[Registry] MiniMax-Music3-condition_encoder-F32.gguf -> mm3-cond
[Registry] MiniMax-Music3-language_model-BF16.gguf -> mm3-lm
[Registry] MiniMax-Music3-language_model-Q5_K_M.gguf -> mm3-lm
[Registry] MiniMax-Music3-language_model-Q6_K.gguf -> mm3-lm
[Registry] MiniMax-Music3-language_model-Q8_0.gguf -> mm3-lm
[Registry] MiniMax-Music3-rvq_depth_decoder-BF16.gguf -> mm3-depth
[Registry] MiniMax-Music3-rvq_depth_decoder-Q8_0.gguf -> mm3-depth
[Registry] MiniMax-Music3-transformer-F32.gguf -> mm3-dit
[Registry] MiniMax-Music3-transformer-Q4_K_M.gguf -> mm3-dit
[Registry] MiniMax-Music3-transformer-Q5_K_M.gguf -> mm3-dit
[Registry] MiniMax-Music3-transformer-Q6_K.gguf -> mm3-dit
[Registry] MiniMax-Music3-transformer-Q8_0.gguf -> mm3-dit
[Registry] MiniMax-Music3-vocoder-F32.gguf -> mm3-vae
[Store] Created (policy=STRICT)
[Pipeline] Flash attention disabled
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/minimaxmusic.cpp/build/libggml-cuda.so
load_backend: loaded Vulkan backend from /mnt/workspace/git/minimaxmusic.cpp/build/libggml-vulkan.so
load_backend: loaded CPU backend from /mnt/workspace/git/minimaxmusic.cpp/build/libggml-cpu-zen4.so
[Load] LM backend: Vulkan0 (CPU threads: 16)
[GGUF] ../models/MiniMax-Music3-language_model-BF16.gguf: 399 tensors, data at offset 5346496
[LM-Config] 36L, H=4096, V=200000, Nh=32, Nkv=8, D=128, tied=0
[Qwen3] Attn: Q+K+V fused
[Qwen3] MLP: gate+up fused
[WeightCtx] Loaded 399 tensors, 16374.2 MB into backend
[LM-KV] Allocated 2 sets x 36 layers (4D batched), 2880.0 MB
[Store] Load LM: 2068 ms
[GGUF] ../models/MiniMax-Music3-rvq_depth_decoder-BF16.gguf: 47 tensors, data at offset 3456
[Load] Depth backend: Vulkan0 (shared)
[WeightCtx] Loaded 47 tensors, 1232.4 MB into backend
[Depth] Loaded: 4 layers, dim 4096, 7 heads
[Store] Load Depth: 143 ms
[BPE] Loaded from GGUF: 151675 vocab, 151387 merges
[Prompt] 40 tokens
[AR] Prefill 133 ms, 40 tokens, CFG=1.50, top_k=50, songs=1
[AR] Frame 0/200
[AR] Frame 100/200
[AR] Song 0: 200 frames (8.0s of music)
[Debug] frame_hiddens: [200, 8, 4096] first4: 0.921327 1.899262 -0.903675 -0.696878
[AR] 200 frames total, 29.7 s (148.4 ms/frame)
[Store] Unload Depth (1232.4 MB)
[Store] Unload LM (16374.2 MB)
[GGUF] ../models/MiniMax-Music3-condition_encoder-F32.gguf: 4 tensors, data at offset 640
load_backend: loaded CUDA backend from /mnt/workspace/git/minimaxmusic.cpp/build/libggml-cuda.so
load_backend: loaded Vulkan backend from /mnt/workspace/git/minimaxmusic.cpp/build/libggml-vulkan.so
load_backend: loaded CPU backend from /mnt/workspace/git/minimaxmusic.cpp/build/libggml-cpu-zen4.so
[Load] Cond backend: Vulkan0 (CPU threads: 16)
[Cond] Loaded: mix over 8 states, proj 4096 -> 2048
[Store] Load Cond: 49 ms
[GGUF] ../models/MiniMax-Music3-transformer-Q8_0.gguf: 441 tensors, data at offset 31712
[Load] DiT backend: Vulkan0 (shared)
[WeightCtx] Loaded 436 tensors, 2477.0 MB into backend
[DiT] Loaded: 36 layers, dim 2048, flash_attn=1
[Store] Load DiT: 332 ms
