 64 frames -[Pipeline] Generation done : 64 frames -[Debug] codes-full: [64, 16] first4: 27.000000 1408.000000 1824.000000 400.000000 -[Debug] output-audio: [122880] first4: 0.000658 0.000394 -0.000943 -0.000566 -[Perf] PromptBuild 2.7 ms -[Perf] Prefill 17.1 ms (T_ctx prefill) -[Perf] TTFA 26.5 ms (first frame codes) -[Perf] TalkerDecode 201.3 ms (64 frames, 3.14 ms/frame) -[Perf] CodePredictor 334.3 ms (5.22 ms/frame) -[Perf] HostCompose 0.1 ms (c0 sample + next emb) -[Perf] CodecDecode 30.3 ms -[Perf] Total 586.9 ms (64 frames, 8.37 ms/frame AR, audio 5.12 s, RTF 0.115) -[WAV] Wrote cpp/clone/clone-cpp.wav: 122880 samples, 24000 Hz, mono S16 -[Pipeline] Wrote 122880 samples (5.12 s) -> cpp/clone/clone-cpp.wav +/mnt/workspace/git/qwentts.cpp/ggml/src/ggml-vulkan/ggml-vulkan.cpp:11497: GGML_ASSERT(dst->op != GGML_OP_GET_ROWS || (a_offset == 0 && b_offset == 0 && d_offset == 0)) failed +[New LWP 121584] +[New LWP 121583] +[New LWP 121582] +[New LWP 121581] +[New LWP 121580] +[New LWP 121579] +[New LWP 121578] +[New LWP 121577] +[New LWP 121576] +[New LWP 121575] +[New LWP 121574] +[New LWP 121573] +[New LWP 121572] +[New LWP 121571] +[New LWP 121570] +[New LWP 121569] +[New LWP 121568] +[New LWP 121567] +[New LWP 121566] +[New LWP 121565] +[Thread debugging using libthread_db enabled] +Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1". +0x00007fd1dd7b69ee in ?? () from /lib/x86_64-linux-gnu/libc.so.6 +#0 0x00007fd1dd7b69ee in ?? () from /lib/x86_64-linux-gnu/libc.so.6 +#1 0x00007fd1dd7ab668 in ?? () from /lib/x86_64-linux-gnu/libc.so.6 +#2 0x00007fd1dd7ab6ad in ?? () from /lib/x86_64-linux-gnu/libc.so.6 +#3 0x00007fd1dd8167c7 in wait4 () from /lib/x86_64-linux-gnu/libc.so.6 +#4 0x00007fd1ddd2ca93 in ggml_print_backtrace () from /mnt/workspace/git/qwentts.cpp/build/libggml-base.so.0 +#5 0x00007fd1ddd2cbfe in ggml_abort () from /mnt/workspace/git/qwentts.cpp/build/libggml-base.so.0 +#6 0x00007fcf948d4bf1 in void init_pushconst_tensor_offsets(ggml_backend_vk_context*, vk_op_binary_push_constants&, ggml_tensor const*, ggml_tensor const*, ggml_tensor const*, ggml_tensor const*, ggml_tensor*) () from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so +#7 0x00007fcf94a49201 in void ggml_vk_op_f32(ggml_backend_vk_context*, std::shared_ptr&, ggml_tensor const*, ggml_tensor const*, ggml_tensor const*, ggml_tensor const*, ggml_tensor*, ggml_op, vk_op_binary_push_constants&&) [clone .constprop.0] () from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so +#8 0x00007fcf94a64f48 in ggml_vk_build_graph(ggml_backend_vk_context*, ggml_cgraph*, int, ggml_tensor*, int, bool, bool, bool) [clone .isra.0] () from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so +#9 0x00007fcf94a679bc in ggml_backend_vk_graph_compute(ggml_backend*, ggml_cgraph*) () from /mnt/workspace/git/qwentts.cpp/build/libggml-vulkan.so +#10 0x00007fd1ddd4496e in ggml_backend_graph_compute () from /mnt/workspace/git/qwentts.cpp/build/libggml-base.so.0 +#11 0x000055b420b50365 in tts_engine_step(TtsEngine*, std::vector >*) () +#12 0x000055b420b5d66a in pipeline_tts_synthesize(PipelineTTS*, BPETokenizer*, qt_tts_params const*, long, qt_audio*) () +#13 