[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: 2061 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 158 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 (148.6 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: 61 ms
[GGUF] ../models/MiniMax-Music3-transformer-F32.gguf: 441 tensors, data at offset 31616
[Load] DiT backend: Vulkan0 (shared)
[WeightCtx] Loaded 436 tensors, 9259.0 MB into backend
[DiT] Loaded: 36 layers, dim 2048, flash_attn=1
[Store] Load DiT: 1020 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: 111 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: 0.059375 3.944638 -0.796440 2.207947
[Debug] hidden_after_proj_in: [689, 2048] first4: 3.502075 -5.731445 -1.940430 -0.170166
[Debug] layer0_sa_output: [690, 2048] first4: 1.555603 -1.825684 0.385040 1.460754
[Debug] hidden_after_layer0: [690, 2048] first4: -2.250368 -1.216694 -3.198996 3.631584
[Debug] hidden_after_layer6: [690, 2048] first4: -2.622287 0.967970 1.847614 0.432568
[Debug] hidden_after_layer12: [690, 2048] first4: -0.679764 0.602127 3.989414 2.947779
[Debug] hidden_after_layer18: [690, 2048] first4: -2.770644 -1.268000 2.978302 -1.054674
[Debug] hidden_after_layer35: [690, 2048] first4: -2.568978 1.871349 -1.545237 -4.797351
[Debug] dit_step0_vt_cond: [689, 128] first4: -0.341919 -2.044861 -0.337341 -0.680054
[Debug] dit_step0_vt_uncond: [689, 128] first4: -0.357056 -2.265656 -0.432228 -0.614746
[Debug] dit_step0_vt: [689, 128] first4: -0.331323 -1.890305 -0.270921 -0.725769
[Debug] dit_step0_xt: [689, 128] first4: 0.182975 2.098364 -0.181081 0.824868
[Debug] dit_step1_vt_cond: [689, 128] first4: -0.287231 -1.806519 -0.145508 -0.561768
[Debug] dit_step1_vt_uncond: [689, 128] first4: -0.626095 -1.615479 -0.265625 -0.356873
[Debug] dit_step1_vt: [689, 128] first4: -0.050027 -1.940247 -0.061426 -0.705194
[Debug] dit_step1_xt: [689, 128] first4: 0.181307 2.033689 -0.183129 0.801361
[Debug] dit_step2_vt_cond: [689, 128] first4: -0.291687 -1.632019 -0.053711 -0.470093
[Debug] dit_step2_vt_uncond: [689, 128] first4: -0.810120 -1.198486 -0.260132 -0.141479
[Debug] dit_step2_vt: [689, 128] first4: 0.071216 -1.935492 0.090784 -0.700122
[Debug] dit_step2_xt: [689, 128] first4: 0.183681 1.969173 -0.180103 0.778024
[Debug] dit_step3_vt_cond: [689, 128] first4: -0.309052 -1.517578 0.079102 -0.413086
[Debug] dit_step3_vt_uncond: [689, 128] first4: -0.253418 -1.445435 -0.372070 -0.540771
[Debug] dit_step3_vt: [689, 128] first4: -0.347995 -1.568079 0.394922 -0.323706
[Debug] dit_step3_xt: [689, 128] first4: 0.172081 1.916903 -0.166939 0.767234
[Debug] dit_step4_vt_cond: [689, 128] first4: -0.324677 -1.381836 0.143555 -0.335205
[Debug] dit_step4_vt_uncond: [689, 128] first4: -0.180054 -1.407349 -0.102539 -0.533936
[Debug] dit_step4_vt: [689, 128] first4: -0.425912 -1.363977 0.315820 -0.196094
[Debug] dit_step4_xt: [689, 128] first4: 0.157884 1.871438 -0.156411 0.760697
[Debug] dit_step5_vt_cond: [689, 128] first4: -0.326973 -1.232666 0.141602 -0.248779
[Debug] dit_step5_vt_uncond: [689, 128] first4: -0.159424 -1.259644 -0.015625 -0.411621
[Debug] dit_step5_vt: [689, 128] first4: -0.444257 -1.213782 0.251660 -0.134790
[Debug] dit_step5_xt: [689, 128] first4: 0.143076 1.830978 -0.148023 0.756204
[Debug] dit_step6_vt_cond: [689, 128] first4: -0.310974 -1.042236 0.065430 -0.139771
