[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: 1354 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-Q8_0.gguf: 441 tensors, data at offset 31712
[Load] DiT backend: CUDA0 (shared)
[WeightCtx] Loaded 436 tensors, 2477.0 MB into backend
[DiT] Loaded: 36 layers, dim 2048, flash_attn=1
[Store] Load DiT: 252 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: 3.626598 -3.071267 -3.962117 4.059601
[Debug] hidden_after_preprocess: [689, 2304] first4: 0.341482 4.083567 -1.024049 1.790691
[Debug] hidden_after_proj_in: [689, 2048] first4: 5.216561 -4.322406 -5.954851 -1.885700
[Debug] layer0_sa_output: [690, 2048] first4: 1.682718 -1.863372 0.463236 1.519236
[Debug] hidden_after_layer0: [690, 2048] first4: -1.968753 -1.311648 -3.165219 3.766338
[Debug] hidden_after_layer6: [690, 2048] first4: -4.127179 -0.219267 0.365011 2.027999
[Debug] hidden_after_layer12: [690, 2048] first4: -2.241782 -1.707766 1.645301 3.533360
[Debug] hidden_after_layer18: [690, 2048] first4: -2.490401 -2.256999 -0.742836 1.760959
[Debug] hidden_after_layer35: [690, 2048] first4: -1.434917 -0.303268 -3.096095 -0.251890
[Debug] dit_step0_vt_cond: [689, 128] first4: -0.608625 -1.875455 -0.511795 -0.374536
[Debug] dit_step0_vt_uncond: [689, 128] first4: -0.374023 -2.273755 -0.423508 -0.607524
[Debug] dit_step0_vt: [689, 128] first4: -0.772846 -1.596645 -0.573595 -0.211444
[Debug] dit_step0_xt: [689, 128] first4: 0.168257 2.108152 -0.191170 0.842012
[Debug] dit_step1_vt_cond: [689, 128] first4: -0.665750 -1.787993 -0.418238 -0.291357
[Debug] dit_step1_vt_uncond: [689, 128] first4: -0.454776 -1.848079 -0.310900 -0.660754
[Debug] dit_step1_vt: [689, 128] first4: -0.813432 -1.745934 -0.493374 -0.032779
[Debug] dit_step1_xt: [689, 128] first4: 0.141143 2.049955 -0.207616 0.840919
[Debug] dit_step2_vt_cond: [689, 128] first4: -0.692018 -1.762044 -0.406717 -0.136544
[Debug] dit_step2_vt_uncond: [689, 128] first4: -0.733205 -1.980581 -0.506883 -0.591616
[Debug] dit_step2_vt: [689, 128] first4: -0.663186 -1.609068 -0.336601 0.182007
[Debug] dit_step2_xt: [689, 128] first4: 0.119037 1.996319 -0.218836 0.846986
[Debug] dit_step3_vt_cond: [689, 128] first4: -0.735789 -1.658086 -0.332508 -0.003189
[Debug] dit_step3_vt_uncond: [689, 128] first4: -0.283393 -1.894445 -0.519584 -0.267739
[Debug] dit_step3_vt: [689, 128] first4: -1.052466 -1.492634 -0.201556 0.181997
[Debug] dit_step3_xt: [689, 128] first4: 0.083955 1.946565 -0.225555 0.853053
[Debug] dit_step4_vt_cond: [689, 128] first4: -0.779378 -1.522730 -0.335224 0.229808
[Debug] dit_step4_vt_uncond: [689, 128] first4: -0.257616 -1.499757 -0.398014 0.071058
[Debug] dit_step4_vt: [689, 128] first4: -1.144611 -1.538811 -0.291270 0.340934
[Debug] dit_step4_xt: [689, 128] first4: 0.045801 1.895271 -0.235264 0.864417
[Debug] dit_step5_vt_cond: [689, 128] first4: -0.830109 -1.399254 -0.289744 0.423313
[Debug] dit_step5_vt_uncond: [689, 128] first4: -0.372653 -1.310444 -0.467421 0.163431
[Debug] dit_step5_vt: [689, 128] first4: -1.150328 -1.461421 -0.165370 0.605229
[Debug] dit_step5_xt: [689, 128] first4: 0.007457 1.846557 -0.240776 0.884592
[Debug] dit_step6_vt_cond: [689, 128] first4: -0.888258 -1.231486 -0.264115 0.644219
[Debug] dit_step6_vt_uncond: [689, 128] first4: -0.459718 -1.318460 -0.631753 0.526502
[Debug] dit_step6_vt: [689, 128] first4: -1.188237 -1.170604 -0.006769 0.726620
[Debug] dit_step6_xt: [689, 128] first4: -0.032151 1.807537 -0.241002 0.908812