[GGUF] ../models/MiniMax-Music3-vocoder-F32.gguf: 182 tensors, data at offset 12896
[Load] VAE backend: Vulkan0 (shared)
[VAE] Backend: Vulkan0, Weight buffer: 103.4 MB
[VAE] Loaded: 4 blocks, upsample=512x, F32 activations
[Store] Load VAE: 103 ms
[Debug] noise: [689, 128] first4: 0.194019 2.161374 -0.172051 0.849060
[DiT] Graph: 1344 nodes, T=689, B=2
[Debug] temb_t: [2048] first4: 3.626598 -3.071267 -3.962117 4.059601
[Debug] hidden_after_preprocess: [689, 2304] first4: 0.059375 3.944638 -0.796440 2.207947
[Debug] hidden_after_proj_in: [689, 2048] first4: 3.501465 -5.733887 -1.914062 -0.181519
[Debug] layer0_sa_output: [690, 2048] first4: 1.568481 -1.834229 0.383606 1.460083
[Debug] hidden_after_layer0: [690, 2048] first4: -2.219043 -1.240525 -3.199054 3.629245
[Debug] hidden_after_layer6: [690, 2048] first4: -2.618366 0.965501 1.851092 0.461585
[Debug] hidden_after_layer12: [690, 2048] first4: -0.687699 0.584312 3.981623 2.956595
[Debug] hidden_after_layer18: [690, 2048] first4: -2.770306 -1.291964 3.050559 -1.029601
[Debug] hidden_after_layer35: [690, 2048] first4: -2.539729 1.861158 -1.487182 -4.743420
[Debug] dit_step0_vt_cond: [689, 128] first4: -0.342529 -2.040405 -0.331146 -0.667786
[Debug] dit_step0_vt_uncond: [689, 128] first4: -0.357056 -2.277603 -0.424774 -0.597168
[Debug] dit_step0_vt: [689, 128] first4: -0.332361 -1.874367 -0.265607 -0.717218
[Debug] dit_step0_xt: [689, 128] first4: 0.182940 2.098895 -0.180904 0.825153
[Debug] dit_step1_vt_cond: [689, 128] first4: -0.288574 -1.807945 -0.138672 -0.547852
[Debug] dit_step1_vt_uncond: [689, 128] first4: -0.626762 -1.614136 -0.260254 -0.346558
[Debug] dit_step1_vt: [689, 128] first4: -0.051843 -1.943612 -0.053564 -0.688757
[Debug] dit_step1_xt: [689, 128] first4: 0.181212 2.034108 -0.182690 0.802194
[Debug] dit_step2_vt_cond: [689, 128] first4: -0.290649 -1.634094 -0.046875 -0.456421
[Debug] dit_step2_vt_uncond: [689, 128] first4: -0.809814 -1.194580 -0.247742 -0.135376
[Debug] dit_step2_vt: [689, 128] first4: 0.072766 -1.941754 0.093732 -0.681152
[Debug] dit_step2_xt: [689, 128] first4: 0.183638 1.969383 -0.179565 0.779489
[Debug] dit_step3_vt_cond: [689, 128] first4: -0.314331 -1.513428 0.082031 -0.401245
[Debug] dit_step3_vt_uncond: [689, 128] first4: -0.253662 -1.459595 -0.361328 -0.531616
[Debug] dit_step3_vt: [689, 128] first4: -0.356799 -1.551111 0.392383 -0.309985
[Debug] dit_step3_xt: [689, 128] first4: 0.171744 1.917679 -0.166486 0.769156
[Debug] dit_step4_vt_cond: [689, 128] first4: -0.330063 -1.381836 0.154297 -0.333374
[Debug] dit_step4_vt_uncond: [689, 128] first4: -0.179321 -1.409424 -0.092285 -0.530762
[Debug] dit_step4_vt: [689, 128] first4: -0.435582 -1.362524 0.326904 -0.195203
[Debug] dit_step4_xt: [689, 128] first4: 0.157225 1.872262 -0.155589 0.762650
[Debug] dit_step5_vt_cond: [689, 128] first4: -0.329012 -1.239502 0.145508 -0.240112
[Debug] dit_step5_vt_uncond: [689, 128] first4: -0.160156 -1.256104 -0.005859 -0.404785
