[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: 1383 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 71 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 (21.0 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: 48 ms
[GGUF] ../models/MiniMax-Music3-transformer-F32.gguf: 441 tensors, data at offset 31616
[Load] DiT backend: CUDA0 (shared)
[WeightCtx] Loaded 436 tensors, 9259.0 MB into backend
[DiT] Loaded: 36 layers, dim 2048, flash_attn=1
[Store] Load DiT: 675 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: 105 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: 3.623775 -3.071606 -3.963248 4.064687
[Debug] hidden_after_preprocess: [689, 2304] first4: 0.341482 4.083567 -1.024049 1.790691
[Debug] hidden_after_proj_in: [689, 2048] first4: 5.232117 -4.336130 -5.957681 -1.879564
[Debug] layer0_sa_output: [690, 2048] first4: 1.653646 -1.851118 0.385252 1.528639
[Debug] hidden_after_layer0: [690, 2048] first4: -2.038190 -1.290454 -3.253103 3.755661
[Debug] hidden_after_layer6: [690, 2048] first4: -4.124695 -0.261449 0.298013 1.999205
[Debug] hidden_after_layer12: [690, 2048] first4: -2.181556 -1.738128 1.567802 3.540758
[Debug] hidden_after_layer18: [690, 2048] first4: -2.424420 -2.252856 -0.846813 1.654849
[Debug] hidden_after_layer35: [690, 2048] first4: -1.418122 -0.296339 -3.227356 -0.374267
[Debug] dit_step0_vt_cond: [689, 128] first4: -0.608823 -1.883970 -0.529756 -0.392248
[Debug] dit_step0_vt_uncond: [689, 128] first4: -0.357136 -2.273930 -0.432869 -0.607793
[Debug] dit_step0_vt: [689, 128] first4: -0.785003 -1.610999 -0.597577 -0.241367
[Debug] dit_step0_xt: [689, 128] first4: 0.167852 2.107674 -0.191970 0.841015
[Debug] dit_step1_vt_cond: [689, 128] first4: -0.655181 -1.791591 -0.424369 -0.303538
[Debug] dit_step1_vt_uncond: [689, 128] first4: -0.445225 -1.847773 -0.312963 -0.673379
[Debug] dit_step1_vt: [689, 128] first4: -0.802151 -1.752263 -0.502354 -0.044650
[Debug] dit_step1_xt: [689, 128] first4: 0.141114 2.049265 -0.208715 0.839526
[Debug] dit_step2_vt_cond: [689, 128] first4: -0.688924 -1.736101 -0.392709 -0.155919
[Debug] dit_step2_vt_uncond: [689, 128] first4: -0.733686 -1.989172 -0.489992 -0.609130
[Debug] dit_step2_vt: [689, 128] first4: -0.657591 -1.558952 -0.324611 0.161329
[Debug] dit_step2_xt: [689, 128] first4: 0.119194 1.997300 -0.219535 0.844904
[Debug] dit_step3_vt_cond: [689, 128] first4: -0.714650 -1.657816 -0.353020 0.017290
[Debug] dit_step3_vt_uncond: [689, 128] first4: -0.311367 -1.875381 -0.518005 -0.268326
[Debug] dit_step3_vt: [689, 128] first4: -0.996949 -1.505520 -0.237531 0.217221
[Debug] dit_step3_xt: [689, 128] first4: 0.085962 1.947116 -0.227453 0.852145
[Debug] dit_step4_vt_cond: [689, 128] first4: -0.760777 -1.524820 -0.334847 0.222214
[Debug] dit_step4_vt_uncond: [689, 128] first4: -0.277488 -1.489908 -0.415575 0.069826
