[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: 1353 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 (20.9 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-Q5_K_M.gguf: 441 tensors, data at offset 31712
[Load] DiT backend: CUDA0 (shared)
[WeightCtx] Loaded 436 tensors, 1611.2 MB into backend
[DiT] Loaded: 36 layers, dim 2048, flash_attn=1
[Store] Load DiT: 136 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: 102 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: 3.715232 -3.035194 -3.836430 4.005782
[Debug] hidden_after_preprocess: [689, 2304] first4: 0.341482 4.083567 -1.024049 1.790691
[Debug] hidden_after_proj_in: [689, 2048] first4: 5.381108 -4.273623 -6.029389 -1.923309
[Debug] layer0_sa_output: [690, 2048] first4: 1.510196 -1.898746 0.511049 1.581268
[Debug] hidden_after_layer0: [690, 2048] first4: -1.910339 -1.241704 -2.974151 3.819715
[Debug] hidden_after_layer6: [690, 2048] first4: -4.044596 -0.117340 0.638660 2.105507
[Debug] hidden_after_layer12: [690, 2048] first4: -2.072731 -1.726063 1.732279 3.752161
[Debug] hidden_after_layer18: [690, 2048] first4: -2.306066 -2.079880 -0.576317 1.797633
[Debug] hidden_after_layer35: [690, 2048] first4: -1.280347 -0.278591 -2.930558 -0.066629
[Debug] dit_step0_vt_cond: [689, 128] first4: -0.587033 -1.877623 -0.574954 -0.329968
[Debug] dit_step0_vt_uncond: [689, 128] first4: -0.322930 -2.235872 -0.432661 -0.565320
[Debug] dit_step0_vt: [689, 128] first4: -0.771904 -1.626848 -0.674559 -0.165222
[Debug] dit_step0_xt: [689, 128] first4: 0.168289 2.107146 -0.194536 0.843553
[Debug] dit_step1_vt_cond: [689, 128] first4: -0.665176 -1.782290 -0.509494 -0.269705
[Debug] dit_step1_vt_uncond: [689, 128] first4: -0.409999 -1.846012 -0.359756 -0.667275
[Debug] dit_step1_vt: [689, 128] first4: -0.843800 -1.737684 -0.614310 0.008593
[Debug] dit_step1_xt: [689, 128] first4: 0.140162 2.049223 -0.215013 0.843839
[Debug] dit_step2_vt_cond: [689, 128] first4: -0.643545 -1.713459 -0.435970 -0.090623
[Debug] dit_step2_vt_uncond: [689, 128] first4: -0.785497 -2.138386 -0.558996 -0.688526
[Debug] dit_step2_vt: [689, 128] first4: -0.544179 -1.416010 -0.349852 0.327909
[Debug] dit_step2_xt: [689, 128] first4: 0.122023 2.002022 -0.226675 0.854769
[Debug] dit_step3_vt_cond: [689, 128] first4: -0.659461 -1.664456 -0.394921 0.066309
[Debug] dit_step3_vt_uncond: [689, 128] first4: -0.318666 -1.918632 -0.551187 -0.259949
[Debug] dit_step3_vt: [689, 128] first4: -0.898017 -1.486534 -0.285535 0.294690
[Debug] dit_step3_xt: [689, 128] first4: 0.092089 1.952471 -0.236192 0.864592
[Debug] dit_step4_vt_cond: [689, 128] first4: -0.716072 -1.461243 -0.319316 0.318580
[Debug] dit_step4_vt_uncond: [689, 128] first4: -0.267847 -1.511250 -0.465324 0.119223
[Debug] dit_step4_vt: [689, 128] first4: -1.029830 -1.426238 -0.217111 0.458131