[Debug] dit_step7_vt_cond: [689, 128] first4: -0.945925 -1.055138 -0.200901 0.858550
[Debug] dit_step7_vt_uncond: [689, 128] first4: -0.552224 -1.244396 -0.810314 0.904151
[Debug] dit_step7_vt: [689, 128] first4: -1.221515 -0.922658 0.225688 0.826629
[Debug] dit_step7_xt: [689, 128] first4: -0.072869 1.776782 -0.233479 0.936366
[Debug] dit_step8_vt_cond: [689, 128] first4: -1.048808 -0.864127 -0.128268 0.977911
[Debug] dit_step8_vt_uncond: [689, 128] first4: -0.689290 -0.976245 -0.627727 1.149060
[Debug] dit_step8_vt: [689, 128] first4: -1.300470 -0.785645 0.221353 0.858106
[Debug] dit_step8_xt: [689, 128] first4: -0.116217 1.750593 -0.226100 0.964970
[Debug] dit_step9_vt_cond: [689, 128] first4: -1.066652 -0.642616 -0.119987 1.128857
[Debug] dit_step9_vt_uncond: [689, 128] first4: -0.746489 -0.771012 -0.388536 1.316881
[Debug] dit_step9_vt: [689, 128] first4: -1.290766 -0.552739 0.067997 0.997239
[Debug] dit_step9_xt: [689, 128] first4: -0.159243 1.732169 -0.223834 0.998211
[Debug] dit_step10_vt_cond: [689, 128] first4: -1.084811 -0.402975 -0.112258 1.284113
[Debug] dit_step10_vt_uncond: [689, 128] first4: -0.891440 -0.590426 -0.215049 1.415634
[Debug] dit_step10_vt: [689, 128] first4: -1.220170 -0.271760 -0.040305 1.192049
[Debug] dit_step10_xt: [689, 128] first4: -0.199915 1.723110 -0.225177 1.037946
[Debug] dit_step11_vt_cond: [689, 128] first4: -1.099982 -0.165073 -0.137851 1.456448
[Debug] dit_step11_vt_uncond: [689, 128] first4: -0.955519 -0.368499 -0.146331 1.579862
[Debug] dit_step11_vt: [689, 128] first4: -1.201106 -0.022674 -0.131914 1.370058
[Debug] dit_step11_xt: [689, 128] first4: -0.239952 1.722354 -0.229575 1.083615
[Debug] dit_step12_vt_cond: [689, 128] first4: -1.109673 0.041330 -0.178791 1.621211
[Debug] dit_step12_vt_uncond: [689, 128] first4: -0.969306 -0.150002 -0.053398 1.652195
[Debug] dit_step12_vt: [689, 128] first4: -1.207929 0.175263 -0.266566 1.599521
[Debug] dit_step12_xt: [689, 128] first4: -0.280217 1.728196 -0.238460 1.136932
[Debug] dit_step13_vt_cond: [689, 128] first4: -1.104385 0.273659 -0.173092 1.773040
[Debug] dit_step13_vt_uncond: [689, 128] first4: -0.994190 0.109622 -0.060111 1.801473
[Debug] dit_step13_vt: [689, 128] first4: -1.181522 0.388486 -0.252178 1.753137
[Debug] dit_step13_xt: [689, 128] first4: -0.319601 1.741146 -0.246866 1.195370
[Debug] dit_step14_vt_cond: [689, 128] first4: -1.064391 0.447022 -0.153509 1.912056
[Debug] dit_step14_vt_uncond: [689, 128] first4: -1.043884 0.357829 -0.074618 1.878090
[Debug] dit_step14_vt: [689, 128] first4: -1.078746 0.509457 -0.208732 1.935832
[Debug] dit_step14_xt: [689, 128] first4: -0.355559 1.758128 -0.253824 1.259898
[Debug] dit_step15_vt_cond: [689, 128] first4: -1.034582 0.642218 -0.136640 1.997462
[Debug] dit_step15_vt_uncond: [689, 128] first4: -1.022784 0.589854 -0.074196 1.961160
[Debug] dit_step15_vt: [689, 128] first4: -1.042841 0.678872 -0.180351 2.022874
[Debug] dit_step15_xt: [689, 128] first4: -0.390320 1.780757 -0.259835 1.327327
[Debug] dit_step16_vt_cond: [689, 128] first4: -0.990166 0.829046 -0.159825 2.094293
[Debug] dit_step16_vt_uncond: [689, 128] first4: -0.987738 0.821537 -0.064192 2.045485
[Debug] dit_step16_vt: [689, 128] first4: -0.991866 0.834302 -0.226769 2.128458
[Debug] dit_step16_xt: [689, 128] first4: -0.423382 1.808567 -0.267394 1.398276
[Debug] dit_step17_vt_cond: [689, 128] first4: -1.002480 1.034342 -0.167460 2.167799