[Debug] dit_step4_vt: [689, 128] first4: -1.099078 -1.549259 -0.278338 0.328885
[Debug] dit_step4_xt: [689, 128] first4: 0.049327 1.895474 -0.236731 0.863107
[Debug] dit_step5_vt_cond: [689, 128] first4: -0.820312 -1.385038 -0.298030 0.417855
[Debug] dit_step5_vt_uncond: [689, 128] first4: -0.389199 -1.305017 -0.463630 0.152670
[Debug] dit_step5_vt: [689, 128] first4: -1.122091 -1.441053 -0.182110 0.603485
[Debug] dit_step5_xt: [689, 128] first4: 0.011923 1.847439 -0.242801 0.883224
[Debug] dit_step6_vt_cond: [689, 128] first4: -0.890532 -1.232958 -0.250340 0.614621
[Debug] dit_step6_vt_uncond: [689, 128] first4: -0.469062 -1.315527 -0.632056 0.502121
[Debug] dit_step6_vt: [689, 128] first4: -1.185562 -1.175160 0.016860 0.693371
[Debug] dit_step6_xt: [689, 128] first4: -0.027595 1.808267 -0.242239 0.906336
[Debug] dit_step7_vt_cond: [689, 128] first4: -0.958613 -1.046672 -0.196916 0.798892
[Debug] dit_step7_vt_uncond: [689, 128] first4: -0.578704 -1.224388 -0.788433 0.876834
[Debug] dit_step7_vt: [689, 128] first4: -1.224549 -0.922272 0.217145 0.744332
[Debug] dit_step7_xt: [689, 128] first4: -0.068414 1.777524 -0.235001 0.931147
[Debug] dit_step8_vt_cond: [689, 128] first4: -1.012681 -0.843172 -0.150924 0.961908
[Debug] dit_step8_vt_uncond: [689, 128] first4: -0.688788 -0.985857 -0.645691 1.118522
[Debug] dit_step8_vt: [689, 128] first4: -1.239406 -0.743291 0.195412 0.852277
[Debug] dit_step8_xt: [689, 128] first4: -0.109727 1.752748 -0.228487 0.959556
[Debug] dit_step9_vt_cond: [689, 128] first4: -1.050610 -0.629035 -0.124843 1.116230
[Debug] dit_step9_vt_uncond: [689, 128] first4: -0.778833 -0.775113 -0.398144 1.276709
[Debug] dit_step9_vt: [689, 128] first4: -1.240853 -0.526781 0.066467 1.003894
[Debug] dit_step9_xt: [689, 128] first4: -0.151089 1.735189 -0.226272 0.993019
[Debug] dit_step10_vt_cond: [689, 128] first4: -1.069182 -0.401572 -0.121762 1.271067
[Debug] dit_step10_vt_uncond: [689, 128] first4: -0.864966 -0.569788 -0.243404 1.421476
[Debug] dit_step10_vt: [689, 128] first4: -1.212133 -0.283821 -0.036612 1.165781
[Debug] dit_step10_xt: [689, 128] first4: -0.191493 1.725728 -0.227492 1.031879
[Debug] dit_step11_vt_cond: [689, 128] first4: -1.072961 -0.174176 -0.138677 1.427288
[Debug] dit_step11_vt_uncond: [689, 128] first4: -0.930071 -0.361018 -0.139194 1.553831
[Debug] dit_step11_vt: [689, 128] first4: -1.172985 -0.043387 -0.138316 1.338708
[Debug] dit_step11_xt: [689, 128] first4: -0.230593 1.724282 -0.232103 1.076502
[Debug] dit_step12_vt_cond: [689, 128] first4: -1.070170 0.047007 -0.162461 1.590281
[Debug] dit_step12_vt_uncond: [689, 128] first4: -0.975616 -0.131881 -0.074206 1.658635
[Debug] dit_step12_vt: [689, 128] first4: -1.136358 0.172228 -0.224240 1.542433
[Debug] dit_step12_xt: [689, 128] first4: -0.268471 1.730023 -0.239577 1.127917