[Debug] dit_step4_xt: [689, 128] first4: 0.057761 1.904930 -0.243430 0.879863
[Debug] dit_step5_vt_cond: [689, 128] first4: -0.832744 -1.393555 -0.348315 0.413933
[Debug] dit_step5_vt_uncond: [689, 128] first4: -0.336876 -1.280974 -0.437869 0.193339
[Debug] dit_step5_vt: [689, 128] first4: -1.179852 -1.472361 -0.285628 0.568348
[Debug] dit_step5_xt: [689, 128] first4: 0.018433 1.855851 -0.252950 0.898808
[Debug] dit_step6_vt_cond: [689, 128] first4: -0.862159 -1.203120 -0.268902 0.668401
[Debug] dit_step6_vt_uncond: [689, 128] first4: -0.417249 -1.299112 -0.674011 0.530152
[Debug] dit_step6_vt: [689, 128] first4: -1.173595 -1.135925 0.014673 0.765176
[Debug] dit_step6_xt: [689, 128] first4: -0.020687 1.817987 -0.252461 0.924314
[Debug] dit_step7_vt_cond: [689, 128] first4: -0.897414 -1.043426 -0.188468 0.856725
[Debug] dit_step7_vt_uncond: [689, 128] first4: -0.522018 -1.230849 -0.773294 0.896454
[Debug] dit_step7_vt: [689, 128] first4: -1.160191 -0.912229 0.220910 0.828914
[Debug] dit_step7_xt: [689, 128] first4: -0.059360 1.787579 -0.245098 0.951945
[Debug] dit_step8_vt_cond: [689, 128] first4: -0.938680 -0.845870 -0.086087 1.011822
[Debug] dit_step8_vt_uncond: [689, 128] first4: -0.635881 -1.013276 -0.668730 1.169125
[Debug] dit_step8_vt: [689, 128] first4: -1.150640 -0.728686 0.321763 0.901710
[Debug] dit_step8_xt: [689, 128] first4: -0.097715 1.763290 -0.234372 0.982002
[Debug] dit_step9_vt_cond: [689, 128] first4: -0.970952 -0.591795 -0.034270 1.185450
[Debug] dit_step9_vt_uncond: [689, 128] first4: -0.793685 -0.803986 -0.356986 1.366580
[Debug] dit_step9_vt: [689, 128] first4: -1.095038 -0.443261 0.191631 1.058659
[Debug] dit_step9_xt: [689, 128] first4: -0.134216 1.748514 -0.227985 1.017290
[Debug] dit_step10_vt_cond: [689, 128] first4: -1.038684 -0.354190 -0.049808 1.315176
[Debug] dit_step10_vt_uncond: [689, 128] first4: -0.865595 -0.596968 -0.275425 1.479196
[Debug] dit_step10_vt: [689, 128] first4: -1.159847 -0.184245 0.108124 1.200363
[Debug] dit_step10_xt: [689, 128] first4: -0.172878 1.742373 -0.224380 1.057302
[Debug] dit_step11_vt_cond: [689, 128] first4: -1.070505 -0.111908 -0.061492 1.454566
[Debug] dit_step11_vt_uncond: [689, 128] first4: -1.049062 -0.372850 -0.180834 1.603965
[Debug] dit_step11_vt: [689, 128] first4: -1.085515 0.070752 0.022047 1.349987
[Debug] dit_step11_xt: [689, 128] first4: -0.209061 1.744731 -0.223646 1.102302
[Debug] dit_step12_vt_cond: [689, 128] first4: -1.095430 0.037983 -0.148812 1.586237
[Debug] dit_step12_vt_uncond: [689, 128] first4: -0.927196 -0.091250 0.028036 1.778040
[Debug] dit_step12_vt: [689, 128] first4: -1.213194 0.128446 -0.272606 1.451975
[Debug] dit_step12_xt: [689, 128] first4: -0.249501 1.749013 -0.232732 1.150701
[Debug] dit_step13_vt_cond: [689, 128] first4: -1.107134 0.301946 -0.092412 1.735084