[Debug] dit_step6_vt_uncond: [689, 128] first4: -0.070801 -1.047485 -0.172119 -0.237793
[Debug] dit_step6_vt: [689, 128] first4: -0.479095 -1.038562 0.231714 -0.071155
[Debug] dit_step6_xt: [689, 128] first4: 0.127106 1.796359 -0.140299 0.753832
[Debug] dit_step7_vt_cond: [689, 128] first4: -0.301270 -0.776749 -0.041016 -0.034180
[Debug] dit_step7_vt_uncond: [689, 128] first4: -0.024750 -0.701904 -0.334229 -0.044800
[Debug] dit_step7_vt: [689, 128] first4: -0.494833 -0.829140 0.164233 -0.026746
[Debug] dit_step7_xt: [689, 128] first4: 0.110611 1.768721 -0.134824 0.752941
[Debug] dit_step8_vt_cond: [689, 128] first4: -0.267822 -0.518555 -0.148682 0.058655
[Debug] dit_step8_vt_uncond: [689, 128] first4: 0.000854 -0.311768 -0.441772 0.112671
[Debug] dit_step8_vt: [689, 128] first4: -0.455896 -0.663306 0.056482 0.020844
[Debug] dit_step8_xt: [689, 128] first4: 0.095415 1.746611 -0.132942 0.753636
[Debug] dit_step9_vt_cond: [689, 128] first4: -0.220215 -0.255554 -0.255981 0.158936
[Debug] dit_step9_vt_uncond: [689, 128] first4: 0.039001 0.051727 -0.539429 0.249069
[Debug] dit_step9_vt: [689, 128] first4: -0.401666 -0.470651 -0.057568 0.095842
[Debug] dit_step9_xt: [689, 128] first4: 0.082026 1.730923 -0.134861 0.756831
[Debug] dit_step10_vt_cond: [689, 128] first4: -0.158447 -0.012192 -0.362000 0.251709
[Debug] dit_step10_vt_uncond: [689, 128] first4: 0.063232 0.348419 -0.582520 0.355347
[Debug] dit_step10_vt: [689, 128] first4: -0.313623 -0.264619 -0.207635 0.179163
[Debug] dit_step10_xt: [689, 128] first4: 0.071572 1.722102 -0.141782 0.762803
[Debug] dit_step11_vt_cond: [689, 128] first4: -0.093018 0.210449 -0.455811 0.332199
[Debug] dit_step11_vt_uncond: [689, 128] first4: 0.098755 0.618301 -0.633057 0.452148
[Debug] dit_step11_vt: [689, 128] first4: -0.227258 -0.075047 -0.331738 0.248235
[Debug] dit_step11_xt: [689, 128] first4: 0.063996 1.719601 -0.152840 0.771077
[Debug] dit_step12_vt_cond: [689, 128] first4: -0.027588 0.427979 -0.544617 0.402100
[Debug] dit_step12_vt_uncond: [689, 128] first4: 0.143799 0.850159 -0.688347 0.516846
[Debug] dit_step12_vt: [689, 128] first4: -0.147559 0.132452 -0.444006 0.321777
[Debug] dit_step12_xt: [689, 128] first4: 0.059078 1.724016 -0.167640 0.781803
[Debug] dit_step13_vt_cond: [689, 128] first4: 0.041870 0.629639 -0.616699 0.454102
[Debug] dit_step13_vt_uncond: [689, 128] first4: 0.172607 1.049377 -0.751709 0.565918
[Debug] dit_step13_vt: [689, 128] first4: -0.049646 0.335822 -0.522192 0.375830
[Debug] dit_step13_xt: [689, 128] first4: 0.057423 1.735210 -0.185046 0.794331
[Debug] dit_step14_vt_cond: [689, 128] first4: 0.098143 0.829346 -0.682617 0.501465
[Debug] dit_step14_vt_uncond: [689, 128] first4: 0.191162 1.198853 -0.806946 0.609375
[Debug] dit_step14_vt: [689, 128] first4: 0.033029 0.570691 -0.595587 0.425928
[Debug] dit_step14_xt: [689, 128] first4: 0.058524 1.754233 -0.204899 0.808528
[Debug] dit_step15_vt_cond: [689, 128] first4: 0.144409 1.017334 -0.730597 0.546631
[Debug] dit_step15_vt_uncond: [689, 128] first4: 0.208252 1.328125 -0.852844 0.664551
[Debug] dit_step15_vt: [689, 128] first4: 0.099719 0.799780 -0.645025 0.464087
[Debug] dit_step15_xt: [689, 128] first4: 0.061848 1.780892 -0.226400 0.823998
[Debug] dit_step16_vt_cond: [689, 128] first4: 0.179321 1.213867 -0.761292 0.596191