[Debug] dit_step13_vt_cond: [689, 128] first4: -1.058271 0.250810 -0.164808 1.745533
[Debug] dit_step13_vt_uncond: [689, 128] first4: -1.006599 0.112671 -0.045902 1.751181
[Debug] dit_step13_vt: [689, 128] first4: -1.094442 0.347508 -0.248042 1.741579
[Debug] dit_step13_xt: [689, 128] first4: -0.304953 1.741606 -0.247845 1.185969
[Debug] dit_step14_vt_cond: [689, 128] first4: -1.027017 0.453218 -0.159213 1.857556
[Debug] dit_step14_vt_uncond: [689, 128] first4: -1.019830 0.360510 -0.048442 1.840388
[Debug] dit_step14_vt: [689, 128] first4: -1.032047 0.518113 -0.236753 1.869574
[Debug] dit_step14_xt: [689, 128] first4: -0.339354 1.758877 -0.255737 1.248289
[Debug] dit_step15_vt_cond: [689, 128] first4: -0.993572 0.651662 -0.156314 1.951779
[Debug] dit_step15_vt_uncond: [689, 128] first4: -1.010006 0.602862 -0.057847 1.921735
[Debug] dit_step15_vt: [689, 128] first4: -0.982067 0.685822 -0.225241 1.972811
[Debug] dit_step15_xt: [689, 128] first4: -0.372090 1.781738 -0.263245 1.314049
[Debug] dit_step16_vt_cond: [689, 128] first4: -0.956465 0.841392 -0.155033 2.039734
[Debug] dit_step16_vt_uncond: [689, 128] first4: -0.986521 0.828547 -0.062920 1.992805
[Debug] dit_step16_vt: [689, 128] first4: -0.935426 0.850383 -0.219512 2.072584
[Debug] dit_step16_xt: [689, 128] first4: -0.403271 1.810084 -0.270562 1.383135
[Debug] dit_step17_vt_cond: [689, 128] first4: -0.914512 1.020183 -0.151478 2.109023
[Debug] dit_step17_vt_uncond: [689, 128] first4: -0.948756 1.038449 -0.059973 2.051810
[Debug] dit_step17_vt: [689, 128] first4: -0.890542 1.007396 -0.215531 2.149072
[Debug] dit_step17_xt: [689, 128] first4: -0.432956 1.843663 -0.277747 1.454771
[Debug] dit_step18_vt_cond: [689, 128] first4: -0.862566 1.180396 -0.151631 2.168334
[Debug] dit_step18_vt_uncond: [689, 128] first4: -0.900106 1.222403 -0.055514 2.099019
[Debug] dit_step18_vt: [689, 128] first4: -0.836289 1.150992 -0.218913 2.216855
[Debug] dit_step18_xt: [689, 128] first4: -0.460832 1.882030 -0.285044 1.528666
[Debug] dit_step19_vt_cond: [689, 128] first4: -0.803991 1.333465 -0.152772 2.220012
[Debug] dit_step19_vt_uncond: [689, 128] first4: -0.843804 1.381734 -0.051089 2.134067
[Debug] dit_step19_vt: [689, 128] first4: -0.776123 1.299676 -0.223950 2.280173
[Debug] dit_step19_xt: [689, 128] first4: -0.486703 1.925352 -0.292509 1.604672
[Debug] dit_step20_vt_cond: [689, 128] first4: -0.739595 1.472323 -0.160349 2.259891
[Debug] dit_step20_vt_uncond: [689, 128] first4: -0.785327 1.529465 -0.052460 2.164472
[Debug] dit_step20_vt: [689, 128] first4: -0.707583 1.432325 -0.235871 2.326684
[Debug] dit_step20_xt: [689, 128] first4: -0.510289 1.973096 -0.300371 1.682228
[Debug] dit_step21_vt_cond: [689, 128] first4: -0.679376 1.606888 -0.171218 2.285494
[Debug] dit_step21_vt_uncond: [689, 128] first4: -0.722700 1.655012 -0.057522 2.183173
