[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: CUDA0 (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: 1359 ms
[GGUF] ../models/MiniMax-Music3-rvq_depth_decoder-BF16.gguf: 47 tensors, data at offset 3456
[Load] Depth backend: CUDA0 (shared)
[WeightCtx] Loaded 47 tensors, 1232.4 MB into backend
[Depth] Loaded: 4 layers, dim 4096, 7 heads
[Store] Load Depth: 101 ms
[BPE] Loaded from GGUF: 151675 vocab, 151387 merges
[Prompt] 40 tokens
[AR] Prefill 93 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: -1.355042 -1.043213 -0.754630 0.104368
[AR] 200 frames total, 4.2 s (21.1 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: CUDA0 (CPU threads: 16)
[Cond] Loaded: mix over 8 states, proj 4096 -> 2048
[Store] Load Cond: 47 ms
[GGUF] ../models/MiniMax-Music3-transformer-Q6_K.gguf: 441 tensors, data at offset 31712
[Load] DiT backend: CUDA0 (shared)
[WeightCtx] Loaded 436 tensors, 1917.9 MB into backend
[DiT] Loaded: 36 layers, dim 2048, flash_attn=1
[Store] Load DiT: 198 ms
[GGUF] ../models/MiniMax-Music3-vocoder-F32.gguf: 182 tensors, data at offset 12896
[Load] VAE backend: CUDA0 (shared)
[VAE] Backend: CUDA0, Weight buffer: 103.4 MB
[VAE] Loaded: 4 blocks, upsample=512x, F32 activations
[Store] Load VAE: 110 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: 3.669599 -3.083768 -3.944149 4.045699
[Debug] hidden_after_preprocess: [689, 2304] first4: 0.341482 4.083567 -1.024049 1.790691
[Debug] hidden_after_proj_in: [689, 2048] first4: 5.213864 -4.338507 -6.016787 -1.824567
[Debug] layer0_sa_output: [690, 2048] first4: 1.649719 -1.855248 0.473579 1.612615
[Debug] hidden_after_layer0: [690, 2048] first4: -1.930172 -1.316381 -3.114213 3.862382
[Debug] hidden_after_layer6: [690, 2048] first4: -3.988987 -0.146251 0.499511 2.006353
[Debug] hidden_after_layer12: [690, 2048] first4: -2.125443 -1.637329 1.712717 3.549122
[Debug] hidden_after_layer18: [690, 2048] first4: -2.451432 -2.217984 -0.637439 1.664280
[Debug] hidden_after_layer35: [690, 2048] first4: -1.509831 -0.621866 -3.126137 -0.091052
[Debug] dit_step0_vt_cond: [689, 128] first4: -0.552450 -1.869192 -0.548412 -0.343643
[Debug] dit_step0_vt_uncond: [689, 128] first4: -0.317998 -2.270719 -0.453220 -0.595697
[Debug] dit_step0_vt: [689, 128] first4: -0.716566 -1.588124 -0.615046 -0.167204
[Debug] dit_step0_xt: [689, 128] first4: 0.170133 2.108436 -0.192552 0.843487
[Debug] dit_step1_vt_cond: [689, 128] first4: -0.626264 -1.785261 -0.446131 -0.269678
[Debug] dit_step1_vt_uncond: [689, 128] first4: -0.387700 -1.839803 -0.355117 -0.657116
[Debug] dit_step1_vt: [689, 128] first4: -0.793258 -1.747082 -0.509840 0.001530
[Debug] dit_step1_xt: [689, 128] first4: 0.143691 2.050200 -0.209547 0.843538
[Debug] dit_step2_vt_cond: [689, 128] first4: -0.644941 -1.741796 -0.413741 -0.117883
[Debug] dit_step2_vt_uncond: [689, 128] first4: -0.614322 -1.973492 -0.507117 -0.553317
[Debug] dit_step2_vt: [689, 128] first4: -0.666374 -1.579608 -0.348378 0.186921