[Debug] dit_step17_vt_uncond: [689, 128] first4: -0.962676 1.051834 -0.037558 2.102093
[Debug] dit_step17_vt: [689, 128] first4: -1.030342 1.022097 -0.258392 2.213793
[Debug] dit_step17_xt: [689, 128] first4: -0.457727 1.842637 -0.276007 1.472069
[Debug] dit_step18_vt_cond: [689, 128] first4: -0.937155 1.225655 -0.161932 2.241035
[Debug] dit_step18_vt_uncond: [689, 128] first4: -0.988231 1.224216 -0.072219 2.146763
[Debug] dit_step18_vt: [689, 128] first4: -0.901402 1.226662 -0.224732 2.307025
[Debug] dit_step18_xt: [689, 128] first4: -0.487774 1.883526 -0.283499 1.548970
[Debug] dit_step19_vt_cond: [689, 128] first4: -0.891931 1.394991 -0.150753 2.279128
[Debug] dit_step19_vt_uncond: [689, 128] first4: -0.923245 1.405508 -0.053652 2.194935
[Debug] dit_step19_vt: [689, 128] first4: -0.870011 1.387629 -0.218724 2.338063
[Debug] dit_step19_xt: [689, 128] first4: -0.516774 1.929780 -0.290789 1.626905
[Debug] dit_step20_vt_cond: [689, 128] first4: -0.800018 1.492523 -0.167347 2.320457
[Debug] dit_step20_vt_uncond: [689, 128] first4: -0.846360 1.558726 -0.082342 2.253069
[Debug] dit_step20_vt: [689, 128] first4: -0.767579 1.446181 -0.226850 2.367628
[Debug] dit_step20_xt: [689, 128] first4: -0.542360 1.977986 -0.298351 1.705826
[Debug] dit_step21_vt_cond: [689, 128] first4: -0.783941 1.626314 -0.186642 2.328343
[Debug] dit_step21_vt_uncond: [689, 128] first4: -0.795719 1.675608 -0.076471 2.251176
[Debug] dit_step21_vt: [689, 128] first4: -0.775696 1.591809 -0.263761 2.382361
[Debug] dit_step21_xt: [689, 128] first4: -0.568217 2.031046 -0.307143 1.785238
[Debug] dit_step22_vt_cond: [689, 128] first4: -0.732728 1.780992 -0.188624 2.361874
[Debug] dit_step22_vt_uncond: [689, 128] first4: -0.755031 1.817601 -0.077244 2.258783
[Debug] dit_step22_vt: [689, 128] first4: -0.717116 1.755365 -0.266590 2.434037
[Debug] dit_step22_xt: [689, 128] first4: -0.592121 2.089559 -0.316029 1.866372
[Debug] dit_step23_vt_cond: [689, 128] first4: -0.703248 1.904437 -0.219809 2.361421
[Debug] dit_step23_vt_uncond: [689, 128] first4: -0.714774 1.949152 -0.106139 2.286816
[Debug] dit_step23_vt: [689, 128] first4: -0.695180 1.873137 -0.299377 2.413644
[Debug] dit_step23_xt: [689, 128] first4: -0.615293 2.151996 -0.326009 1.946827
[Debug] dit_step24_vt_cond: [689, 128] first4: -0.720546 2.041474 -0.237392 2.381482
[Debug] dit_step24_vt_uncond: [689, 128] first4: -0.691401 2.070168 -0.124165 2.299955
[Debug] dit_step24_vt: [689, 128] first4: -0.740948 2.021389 -0.316651 2.438551
[Debug] dit_step24_xt: [689, 128] first4: -0.639992 2.219376 -0.336564 2.028112
[Debug] dit_step25_vt_cond: [689, 128] first4: -0.730121 2.186500 -0.237546 2.409747
[Debug] dit_step25_vt_uncond: [689, 128] first4: -0.688904 2.171427 -0.165631 2.314910
[Debug] dit_step25_vt: [689, 128] first4: -0.758974 2.197051 -0.287887 2.476133
[Debug] dit_step25_xt: [689, 128] first4: -0.665291 2.292611 -0.346160 2.110650
[Debug] dit_step26_vt_cond: [689, 128] first4: -0.706563 2.307671 -0.296049 2.437426
[Debug] dit_step26_vt_uncond: [689, 128] first4: -0.734823 2.299829 -0.219607 2.334319
[Debug] dit_step26_vt: [689, 128] first4: -0.686780 2.313159 -0.349558 2.509600
[Debug] dit_step26_xt: [689, 128] first4: -0.688183 2.369716 -0.357812 2.194303
[Debug] dit_step27_vt_cond: [689, 128] first4: -0.733301 2.448322 -0.344103 2.453575
[Debug] dit_step27_vt_uncond: [689, 128] first4: -0.734787 2.409380 -0.285081 2.387823