0x000055b420b39b80 in qt_synthesize () +#14 0x000055b420b30cdd in run(Args const&) () +#15 0x000055b420b2ea83 in main () +[Inferior 1 (process 121564) detached] [Input] Prompt: 100 chars: qwentts.cpp is a minimal C++17 and GGML port of Qwen3-TTS fo... [Input] RefAudio: ../examples/freeman.wav [Input] RefText: 213 chars: If you go into different cultures, they have different conce... @@ -147,38 +136,3 @@ load_backend: loaded CPU backend from /mnt/workspace/git/qwentts.cpp/build/libgg [Python] Audio: 120960 samples 24000 Hz 5.04s -> python/clone/clone-python.wav [Quant] Q8_0 -> ../models/qwen-talker-1.7b-base-Q8_0.gguf + ../models/qwen-tokenizer-12hz-Q8_0.gguf [GGML] Cmd: ../build/qwen-tts --model ../models/qwen-talker-1.7b-base-Q8_0.gguf --codec ../models/qwen-tokenizer-12hz-Q8_0.gguf --seed 42 --ref-wav ../examples/freeman.wav --ref-text ../examples/freeman.txt --lang english --max-new 64 --dump cpp/clone -o cpp/clone/clone-cpp.wav --greedy -[GGML] Audio: 122880 samples 24000 Hz 5.12s -> cpp/clone/clone-cpp.wav -[Cossim] PromptIDs exact: 100.00% (38 values) -[Cossim] RefCodes exact: 81.16% (3440 values) -[Cossim] Embed cos: 0.970174 max: 1.9001e-01 mean: 8.9738e-03 -[Cossim] TrailingText cos: 0.999999 max: 1.9968e-04 mean: 4.5182e-05 -[Cossim] TTSPadEmbed cos: 0.999999 max: 1.9968e-04 mean: 4.5182e-05 -[Cossim] L0 cos: 0.985826 max: 7.7038e-01 mean: 1.1813e-02 -[Cossim] L7 cos: 0.999979 max: 5.4011e+00 mean: 3.6051e-02 -[Cossim] L14 cos: 0.999957 max: 5.4243e+00 mean: 7.0612e-02 -[Cossim] L21 cos: 0.999605 max: 1.8733e+01 mean: 2.5057e-01 -[Cossim] L27 cos: 0.998543 max: 5.3200e+01 mean: 5.6948e-01 -[Cossim] Final cos: 0.992844 max: 1.9032e+01 mean: 1.8018e-01 -[Cossim] Logits cos: 0.999587 max: 7.9291e-01 mean: 1.0517e-01 -[Cossim] NextEmbStep0 cos: 0.848337 max: 2.7267e-01 mean: 2.8037e-02 -[Cossim] TalkerHiddenStep1 cos: 0.993831 max: 1.5021e+00 mean: 2.3902e-01 -[Cossim] MelHann cos: 1.000000 max: 1.7881e-07 mean: 3.2043e-08 -[Cossim] MelBasis cos: 1.000000 max: 1.7881e-07 mean: 3.9379e-10 -[Cossim] MelMag cos: 1.000000 max: 3.1307e-05 mean: 5.6396e-07 -[Cossim] MelSpk cos: 1.000000 max: 2.5492e-03 mean: 9.3754e-06 -[Cossim] SeanetInit cos: 1.000000 max: 4.1014e-04 mean: 3.4146e-06 -[Cossim] SeanetResnet0 cos: 0.999987 max: 1.9526e-03 mean: 9.4296e-05 -[Cossim] SeanetStage0 cos: 0.999985 max: 4.2238e-02 mean: 6.5089e-04 -[Cossim] SeanetStage1 cos: 0.999866 max: 7.7454e-01 mean: 4.9543e-03 -[Cossim] SeanetStage3 cos: 0.999995 max: 1.4657e+00 mean: 3.8722e-02 -[Cossim] SeanetOut cos: 0.999963 max: 4.0791e-01 mean: 1.6984e-02 -[Cossim] EncTransformer cos: 0.999932 max: 3.3955e-01 mean: 1.5945e-02 -[Cossim] CodecPreFSQ cos: 0.999951 max: 4.5186e-01 mean: 2.6992e-02 -[Cossim] SpkFrontend cos: 1.000000 max: 2.6600e-03 mean: 1.1964e-04 -[Cossim] SpkBlock3 cos: 1.000000 max: 1.2137e-02 mean: 5.2085e-04 -[Cossim] SpkMFA cos: 1.000000 max: 8.9911e-03 mean: 1.3079e-04 -[Cossim] SpkASP cos: 1.000000 max: 2.2919e-04 mean: 1.7381e-05 -[Cossim] SpeakerEmb cos: 1.000000 max: 8.0518e-05 mean: 1.3621e-05 -[Cossim] CodesFull exact: 2.38% (1008 values) -[Cossim] Audio cos: 0.013854 -[Cossim] WAV stft_cos: 0.183103 samples: 120960