[Debug] dit_step2_xt: [689, 128] first4: 0.121479 1.997547 -0.221159 0.849768
[Debug] dit_step3_vt_cond: [689, 128] first4: -0.675567 -1.652753 -0.354229 0.007296
[Debug] dit_step3_vt_uncond: [689, 128] first4: -0.201396 -1.873950 -0.545188 -0.242176
[Debug] dit_step3_vt: [689, 128] first4: -1.007486 -1.497915 -0.220558 0.181927
[Debug] dit_step3_xt: [689, 128] first4: 0.087896 1.947616 -0.228511 0.855833
[Debug] dit_step4_vt_cond: [689, 128] first4: -0.728932 -1.525487 -0.362937 0.237892
[Debug] dit_step4_vt_uncond: [689, 128] first4: -0.195456 -1.475816 -0.431917 0.070558
[Debug] dit_step4_vt: [689, 128] first4: -1.102365 -1.560257 -0.314651 0.355026
[Debug] dit_step4_xt: [689, 128] first4: 0.051151 1.895608 -0.239000 0.867667
[Debug] dit_step5_vt_cond: [689, 128] first4: -0.768290 -1.378217 -0.318106 0.436026
[Debug] dit_step5_vt_uncond: [689, 128] first4: -0.340211 -1.309622 -0.473723 0.156722
[Debug] dit_step5_vt: [689, 128] first4: -1.067945 -1.426233 -0.209175 0.631539
[Debug] dit_step5_xt: [689, 128] first4: 0.015552 1.848066 -0.245972 0.888718
[Debug] dit_step6_vt_cond: [689, 128] first4: -0.816318 -1.227134 -0.254824 0.630461
[Debug] dit_step6_vt_uncond: [689, 128] first4: -0.417623 -1.299827 -0.684665 0.513208
[Debug] dit_step6_vt: [689, 128] first4: -1.095405 -1.176249 0.046065 0.712539
[Debug] dit_step6_xt: [689, 128] first4: -0.020961 1.808858 -0.244437 0.912469
[Debug] dit_step7_vt_cond: [689, 128] first4: -0.885737 -1.035193 -0.219223 0.821847
[Debug] dit_step7_vt_uncond: [689, 128] first4: -0.476044 -1.224182 -0.827181 0.872967
[Debug] dit_step7_vt: [689, 128] first4: -1.172522 -0.902901 0.206348 0.786064
[Debug] dit_step7_xt: [689, 128] first4: -0.060045 1.778762 -0.237558 0.938671
[Debug] dit_step8_vt_cond: [689, 128] first4: -0.973357 -0.839315 -0.171714 0.943764
[Debug] dit_step8_vt_uncond: [689, 128] first4: -0.596211 -0.959337 -0.624549 1.077352
[Debug] dit_step8_vt: [689, 128] first4: -1.237358 -0.755300 0.145270 0.850253
[Debug] dit_step8_xt: [689, 128] first4: -0.101290 1.753585 -0.232716 0.967013
[Debug] dit_step9_vt_cond: [689, 128] first4: -0.994998 -0.602157 -0.152540 1.118821
[Debug] dit_step9_vt_uncond: [689, 128] first4: -0.664822 -0.746920 -0.382141 1.223876
[Debug] dit_step9_vt: [689, 128] first4: -1.226122 -0.500824 0.008180 1.045282
[Debug] dit_step9_xt: [689, 128] first4: -0.142161 1.736891 -0.232443 1.001856
[Debug] dit_step10_vt_cond: [689, 128] first4: -0.997056 -0.387540 -0.172492 1.282439
[Debug] dit_step10_vt_uncond: [689, 128] first4: -0.790595 -0.528126 -0.237302 1.344292
[Debug] dit_step10_vt: [689, 128] first4: -1.141578 -0.289129 -0.127125 1.239142
[Debug] dit_step10_xt: [689, 128] first4: -0.180214 1.727253 -0.236681 1.043161
[Debug] dit_step11_vt_cond: [689, 128] first4: -1.001053 -0.148552 -0.176695 1.416358
[Debug] dit_step11_vt_uncond: [689, 128] first4: -0.852680 -0.347819 -0.135213 1.468108
[Debug] dit_step11_vt: [689, 128] first4: -1.104914 -0.009065 -0.205733 1.380133