[Debug] dit_step16_vt_uncond: [689, 128] first4: 0.213379 1.485352 -0.882629 0.731934
[Debug] dit_step16_vt: [689, 128] first4: 0.155481 1.023828 -0.676355 0.501172
[Debug] dit_step16_xt: [689, 128] first4: 0.067031 1.815020 -0.248945 0.840704
[Debug] dit_step17_vt_cond: [689, 128] first4: 0.212708 1.395508 -0.768555 0.645020
[Debug] dit_step17_vt_uncond: [689, 128] first4: 0.227722 1.631836 -0.878418 0.783447
[Debug] dit_step17_vt: [689, 128] first4: 0.202197 1.230078 -0.691650 0.548120
[Debug] dit_step17_xt: [689, 128] first4: 0.073771 1.856022 -0.272000 0.858974
[Debug] dit_step18_vt_cond: [689, 128] first4: 0.240250 1.545898 -0.764648 0.691650
[Debug] dit_step18_vt_uncond: [689, 128] first4: 0.246216 1.788086 -0.867676 0.836670
[Debug] dit_step18_vt: [689, 128] first4: 0.236073 1.376367 -0.692529 0.590137
[Debug] dit_step18_xt: [689, 128] first4: 0.081640 1.901901 -0.295085 0.878645
[Debug] dit_step19_vt_cond: [689, 128] first4: 0.260376 1.706787 -0.756836 0.739502
[Debug] dit_step19_vt_uncond: [689, 128] first4: 0.270020 1.906372 -0.827637 0.875000
[Debug] dit_step19_vt: [689, 128] first4: 0.253625 1.567078 -0.707275 0.644653
[Debug] dit_step19_xt: [689, 128] first4: 0.090094 1.954137 -0.318660 0.900134
[Debug] dit_step20_vt_cond: [689, 128] first4: 0.282532 1.819580 -0.737305 0.796387
[Debug] dit_step20_vt_uncond: [689, 128] first4: 0.296997 2.020935 -0.787598 0.929932
[Debug] dit_step20_vt: [689, 128] first4: 0.272406 1.678632 -0.702100 0.702905
[Debug] dit_step20_xt: [689, 128] first4: 0.099174 2.010091 -0.342064 0.923564
[Debug] dit_step21_vt_cond: [689, 128] first4: 0.304553 1.948364 -0.721191 0.844604
[Debug] dit_step21_vt_uncond: [689, 128] first4: 0.324951 2.098846 -0.735840 0.960571
[Debug] dit_step21_vt: [689, 128] first4: 0.290274 1.843027 -0.710938 0.763428
[Debug] dit_step21_xt: [689, 128] first4: 0.108850 2.071526 -0.365762 0.949012
[Debug] dit_step22_vt_cond: [689, 128] first4: 0.309601 2.055435 -0.694336 0.895508
[Debug] dit_step22_vt_uncond: [689, 128] first4: 0.361572 2.195679 -0.692993 0.986084
[Debug] dit_step22_vt: [689, 128] first4: 0.273221 1.957265 -0.695276 0.832105
[Debug] dit_step22_xt: [689, 128] first4: 0.117957 2.136768 -0.388937 0.976748
[Debug] dit_step23_vt_cond: [689, 128] first4: 0.317322 2.169800 -0.674774 0.952637
[Debug] dit_step23_vt_uncond: [689, 128] first4: 0.378296 2.272552 -0.643738 1.003906
[Debug] dit_step23_vt: [689, 128] first4: 0.274640 2.097873 -0.696500 0.916748
[Debug] dit_step23_xt: [689, 128] first4: 0.127112 2.206697 -0.412154 1.007307
[Debug] dit_step24_vt_cond: [689, 128] first4: 0.330261 2.263428 -0.652344 1.007324
[Debug] dit_step24_vt_uncond: [689, 128] first4: 0.394814 2.352158 -0.601074 1.035156
[Debug] dit_step24_vt: [689, 128] first4: 0.285075 2.201317 -0.688232 0.987842
[Debug] dit_step24_xt: [689, 128] first4: 0.136614 2.280074 -0.435095 1.040235
[Debug] dit_step25_vt_cond: [689, 128] first4: 0.339569 2.362244 -0.626953 1.059326
[Debug] dit_step25_vt_uncond: [689, 128] first4: 0.395264 2.387390 -0.567627 1.080078
[Debug] dit_step25_vt: [689, 128] first4: 0.300583 2.344641 -0.668481 1.044800
[Debug] dit_step25_xt: [689, 128] first4: 0.146634 2.358229 -0.457378 1.075061
[Debug] dit_step26_vt_cond: [689, 128] first4: 0.334991 2.441923 -0.614746 1.125000