[Debug] dit_step5_vt: [689, 128] first4: -0.447211 -1.227881 0.251465 -0.124841
[Debug] dit_step5_xt: [689, 128] first4: 0.142318 1.831332 -0.147207 0.758488
[Debug] dit_step6_vt_cond: [689, 128] first4: -0.312714 -1.048096 0.070312 -0.128418
[Debug] dit_step6_vt_uncond: [689, 128] first4: -0.068481 -1.052002 -0.170898 -0.232788
[Debug] dit_step6_vt: [689, 128] first4: -0.483676 -1.045361 0.239160 -0.055359
[Debug] dit_step6_xt: [689, 128] first4: 0.126195 1.796487 -0.139235 0.756643
[Debug] dit_step7_vt_cond: [689, 128] first4: -0.295288 -0.788818 -0.037598 -0.030029
[Debug] dit_step7_vt_uncond: [689, 128] first4: -0.021984 -0.718750 -0.320923 -0.037354
[Debug] dit_step7_vt: [689, 128] first4: -0.486601 -0.837866 0.160730 -0.024902
[Debug] dit_step7_xt: [689, 128] first4: 0.109975 1.768558 -0.133877 0.755813
[Debug] dit_step8_vt_cond: [689, 128] first4: -0.263428 -0.521973 -0.148438 0.068054
[Debug] dit_step8_vt_uncond: [689, 128] first4: 0.002563 -0.327591 -0.437622 0.119263
[Debug] dit_step8_vt: [689, 128] first4: -0.449622 -0.658040 0.053992 0.032208
[Debug] dit_step8_xt: [689, 128] first4: 0.094988 1.746623 -0.132077 0.756886
[Debug] dit_step9_vt_cond: [689, 128] first4: -0.214966 -0.257141 -0.253418 0.165039
[Debug] dit_step9_vt_uncond: [689, 128] first4: 0.041809 0.037811 -0.530579 0.255432
[Debug] dit_step9_vt: [689, 128] first4: -0.394708 -0.463608 -0.059406 0.101764
[Debug] dit_step9_xt: [689, 128] first4: 0.081831 1.731170 -0.134058 0.760279
[Debug] dit_step10_vt_cond: [689, 128] first4: -0.150635 -0.012056 -0.352205 0.258301
[Debug] dit_step10_vt_uncond: [689, 128] first4: 0.070923 0.341614 -0.580322 0.368530
[Debug] dit_step10_vt: [689, 128] first4: -0.305725 -0.259625 -0.192523 0.181140
[Debug] dit_step10_xt: [689, 128] first4: 0.071640 1.722516 -0.140475 0.766317
[Debug] dit_step11_vt_cond: [689, 128] first4: -0.084473 0.223145 -0.448975 0.338055
[Debug] dit_step11_vt_uncond: [689, 128] first4: 0.108521 0.618500 -0.635010 0.469727
[Debug] dit_step11_vt: [689, 128] first4: -0.219568 -0.053604 -0.318750 0.245884
[Debug] dit_step11_xt: [689, 128] first4: 0.064321 1.720729 -0.151100 0.774513
[Debug] dit_step12_vt_cond: [689, 128] first4: -0.015137 0.429077 -0.539124 0.406982
[Debug] dit_step12_vt_uncond: [689, 128] first4: 0.153320 0.841801 -0.688644 0.535156
[Debug] dit_step12_vt: [689, 128] first4: -0.133057 0.140171 -0.434459 0.317261
[Debug] dit_step12_xt: [689, 128] first4: 0.059886 1.725401 -0.165582 0.785088
[Debug] dit_step13_vt_cond: [689, 128] first4: 0.050964 0.646484 -0.621094 0.468262
[Debug] dit_step13_vt_uncond: [689, 128] first4: 0.181396 1.049561 -0.750244 0.583984
[Debug] dit_step13_vt: [689, 128] first4: -0.040338 0.364331 -0.530688 0.387256
[Debug] dit_step13_xt: [689, 128] first4: 0.058541 1.737545 -0.183272 0.797997
[Debug] dit_step14_vt_cond: [689, 128] first4: 0.108459 0.836426 -0.680542 0.514404
[Debug] dit_step14_vt_uncond: [689, 128] first4: 0.205078 1.196411 -0.814880 0.627441