[Debug] dit_step11_xt: [689, 128] first4: -0.217044 1.726951 -0.243539 1.089165
[Debug] dit_step12_vt_cond: [689, 128] first4: -0.963613 0.071315 -0.217645 1.585715
[Debug] dit_step12_vt_uncond: [689, 128] first4: -0.897861 -0.106945 -0.074975 1.593587
[Debug] dit_step12_vt: [689, 128] first4: -1.009638 0.196098 -0.317513 1.580204
[Debug] dit_step12_xt: [689, 128] first4: -0.250699 1.733487 -0.254122 1.141839
[Debug] dit_step13_vt_cond: [689, 128] first4: -0.994768 0.287183 -0.216154 1.727730
[Debug] dit_step13_vt_uncond: [689, 128] first4: -0.917938 0.139723 -0.086898 1.711914
[Debug] dit_step13_vt: [689, 128] first4: -1.048548 0.390405 -0.306633 1.738802
[Debug] dit_step13_xt: [689, 128] first4: -0.285650 1.746501 -0.264344 1.199799
[Debug] dit_step14_vt_cond: [689, 128] first4: -0.954422 0.476453 -0.220675 1.839005
[Debug] dit_step14_vt_uncond: [689, 128] first4: -0.920012 0.370986 -0.089116 1.778936
[Debug] dit_step14_vt: [689, 128] first4: -0.978508 0.550281 -0.312766 1.881054
[Debug] dit_step14_xt: [689, 128] first4: -0.318267 1.764844 -0.274769 1.262501
[Debug] dit_step15_vt_cond: [689, 128] first4: -0.908813 0.670382 -0.195800 1.945107
[Debug] dit_step15_vt_uncond: [689, 128] first4: -0.878525 0.628710 -0.106494 1.888945
[Debug] dit_step15_vt: [689, 128] first4: -0.930015 0.699553 -0.258313 1.984420
[Debug] dit_step15_xt: [689, 128] first4: -0.349268 1.788162 -0.283380 1.328648
[Debug] dit_step16_vt_cond: [689, 128] first4: -0.905340 0.850709 -0.178725 1.989588
[Debug] dit_step16_vt_uncond: [689, 128] first4: -0.883172 0.858985 -0.100781 1.947101
[Debug] dit_step16_vt: [689, 128] first4: -0.920857 0.844915 -0.233285 2.019329
[Debug] dit_step16_xt: [689, 128] first4: -0.379963 1.816326 -0.291156 1.395959
[Debug] dit_step17_vt_cond: [689, 128] first4: -0.806209 1.047462 -0.201343 2.073696
[Debug] dit_step17_vt_uncond: [689, 128] first4: -0.819113 1.067783 -0.133894 2.018428
[Debug] dit_step17_vt: [689, 128] first4: -0.797176 1.033236 -0.248557 2.112384
[Debug] dit_step17_xt: [689, 128] first4: -0.406536 1.850767 -0.299441 1.466372
[Debug] dit_step18_vt_cond: [689, 128] first4: -0.786276 1.222453 -0.195032 2.124341
[Debug] dit_step18_vt_uncond: [689, 128] first4: -0.803370 1.265615 -0.093096 2.062469
[Debug] dit_step18_vt: [689, 128] first4: -0.774311 1.192240 -0.266386 2.167651
[Debug] dit_step18_xt: [689, 128] first4: -0.432346 1.890509 -0.308320 1.538627
[Debug] dit_step19_vt_cond: [689, 128] first4: -0.746320 1.404059 -0.221681 2.190483
[Debug] dit_step19_vt_uncond: [689, 128] first4: -0.772300 1.415773 -0.082144 2.103703
[Debug] dit_step19_vt: [689, 128] first4: -0.728134 1.395860 -0.319356 2.251230
[Debug] dit_step19_xt: [689, 128] first4: -0.456617 1.937037 -0.318966 1.613668
[Debug] dit_step20_vt_cond: [689, 128] first4: -0.673041 1.530769 -0.202325 2.220711
[Debug] dit_step20_vt_uncond: [689, 128] first4: -0.740136 1.588733 -0.088098 2.112392
[Debug] dit_step20_vt: [689, 128] first4: -0.626074 1.490195 -0.282284 2.296534