[Debug] dit_step13_vt_uncond: [689, 128] first4: -0.971837 0.103847 -0.012521 1.813718
[Debug] dit_step13_vt: [689, 128] first4: -1.201842 0.440615 -0.148335 1.680039
[Debug] dit_step13_xt: [689, 128] first4: -0.289563 1.763700 -0.237677 1.206702
[Debug] dit_step14_vt_cond: [689, 128] first4: -1.054282 0.442640 -0.090387 1.838585
[Debug] dit_step14_vt_uncond: [689, 128] first4: -1.069360 0.324730 -0.023474 1.856964
[Debug] dit_step14_vt: [689, 128] first4: -1.043728 0.525177 -0.137226 1.825721
[Debug] dit_step14_xt: [689, 128] first4: -0.324354 1.781206 -0.242251 1.267560
[Debug] dit_step15_vt_cond: [689, 128] first4: -1.133058 0.628273 -0.134066 1.936895
[Debug] dit_step15_vt_uncond: [689, 128] first4: -1.043040 0.604638 -0.048391 1.895597
[Debug] dit_step15_vt: [689, 128] first4: -1.196070 0.644817 -0.194038 1.965803
[Debug] dit_step15_xt: [689, 128] first4: -0.364223 1.802700 -0.248719 1.333087
[Debug] dit_step16_vt_cond: [689, 128] first4: -1.164255 0.772181 -0.210424 2.029953
[Debug] dit_step16_vt_uncond: [689, 128] first4: -1.068349 0.805390 -0.068238 1.976636
[Debug] dit_step16_vt: [689, 128] first4: -1.231388 0.748934 -0.309955 2.067275
[Debug] dit_step16_xt: [689, 128] first4: -0.405269 1.827664 -0.259051 1.401996
[Debug] dit_step17_vt_cond: [689, 128] first4: -1.074251 1.008342 -0.197977 2.107201
[Debug] dit_step17_vt_uncond: [689, 128] first4: -0.945657 1.018885 -0.030694 2.056606
[Debug] dit_step17_vt: [689, 128] first4: -1.164267 1.000962 -0.315075 2.142617
[Debug] dit_step17_xt: [689, 128] first4: -0.444078 1.861030 -0.269553 1.473416
[Debug] dit_step18_vt_cond: [689, 128] first4: -1.054214 1.190401 -0.101642 2.187708
[Debug] dit_step18_vt_uncond: [689, 128] first4: -0.942127 1.300114 0.041751 2.122754
[Debug] dit_step18_vt: [689, 128] first4: -1.132675 1.113602 -0.202016 2.233176
[Debug] dit_step18_xt: [689, 128] first4: -0.481834 1.898150 -0.276287 1.547856
[Debug] dit_step19_vt_cond: [689, 128] first4: -0.977686 1.318038 -0.226178 2.190871
[Debug] dit_step19_vt_uncond: [689, 128] first4: -0.851175 1.413242 -0.023044 2.188907
[Debug] dit_step19_vt: [689, 128] first4: -1.066244 1.251395 -0.368372 2.192246
[Debug] dit_step19_xt: [689, 128] first4: -0.517375 1.939863 -0.288566 1.620930
[Debug] dit_step20_vt_cond: [689, 128] first4: -0.950030 1.469963 -0.176881 2.262784
[Debug] dit_step20_vt_uncond: [689, 128] first4: -0.857834 1.581988 -0.019445 2.252644
[Debug] dit_step20_vt: [689, 128] first4: -1.014567 1.391545 -0.287087 2.269882
[Debug] dit_step20_xt: [689, 128] first4: -0.551194 1.986248 -0.298136 1.696593
[Debug] dit_step21_vt_cond: [689, 128] first4: -0.910333 1.638491 -0.174730 2.292962
[Debug] dit_step21_vt_uncond: [689, 128] first4: -0.843728 1.693350 -0.109596 2.178853
[Debug] dit_step21_vt: [689, 128] first4: -0.956957 1.600089 -0.220324 2.372838