[Debug] dit_step14_vt: [689, 128] first4: 0.040826 0.584436 -0.586505 0.435278
[Debug] dit_step14_xt: [689, 128] first4: 0.059902 1.757027 -0.202822 0.812506
[Debug] dit_step15_vt_cond: [689, 128] first4: 0.156738 1.037598 -0.728714 0.557861
[Debug] dit_step15_vt_uncond: [689, 128] first4: 0.216553 1.343994 -0.854187 0.677490
[Debug] dit_step15_vt: [689, 128] first4: 0.114868 0.823120 -0.640883 0.474121
[Debug] dit_step15_xt: [689, 128] first4: 0.063731 1.784464 -0.224185 0.828310
[Debug] dit_step16_vt_cond: [689, 128] first4: 0.195312 1.223877 -0.761719 0.605713
[Debug] dit_step16_vt_uncond: [689, 128] first4: 0.228226 1.487305 -0.877625 0.744385
[Debug] dit_step16_vt: [689, 128] first4: 0.172273 1.039478 -0.680585 0.508643
[Debug] dit_step16_xt: [689, 128] first4: 0.069474 1.819113 -0.246871 0.845265
[Debug] dit_step17_vt_cond: [689, 128] first4: 0.220886 1.410156 -0.775146 0.654541
[Debug] dit_step17_vt_uncond: [689, 128] first4: 0.239807 1.649658 -0.884521 0.799316
[Debug] dit_step17_vt: [689, 128] first4: 0.207642 1.242505 -0.698584 0.553198
[Debug] dit_step17_xt: [689, 128] first4: 0.076395 1.860530 -0.270157 0.863705
[Debug] dit_step18_vt_cond: [689, 128] first4: 0.251923 1.552734 -0.773926 0.703613
[Debug] dit_step18_vt_uncond: [689, 128] first4: 0.251953 1.814941 -0.868652 0.857178
[Debug] dit_step18_vt: [689, 128] first4: 0.251901 1.369190 -0.707617 0.596118
[Debug] dit_step18_xt: [689, 128] first4: 0.084792 1.906170 -0.293744 0.883575
[Debug] dit_step19_vt_cond: [689, 128] first4: 0.268829 1.729492 -0.757324 0.752930
[Debug] dit_step19_vt_uncond: [689, 128] first4: 0.276733 1.925537 -0.833496 0.898193
[Debug] dit_step19_vt: [689, 128] first4: 0.263297 1.592261 -0.704004 0.651245
[Debug] dit_step19_xt: [689, 128] first4: 0.093568 1.959245 -0.317211 0.905283
[Debug] dit_step20_vt_cond: [689, 128] first4: 0.294128 1.822510 -0.740723 0.807129
[Debug] dit_step20_vt_uncond: [689, 128] first4: 0.306274 2.021484 -0.794922 0.929443
[Debug] dit_step20_vt: [689, 128] first4: 0.285626 1.683228 -0.702783 0.721509
[Debug] dit_step20_xt: [689, 128] first4: 0.103089 2.015352 -0.340637 0.929334
[Debug] dit_step21_vt_cond: [689, 128] first4: 0.308182 1.972717 -0.723145 0.857605
[Debug] dit_step21_vt_uncond: [689, 128] first4: 0.333679 2.111481 -0.742432 0.966797
[Debug] dit_step21_vt: [689, 128] first4: 0.290334 1.875583 -0.709644 0.781171
[Debug] dit_step21_xt: [689, 128] first4: 0.112767 2.077872 -0.364292 0.955373
[Debug] dit_step22_vt_cond: [689, 128] first4: 0.320465 2.078583 -0.697632 0.898560
[Debug] dit_step22_vt_uncond: [689, 128] first4: 0.360352 2.224945 -0.690125 0.998657
[Debug] dit_step22_vt: [689, 128] first4: 0.292545 1.976129 -0.702887 0.828491
[Debug] dit_step22_xt: [689, 128] first4: 0.122518 2.143743 -0.387721 0.982989
[Debug] dit_step23_vt_cond: [689, 128] first4: 0.330688 2.188232 -0.675476 0.952881
[Debug] dit_step23_vt_uncond: [689, 128] first4: 0.388672 2.273560 -0.644470 1.019043