[Debug] dit_step21_vt: [689, 128] first4: -0.649049 1.573201 -0.250806 2.357118
[Debug] dit_step21_xt: [689, 128] first4: -0.531924 2.025537 -0.308731 1.760799
[Debug] dit_step22_vt_cond: [689, 128] first4: -0.630460 1.734258 -0.188560 2.307584
[Debug] dit_step22_vt_uncond: [689, 128] first4: -0.669991 1.785223 -0.070801 2.198204
[Debug] dit_step22_vt: [689, 128] first4: -0.602788 1.698582 -0.270990 2.384150
[Debug] dit_step22_xt: [689, 128] first4: -0.552017 2.082156 -0.317764 1.840270
[Debug] dit_step23_vt_cond: [689, 128] first4: -0.599854 1.869156 -0.203529 2.323054
[Debug] dit_step23_vt_uncond: [689, 128] first4: -0.631600 1.898766 -0.089291 2.215920
[Debug] dit_step23_vt: [689, 128] first4: -0.577632 1.848429 -0.283496 2.398048
[Debug] dit_step23_xt: [689, 128] first4: -0.571271 2.143770 -0.327214 1.920205
[Debug] dit_step24_vt_cond: [689, 128] first4: -0.592054 2.007754 -0.222956 2.334580
[Debug] dit_step24_vt_uncond: [689, 128] first4: -0.610900 2.029048 -0.119060 2.235086
[Debug] dit_step24_vt: [689, 128] first4: -0.578861 1.992849 -0.295684 2.404226
[Debug] dit_step24_xt: [689, 128] first4: -0.590566 2.210199 -0.337070 2.000346
[Debug] dit_step25_vt_cond: [689, 128] first4: -0.601179 2.136873 -0.246367 2.354670
[Debug] dit_step25_vt_uncond: [689, 128] first4: -0.609729 2.131485 -0.160178 2.259896
[Debug] dit_step25_vt: [689, 128] first4: -0.595193 2.140645 -0.306700 2.421011
[Debug] dit_step25_xt: [689, 128] first4: -0.610406 2.281554 -0.347294 2.081046
[Debug] dit_step26_vt_cond: [689, 128] first4: -0.624620 2.264521 -0.283070 2.371584
[Debug] dit_step26_vt_uncond: [689, 128] first4: -0.621158 2.245090 -0.215495 2.288057
[Debug] dit_step26_vt: [689, 128] first4: -0.627043 2.278122 -0.330373 2.430053
[Debug] dit_step26_xt: [689, 128] first4: -0.631308 2.357491 -0.358306 2.162048
[Debug] dit_step27_vt_cond: [689, 128] first4: -0.654016 2.391615 -0.340114 2.383453
[Debug] dit_step27_vt_uncond: [689, 128] first4: -0.642187 2.359852 -0.289932 2.321701
[Debug] dit_step27_vt: [689, 128] first4: -0.662297 2.413849 -0.375241 2.426680
[Debug] dit_step27_xt: [689, 128] first4: -0.653384 2.437952 -0.370814 2.242937
[Debug] dit_step28_vt_cond: [689, 128] first4: -0.681170 2.493784 -0.399420 2.373285
[Debug] dit_step28_vt_uncond: [689, 128] first4: -0.665895 2.460426 -0.364910 2.327788
[Debug] dit_step28_vt: [689, 128] first4: -0.691862 2.517134 -0.423578 2.405133
[Debug] dit_step28_xt: [689, 128] first4: -0.676446 2.521857 -0.384933 2.323108
[Debug] dit_step29_vt_cond: [689, 128] first4: -0.708858 2.576212 -0.457819 2.343248
[Debug] dit_step29_vt_uncond: [689, 128] first4: -0.684115 2.544058 -0.438635 2.310514
[Debug] dit_step29_vt: [689, 128] first4: -0.726177 2.598720 -0.471248 2.366162
[Debug] dit_step29_xt: [689, 128] first4: -0.700652 2.608481 -0.400642 2.401981