[Debug] dit_step21_xt: [689, 128] first4: -0.583092 2.039584 -0.305480 1.775688
[Debug] dit_step22_vt_cond: [689, 128] first4: -0.789343 1.772145 -0.181588 2.322020
[Debug] dit_step22_vt_uncond: [689, 128] first4: -0.809193 1.818001 -0.104393 2.280091
[Debug] dit_step22_vt: [689, 128] first4: -0.775448 1.740046 -0.235625 2.351369
[Debug] dit_step22_xt: [689, 128] first4: -0.608941 2.097586 -0.313334 1.854067
[Debug] dit_step23_vt_cond: [689, 128] first4: -0.737341 1.898174 -0.273008 2.301960
[Debug] dit_step23_vt_uncond: [689, 128] first4: -0.877502 1.968359 -0.119688 2.237705
[Debug] dit_step23_vt: [689, 128] first4: -0.639229 1.849045 -0.380332 2.346938
[Debug] dit_step23_xt: [689, 128] first4: -0.630248 2.159220 -0.326012 1.932298
[Debug] dit_step24_vt_cond: [689, 128] first4: -0.746968 2.047130 -0.273708 2.379160
[Debug] dit_step24_vt_uncond: [689, 128] first4: -0.730740 2.035315 -0.196634 2.225999
[Debug] dit_step24_vt: [689, 128] first4: -0.758328 2.055400 -0.327661 2.486373
[Debug] dit_step24_xt: [689, 128] first4: -0.655526 2.227734 -0.336934 2.015177
[Debug] dit_step25_vt_cond: [689, 128] first4: -0.730352 2.137481 -0.332086 2.362785
[Debug] dit_step25_vt_uncond: [689, 128] first4: -0.717571 2.211388 -0.173997 2.295390
[Debug] dit_step25_vt: [689, 128] first4: -0.739298 2.085746 -0.442747 2.409962
[Debug] dit_step25_xt: [689, 128] first4: -0.680169 2.297259 -0.351692 2.095509
[Debug] dit_step26_vt_cond: [689, 128] first4: -0.823129 2.314955 -0.365102 2.445317
[Debug] dit_step26_vt_uncond: [689, 128] first4: -0.811401 2.279848 -0.257891 2.295869
[Debug] dit_step26_vt: [689, 128] first4: -0.831338 2.339530 -0.440150 2.549931
[Debug] dit_step26_xt: [689, 128] first4: -0.707880 2.375243 -0.366364 2.180507
[Debug] dit_step27_vt_cond: [689, 128] first4: -0.788166 2.428644 -0.436154 2.365030
[Debug] dit_step27_vt_uncond: [689, 128] first4: -0.832364 2.411505 -0.363371 2.389271
[Debug] dit_step27_vt: [689, 128] first4: -0.757228 2.440642 -0.487102 2.348062
[Debug] dit_step27_xt: [689, 128] first4: -0.733121 2.456598 -0.382601 2.258776
[Debug] dit_step28_vt_cond: [689, 128] first4: -0.840615 2.538517 -0.479540 2.402374
[Debug] dit_step28_vt_uncond: [689, 128] first4: -0.770583 2.568447 -0.404145 2.410012
[Debug] dit_step28_vt: [689, 128] first4: -0.889637 2.517566 -0.532317 2.397028
[Debug] dit_step28_xt: [689, 128] first4: -0.762776 2.540517 -0.400344 2.338677
[Debug] dit_step29_vt_cond: [689, 128] first4: -0.791307 2.647017 -0.506605 2.440136
[Debug] dit_step29_vt_uncond: [689, 128] first4: -0.800822 2.620979 -0.483372 2.340769
[Debug] dit_step29_vt: [689, 128] first4: -0.784646 2.665243 -0.522869 2.509693
[Debug] dit_step29_xt: [689, 128] first4: -0.788931 2.629358 -0.417773 2.422333
[Debug] dit_x0: [689, 128] first4: -0.788931 2.629358 -0.417773 2.422333