[Debug] dit_step23_vt: [689, 128] first4: 0.290100 2.128503 -0.697180 0.906567
[Debug] dit_step23_xt: [689, 128] first4: 0.132188 2.214693 -0.410961 1.013208
[Debug] dit_step24_vt_cond: [689, 128] first4: 0.338226 2.285034 -0.650879 1.012207
[Debug] dit_step24_vt_uncond: [689, 128] first4: 0.402557 2.374939 -0.603027 1.046875
[Debug] dit_step24_vt: [689, 128] first4: 0.293195 2.222101 -0.684375 0.987939
[Debug] dit_step24_xt: [689, 128] first4: 0.141962 2.288763 -0.433773 1.046139
[Debug] dit_step25_vt_cond: [689, 128] first4: 0.340805 2.376465 -0.630371 1.067383
[Debug] dit_step25_vt_uncond: [689, 128] first4: 0.403748 2.422028 -0.565918 1.081299
[Debug] dit_step25_vt: [689, 128] first4: 0.296745 2.344571 -0.675488 1.057642
[Debug] dit_step25_xt: [689, 128] first4: 0.151853 2.366915 -0.456289 1.081394
[Debug] dit_step26_vt_cond: [689, 128] first4: 0.341675 2.469788 -0.613770 1.122070
[Debug] dit_step26_vt_uncond: [689, 128] first4: 0.391281 2.495148 -0.555176 1.104248
[Debug] dit_step26_vt: [689, 128] first4: 0.306950 2.452035 -0.654785 1.134546
[Debug] dit_step26_xt: [689, 128] first4: 0.162085 2.448650 -0.478116 1.119212
[Debug] dit_step27_vt_cond: [689, 128] first4: 0.332275 2.482941 -0.602539 1.176270
[Debug] dit_step27_vt_uncond: [689, 128] first4: 0.361450 2.553589 -0.548828 1.141602
[Debug] dit_step27_vt: [689, 128] first4: 0.311853 2.433487 -0.640137 1.200537
[Debug] dit_step27_xt: [689, 128] first4: 0.172480 2.529766 -0.499453 1.159230
[Debug] dit_step28_vt_cond: [689, 128] first4: 0.305176 2.548462 -0.589844 1.193604
[Debug] dit_step28_vt_uncond: [689, 128] first4: 0.316162 2.582947 -0.555664 1.158936
[Debug] dit_step28_vt: [689, 128] first4: 0.297485 2.524323 -0.613770 1.217871
[Debug] dit_step28_xt: [689, 128] first4: 0.182396 2.613910 -0.519912 1.199826
[Debug] dit_step29_vt_cond: [689, 128] first4: 0.260254 2.614868 -0.614258 1.199463
[Debug] dit_step29_vt_uncond: [689, 128] first4: 0.255615 2.597168 -0.578125 1.152588
[Debug] dit_step29_vt: [689, 128] first4: 0.263501 2.627258 -0.639551 1.232275
[Debug] dit_step29_xt: [689, 128] first4: 0.191179 2.701486 -0.541231 1.240902
[Debug] dit_x0: [689, 128] first4: 0.191179 2.701486 -0.541231 1.240902
[Debug] window0_cond: [689, 2048] first4: 0.292188 -0.052068 0.070921 0.078521
[Debug] window0_latent: [689, 128] first4: 0.191179 2.701486 -0.541231 1.240902
[DiT] Window 1/1: T=689, 30 steps, 1917 ms (63.9 ms/step)
[DiT] CFG=1.70, 1 windows, 1.9 s
[VAE] Graph: 398 nodes, T_latent=689
[VAE] Decoded: T_latent=689 -> T_audio=352768 (8.00s @ 44.1kHz)
[Debug] vae_audio: [352768, 2] first4: 0.014720 -0.001610 0.004683 0.000531
[VAE] Decode: 1 windows -> 8.0s of audio, 87 ms
[Done] 34.6 s total
[Store] Unload VAE (103.4 MB)
[Store] Unload DiT (2477.0 MB)
[Store] Unload Cond (48.0 MB)
[WAV] Wrote ggml-dit/output.wav: 352768 samples, 44100 Hz, stereo
[Out] ggml-dit/output.wav
[Out] ggml-dit/output.json
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with: 
```
pip install accelerate
```
.
Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with: 
```
pip install accelerate
```
.
[Request] Loaded request0.json
[DiT] steps=30, CFG=1.7 | mm3-dit-Q8_0.gguf
[GGML] Running MiniMax-Music3-transformer-Q8_0.gguf...
[GGML] Done, 135 dump files
[Python] Initializing transformer on cuda (T=689, 30 steps, CFG=1.7)...
[Python] Denoising...
[Python] Decoding audio...
[Python] Done, 104 dump files
[DiT] Cosine similarities GGML vs Python
noise: ggml vs python 1.000000
temb_t: ggml vs python 1.000000
hidden_after_preprocess: ggml vs python 1.000000
hidden_after_proj_in: ggml vs python 0.999994
layer0_sa_output: ggml vs python 0.999999
hidden_after_layer0: ggml vs python 1.000000
hidden_after_layer6: ggml vs python 0.999999
hidden_after_layer12: ggml vs python 0.999999
hidden_after_layer18: ggml vs python 1.000000
hidden_after_layer35: ggml vs python 0.999997
dit_step0_vt_cond: ggml vs python 0.999969
dit_step0_vt_uncond: ggml vs python 0.999961
dit_step0_vt: ggml vs python 0.999972
dit_step0_xt: ggml vs python 1.000000
dit_step5_vt_cond: ggml vs python 0.999922
dit_step5_vt: ggml vs python 0.999903
dit_step5_xt: ggml vs python 0.999994
dit_step10_vt_cond: ggml vs python 0.999912
dit_step10_vt: ggml vs python 0.999892
dit_step10_xt: ggml vs python 0.999975
dit_step15_vt_cond: ggml vs python 0.999888
dit_step15_vt: ggml vs python 0.999868
dit_step15_xt: ggml vs python 0.999954
dit_step20_vt_cond: ggml vs python 0.999861
dit_step20_vt: ggml vs python 0.999842
dit_step20_xt: ggml vs python 0.999938
dit_step25_vt_cond: ggml vs python 0.999830
dit_step25_vt: ggml vs python 0.999813
dit_step25_xt: ggml vs python 0.999927
dit_step29_vt_cond: ggml vs python 0.999874
dit_step29_vt: ggml vs python 0.999860
dit_x0: ggml vs python 0.999921
vae_audio: ggml vs python 0.999801
vae_audio (STFT cosine): ggml vs python 0.999881
[DiT] Error growth GGML vs Python
dit_step0_xt: cos 1.000000, max_err 0.002997, mean_err 0.000343, mean_A -0.001845, std_A 0.964987, mean_B -0.001888, std_B 0.964984
dit_step5_xt: cos 0.999994, max_err 0.024234, mean_err 0.002288, mean_A 0.002223, std_A 0.875719, mean_B 0.001975, std_B 0.875768
dit_step10_xt: cos 0.999975, max_err 0.062484, mean_err 0.004758, mean_A 0.007510, std_A 0.919199, mean_B 0.007025, std_B 0.919308
dit_step15_xt: cos 0.999954, max_err 0.093035, mean_err 0.007589, mean_A 0.012979, std_A 1.078866, mean_B 0.012237, std_B 1.079004
dit_step20_xt: cos 0.999938, max_err 0.133457, mean_err 0.010713, mean_A 0.018319, std_A 1.312090, mean_B 0.017310, std_B 1.312272
dit_step25_xt: cos 0.999927, max_err 0.212113, mean_err 0.014085, mean_A 0.023562, std_A 1.585850, mean_B 0.022269, std_B 1.586108