[Debug] dit_step20_xt: [689, 128] first4: -0.477486 1.986710 -0.328375 1.690219
[Debug] dit_step21_vt_cond: [689, 128] first4: -0.659917 1.664956 -0.222575 2.235505
[Debug] dit_step21_vt_uncond: [689, 128] first4: -0.622951 1.716570 -0.129909 2.170676
[Debug] dit_step21_vt: [689, 128] first4: -0.685793 1.628826 -0.287441 2.280885
[Debug] dit_step21_xt: [689, 128] first4: -0.500346 2.041005 -0.337957 1.766248
[Debug] dit_step22_vt_cond: [689, 128] first4: -0.622696 1.800226 -0.240113 2.247842
[Debug] dit_step22_vt_uncond: [689, 128] first4: -0.631829 1.819317 -0.133128 2.162844
[Debug] dit_step22_vt: [689, 128] first4: -0.616304 1.786863 -0.315003 2.307341
[Debug] dit_step22_xt: [689, 128] first4: -0.520890 2.100567 -0.348457 1.843160
[Debug] dit_step23_vt_cond: [689, 128] first4: -0.588948 1.936145 -0.249065 2.255521
[Debug] dit_step23_vt_uncond: [689, 128] first4: -0.585005 1.962578 -0.156401 2.196611
[Debug] dit_step23_vt: [689, 128] first4: -0.591707 1.917642 -0.313931 2.296757
[Debug] dit_step23_xt: [689, 128] first4: -0.540613 2.164488 -0.358921 1.919718
[Debug] dit_step24_vt_cond: [689, 128] first4: -0.551190 2.045228 -0.275306 2.275240
[Debug] dit_step24_vt_uncond: [689, 128] first4: -0.556503 2.066789 -0.161438 2.167285
[Debug] dit_step24_vt: [689, 128] first4: -0.547472 2.030136 -0.355014 2.350808
[Debug] dit_step24_xt: [689, 128] first4: -0.558862 2.232159 -0.370755 1.998079
[Debug] dit_step25_vt_cond: [689, 128] first4: -0.583280 2.203130 -0.299827 2.293192
[Debug] dit_step25_vt_uncond: [689, 128] first4: -0.554221 2.205715 -0.205401 2.203520
[Debug] dit_step25_vt: [689, 128] first4: -0.603621 2.201320 -0.365925 2.355962
[Debug] dit_step25_xt: [689, 128] first4: -0.578983 2.305537 -0.382952 2.076611
[Debug] dit_step26_vt_cond: [689, 128] first4: -0.575561 2.301609 -0.338514 2.323679
[Debug] dit_step26_vt_uncond: [689, 128] first4: -0.572837 2.281317 -0.277718 2.235267
[Debug] dit_step26_vt: [689, 128] first4: -0.577467 2.315814 -0.381071 2.385568
[Debug] dit_step26_xt: [689, 128] first4: -0.598232 2.382730 -0.395655 2.156130
[Debug] dit_step27_vt_cond: [689, 128] first4: -0.610298 2.421837 -0.385330 2.340446
[Debug] dit_step27_vt_uncond: [689, 128] first4: -0.583893 2.385973 -0.345566 2.275722
[Debug] dit_step27_vt: [689, 128] first4: -0.628782 2.446942 -0.413165 2.385754
[Debug] dit_step27_xt: [689, 128] first4: -0.619191 2.464295 -0.409427 2.235655
[Debug] dit_step28_vt_cond: [689, 128] first4: -0.618557 2.526690 -0.443659 2.328060
[Debug] dit_step28_vt_uncond: [689, 128] first4: -0.609120 2.468060 -0.428395 2.290226
[Debug] dit_step28_vt: [689, 128] first4: -0.625163 2.567730 -0.454344 2.354544
[Debug] dit_step28_xt: [689, 128] first4: -0.640030 2.549886 -0.424572 2.314139
[Debug] dit_step29_vt_cond: [689, 128] first4: -0.641271 2.575329 -0.533827 2.329322
[Debug] dit_step29_vt_uncond: [689, 128] first4: -0.621008 2.560093 -0.523252 2.270237
[Debug] dit_step29_vt: [689, 128] first4: -0.655454 2.585994 -0.541230 2.370681