[Debug] window0_cond: [689, 2048] first4: -0.290057 -0.003712 -0.019914 -0.070771
[Debug] window0_latent: [689, 128] first4: -0.788931 2.629358 -0.417773 2.422333
[DiT] Window 1/1: T=689, 30 steps, 2393 ms (79.8 ms/step)
[DiT] CFG=1.70, 1 windows, 2.4 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: -0.055705 -0.008934 -0.035210 0.003286
[VAE] Decode: 1 windows -> 8.0s of audio, 60 ms
[Done] 8.4 s total
[Store] Unload VAE (103.4 MB)
[Store] Unload DiT (1611.2 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-Q5_K_M.gguf
[GGML] Running MiniMax-Music3-transformer-Q5_K_M.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 0.999983
hidden_after_preprocess: ggml vs python 1.000000
hidden_after_proj_in: ggml vs python 0.999718
layer0_sa_output: ggml vs python 0.999955
hidden_after_layer0: ggml vs python 0.999956
hidden_after_layer6: ggml vs python 0.999973
hidden_after_layer12: ggml vs python 0.999954
hidden_after_layer18: ggml vs python 0.999950
hidden_after_layer35: ggml vs python 0.999828
dit_step0_vt_cond: ggml vs python 0.998750
dit_step0_vt_uncond: ggml vs python 0.998036
dit_step0_vt: ggml vs python 0.998357
dit_step0_xt: ggml vs python 0.999995
dit_step5_vt_cond: ggml vs python 0.996741
dit_step5_vt: ggml vs python 0.995041
dit_step5_xt: ggml vs python 0.999733
dit_step10_vt_cond: ggml vs python 0.996305
dit_step10_vt: ggml vs python 0.995029
dit_step10_xt: ggml vs python 0.998960
dit_step15_vt_cond: ggml vs python 0.995727
dit_step15_vt: ggml vs python 0.994532
dit_step15_xt: ggml vs python 0.998178
dit_step20_vt_cond: ggml vs python 0.995013
dit_step20_vt: ggml vs python 0.993915
dit_step20_xt: ggml vs python 0.997650
dit_step25_vt_cond: ggml vs python 0.995008
dit_step25_vt: ggml vs python 0.994168
dit_step25_xt: ggml vs python 0.997318
dit_step29_vt_cond: ggml vs python 0.995478
dit_step29_vt: ggml vs python 0.994751
dit_x0: ggml vs python 0.997171
vae_audio: ggml vs python 0.989955
vae_audio (STFT cosine): ggml vs python 0.996030
[DiT] Error growth GGML vs Python
dit_step0_xt: cos 0.999995, max_err 0.019520, mean_err 0.002402, mean_A -0.001320, std_A 0.965140, mean_B -0.002135, std_B 0.965041
dit_step5_xt: cos 0.999733, max_err 0.264370, mean_err 0.014831, mean_A 0.003642, std_A 0.885143, mean_B 0.002533, std_B 0.884593
dit_step10_xt: cos 0.998960, max_err 0.690633, mean_err 0.030889, mean_A 0.009218, std_A 0.951070, mean_B 0.007973, std_B 0.950289
dit_step15_xt: cos 0.998178, max_err 1.080672, mean_err 0.049094, mean_A 0.015412, std_A 1.133625, mean_B 0.013735, std_B 1.132820
dit_step20_xt: cos 0.997650, max_err 1.469190, mean_err 0.068823, mean_A 0.021504, std_A 1.386737, mean_B 0.019518, std_B 1.385859
dit_step25_xt: cos 0.997318, max_err 1.864732, mean_err 0.089731, mean_A 0.027225, std_A 1.678486, mean_B 0.025057, std_B 1.677655