[Debug] dit_x0: [689, 128] first4: -0.700652 2.608481 -0.400642 2.401981
[Debug] window0_cond: [689, 2048] first4: -0.290057 -0.003712 -0.019914 -0.070771
[Debug] window0_latent: [689, 128] first4: -0.700652 2.608481 -0.400642 2.401981
[DiT] Window 1/1: T=689, 30 steps, 3122 ms (104.1 ms/step)
[DiT] CFG=1.70, 1 windows, 3.1 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: -0.047609 -0.005859 -0.031278 0.004784
[VAE] Decode: 1 windows -> 8.0s of audio, 85 ms
[Done] 9.8 s total
[Store] Unload VAE (103.4 MB)
[Store] Unload DiT (9259.0 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-F32.gguf
[GGML] Running MiniMax-Music3-transformer-F32.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 1.000000
layer0_sa_output: ggml vs python 1.000000
hidden_after_layer0: ggml vs python 0.999999
hidden_after_layer6: ggml vs python 1.000001
hidden_after_layer12: ggml vs python 0.999999
hidden_after_layer18: ggml vs python 1.000000
hidden_after_layer35: ggml vs python 1.000000
dit_step0_vt_cond: ggml vs python 0.999999
dit_step0_vt_uncond: ggml vs python 0.999999
dit_step0_vt: ggml vs python 0.999998
dit_step0_xt: ggml vs python 1.000000
dit_step5_vt_cond: ggml vs python 0.999999
dit_step5_vt: ggml vs python 0.999997
dit_step5_xt: ggml vs python 1.000000
dit_step10_vt_cond: ggml vs python 0.999998
dit_step10_vt: ggml vs python 0.999996
dit_step10_xt: ggml vs python 1.000000
dit_step15_vt_cond: ggml vs python 0.999998
dit_step15_vt: ggml vs python 0.999996
dit_step15_xt: ggml vs python 1.000000
dit_step20_vt_cond: ggml vs python 0.999998
dit_step20_vt: ggml vs python 0.999995
dit_step20_xt: ggml vs python 1.000000
dit_step25_vt_cond: ggml vs python 0.999998
dit_step25_vt: ggml vs python 0.999995
dit_step25_xt: ggml vs python 1.000000
dit_step29_vt_cond: ggml vs python 0.999999
dit_step29_vt: ggml vs python 0.999995
dit_x0: ggml vs python 1.000000
vae_audio: ggml vs python 0.999997
vae_audio (STFT cosine): ggml vs python 0.999998
[DiT] Error growth GGML vs Python
dit_step0_xt: cos 1.000000, max_err 0.001363, mean_err 0.000076, mean_A -0.002134, std_A 0.965041, mean_B -0.002135, std_B 0.965041
dit_step5_xt: cos 1.000000, max_err 0.003739, mean_err 0.000242, mean_A 0.002534, std_A 0.884601, mean_B 0.002533, std_B 0.884593
dit_step10_xt: cos 1.000000, max_err 0.005110, mean_err 0.000395, mean_A 0.007977, std_A 0.950303, mean_B 0.007973, std_B 0.950289
dit_step15_xt: cos 1.000000, max_err 0.006250, mean_err 0.000557, mean_A 0.013742, std_A 1.132829, mean_B 0.013735, std_B 1.132820
dit_step20_xt: cos 1.000000, max_err 0.008628, mean_err 0.000741, mean_A 0.019527, std_A 1.385866, mean_B 0.019518, std_B 1.385859
dit_step25_xt: cos 1.000000, max_err 0.015975, mean_err 0.000953, mean_A 0.025070, std_A 1.677655, mean_B 0.025057, std_B 1.677655