[Debug] dit_step29_xt: [689, 128] first4: -0.661879 2.636086 -0.442613 2.393162
[Debug] dit_x0: [689, 128] first4: -0.661879 2.636086 -0.442613 2.393162
[Debug] window0_cond: [689, 2048] first4: -0.290057 -0.003712 -0.019914 -0.070771
[Debug] window0_latent: [689, 128] first4: -0.661879 2.636086 -0.442613 2.393162
[DiT] Window 1/1: T=689, 30 steps, 2629 ms (87.6 ms/step)
[DiT] CFG=1.70, 1 windows, 2.6 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: -0.029540 -0.006614 -0.031705 0.003566
[VAE] Decode: 1 windows -> 8.0s of audio, 85 ms
[Done] 8.8 s total
[Store] Unload VAE (103.4 MB)
[Store] Unload DiT (1917.9 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-Q6_K.gguf
[GGML] Running MiniMax-Music3-transformer-Q6_K.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 0.999998
hidden_after_preprocess: ggml vs python 1.000000
hidden_after_proj_in: ggml vs python 0.999936
layer0_sa_output: ggml vs python 0.999985
hidden_after_layer0: ggml vs python 0.999986
hidden_after_layer6: ggml vs python 0.999993
hidden_after_layer12: ggml vs python 0.999989
hidden_after_layer18: ggml vs python 0.999987
hidden_after_layer35: ggml vs python 0.999960
dit_step0_vt_cond: ggml vs python 0.999588
dit_step0_vt_uncond: ggml vs python 0.999533
dit_step0_vt: ggml vs python 0.999622
dit_step0_xt: ggml vs python 0.999999
dit_step5_vt_cond: ggml vs python 0.999263
dit_step5_vt: ggml vs python 0.998962
dit_step5_xt: ggml vs python 0.999935
dit_step10_vt_cond: ggml vs python 0.998958
dit_step10_vt: ggml vs python 0.998651
dit_step10_xt: ggml vs python 0.999727
dit_step15_vt_cond: ggml vs python 0.998827
dit_step15_vt: ggml vs python 0.998593
dit_step15_xt: ggml vs python 0.999506
dit_step20_vt_cond: ggml vs python 0.998640
dit_step20_vt: ggml vs python 0.998464
dit_step20_xt: ggml vs python 0.999354
dit_step25_vt_cond: ggml vs python 0.998610
dit_step25_vt: ggml vs python 0.998483
dit_step25_xt: ggml vs python 0.999255
dit_step29_vt_cond: ggml vs python 0.998722
dit_step29_vt: ggml vs python 0.998623
dit_x0: ggml vs python 0.999206
vae_audio: ggml vs python 0.997531
vae_audio (STFT cosine): ggml vs python 0.998845
[DiT] Error growth GGML vs Python
dit_step0_xt: cos 0.999999, max_err 0.008337, mean_err 0.001149, mean_A -0.002030, std_A 0.965046, mean_B -0.002135, std_B 0.965041
dit_step5_xt: cos 0.999935, max_err 0.095723, mean_err 0.007673, mean_A 0.002991, std_A 0.885171, mean_B 0.002533, std_B 0.884593
dit_step10_xt: cos 0.999727, max_err 0.292280, mean_err 0.016502, mean_A 0.008792, std_A 0.952495, mean_B 0.007973, std_B 0.950289
dit_step15_xt: cos 0.999506, max_err 0.598045, mean_err 0.026247, mean_A 0.014815, std_A 1.137001, mean_B 0.013735, std_B 1.132820
dit_step20_xt: cos 0.999354, max_err 0.842535, mean_err 0.036800, mean_A 0.020878, std_A 1.392115, mean_B 0.019518, std_B 1.385859
dit_step25_xt: cos 0.999255, max_err 1.041279, mean_err 0.048136, mean_A 0.026660, std_A 1.685855, mean_B 0.025057, std_B 1.677655