[Debug] dit_step27_vt: [689, 128] first4: -0.732261 2.475582 -0.385418 2.499602
[Debug] dit_step27_xt: [689, 128] first4: -0.712592 2.452236 -0.370659 2.277623
[Debug] dit_step28_vt_cond: [689, 128] first4: -0.741607 2.530958 -0.386780 2.427095
[Debug] dit_step28_vt_uncond: [689, 128] first4: -0.735079 2.495066 -0.354605 2.381951
[Debug] dit_step28_vt: [689, 128] first4: -0.746177 2.556083 -0.409303 2.458696
[Debug] dit_step28_xt: [689, 128] first4: -0.737465 2.537439 -0.384302 2.359580
[Debug] dit_step29_vt_cond: [689, 128] first4: -0.777543 2.599171 -0.439949 2.376674
[Debug] dit_step29_vt_uncond: [689, 128] first4: -0.747646 2.575167 -0.444840 2.362006
[Debug] dit_step29_vt: [689, 128] first4: -0.798471 2.615973 -0.436525 2.386941
[Debug] dit_step29_xt: [689, 128] first4: -0.764080 2.624638 -0.398853 2.439145
[Debug] dit_x0: [689, 128] first4: -0.764080 2.624638 -0.398853 2.439145
[Debug] window0_cond: [689, 2048] first4: -0.290057 -0.003712 -0.019914 -0.070771
[Debug] window0_latent: [689, 128] first4: -0.764080 2.624638 -0.398853 2.439145
[DiT] Window 1/1: T=689, 30 steps, 2312 ms (77.1 ms/step)
[DiT] CFG=1.70, 1 windows, 2.3 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: -0.043346 -0.006676 -0.033961 0.004513
[VAE] Decode: 1 windows -> 8.0s of audio, 85 ms
[Done] 8.5 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 0.999988
layer0_sa_output: ggml vs python 0.999986
hidden_after_layer0: ggml vs python 0.999993
hidden_after_layer6: ggml vs python 0.999999
hidden_after_layer12: ggml vs python 0.999995
hidden_after_layer18: ggml vs python 0.999998
hidden_after_layer35: ggml vs python 0.999987
dit_step0_vt_cond: ggml vs python 0.999924
dit_step0_vt_uncond: ggml vs python 0.999916
dit_step0_vt: ggml vs python 0.999884
dit_step0_xt: ggml vs python 1.000000
dit_step5_vt_cond: ggml vs python 0.999845
dit_step5_vt: ggml vs python 0.999719
dit_step5_xt: ggml vs python 0.999988
dit_step10_vt_cond: ggml vs python 0.999810
dit_step10_vt: ggml vs python 0.999676
dit_step10_xt: ggml vs python 0.999956
dit_step15_vt_cond: ggml vs python 0.999807
dit_step15_vt: ggml vs python 0.999676
dit_step15_xt: ggml vs python 0.999925
dit_step20_vt_cond: ggml vs python 0.999774
dit_step20_vt: ggml vs python 0.999634
dit_step20_xt: ggml vs python 0.999907
dit_step25_vt_cond: ggml vs python 0.999783
dit_step25_vt: ggml vs python 0.999670
dit_step25_xt: ggml vs python 0.999895
dit_step29_vt_cond: ggml vs python 0.999809
dit_step29_vt: ggml vs python 0.999713
dit_x0: ggml vs python 0.999890
vae_audio: ggml vs python 0.999552
vae_audio (STFT cosine): ggml vs python 0.999822
[DiT] Error growth GGML vs Python
dit_step0_xt: cos 1.000000, max_err 0.004102, mean_err 0.000642, mean_A -0.002090, std_A 0.965049, mean_B -0.002135, std_B 0.965041
dit_step5_xt: cos 0.999988, max_err 0.056223, mean_err 0.003284, mean_A 0.002683, std_A 0.884447, mean_B 0.002533, std_B 0.884593
dit_step10_xt: cos 0.999956, max_err 0.151226, mean_err 0.006462, mean_A 0.008295, std_A 0.949849, mean_B 0.007973, std_B 0.950289
dit_step15_xt: cos 0.999925, max_err 0.226379, mean_err 0.009919, mean_A 0.014218, std_A 1.132018, mean_B 0.013735, std_B 1.132820
dit_step20_xt: cos 0.999907, max_err 0.306784, mean_err 0.013583, mean_A 0.020177, std_A 1.384799, mean_B 0.019518, std_B 1.385859
dit_step25_xt: cos 0.999895, max_err 0.470284, mean_err 0.017458, mean_A 0.025945, std_A 1.676333, mean_B 0.025057, std_B 1.677655