[Debug] dit_step26_vt_uncond: [689, 128] first4: 0.389496 2.485474 -0.553223 1.114014
[Debug] dit_step26_vt: [689, 128] first4: 0.296838 2.411438 -0.657812 1.132690
[Debug] dit_step26_xt: [689, 128] first4: 0.156528 2.438610 -0.479305 1.112818
[Debug] dit_step27_vt_cond: [689, 128] first4: 0.333008 2.520691 -0.606445 1.167236
[Debug] dit_step27_vt_uncond: [689, 128] first4: 0.361816 2.526611 -0.557129 1.142822
[Debug] dit_step27_vt: [689, 128] first4: 0.312842 2.516547 -0.640967 1.184326
[Debug] dit_step27_xt: [689, 128] first4: 0.166956 2.522495 -0.500671 1.152295
[Debug] dit_step28_vt_cond: [689, 128] first4: 0.309814 2.536377 -0.602539 1.197754
[Debug] dit_step28_vt_uncond: [689, 128] first4: 0.316895 2.580505 -0.553223 1.157959
[Debug] dit_step28_vt: [689, 128] first4: 0.304858 2.505487 -0.637061 1.225610
[Debug] dit_step28_xt: [689, 128] first4: 0.177118 2.606011 -0.521906 1.193149
[Debug] dit_step29_vt_cond: [689, 128] first4: 0.264404 2.586731 -0.625977 1.204346
[Debug] dit_step29_vt_uncond: [689, 128] first4: 0.262451 2.587433 -0.585938 1.161499
[Debug] dit_step29_vt: [689, 128] first4: 0.265771 2.586240 -0.654004 1.234338
[Debug] dit_step29_xt: [689, 128] first4: 0.185977 2.692219 -0.543706 1.234294
[Debug] dit_x0: [689, 128] first4: 0.185977 2.692219 -0.543706 1.234294
[Debug] window0_cond: [689, 2048] first4: 0.292188 -0.052068 0.070921 0.078521
[Debug] window0_latent: [689, 128] first4: 0.185977 2.692219 -0.543706 1.234294
[DiT] Window 1/1: T=689, 30 steps, 2122 ms (70.7 ms/step)
[DiT] CFG=1.70, 1 windows, 2.1 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: 0.014655 -0.001303 0.004713 0.000821
[VAE] Decode: 1 windows -> 8.0s of audio, 93 ms
[Done] 35.6 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 0.999999
layer0_sa_output: ggml vs python 1.000001
hidden_after_layer0: ggml vs python 1.000000
hidden_after_layer6: ggml vs python 1.000000
hidden_after_layer12: ggml vs python 1.000001
hidden_after_layer18: ggml vs python 0.999999
hidden_after_layer35: ggml vs python 0.999999
dit_step0_vt_cond: ggml vs python 0.999995
dit_step0_vt_uncond: ggml vs python 0.999994
dit_step0_vt: ggml vs python 0.999991
dit_step0_xt: ggml vs python 1.000000
dit_step5_vt_cond: ggml vs python 0.999993
dit_step5_vt: ggml vs python 0.999983
dit_step5_xt: ggml vs python 0.999999
dit_step10_vt_cond: ggml vs python 0.999992
dit_step10_vt: ggml vs python 0.999981
dit_step10_xt: ggml vs python 0.999999
dit_step15_vt_cond: ggml vs python 0.999991
dit_step15_vt: ggml vs python 0.999977
dit_step15_xt: ggml vs python 0.999998
dit_step20_vt_cond: ggml vs python 0.999989
dit_step20_vt: ggml vs python 0.999973
dit_step20_xt: ggml vs python 0.999998
dit_step25_vt_cond: ggml vs python 0.999988
dit_step25_vt: ggml vs python 0.999972
dit_step25_xt: ggml vs python 0.999997
dit_step29_vt_cond: ggml vs python 0.999991
dit_step29_vt: ggml vs python 0.999977
dit_x0: ggml vs python 0.999997
vae_audio: ggml vs python 0.999988
vae_audio (STFT cosine): ggml vs python 0.999994
[DiT] Error growth GGML vs Python
dit_step0_xt: cos 1.000000, max_err 0.003108, mean_err 0.000163, mean_A -0.001888, std_A 0.964985, mean_B -0.001888, std_B 0.964984
dit_step5_xt: cos 0.999999, max_err 0.009830, mean_err 0.000606, mean_A 0.001978, std_A 0.875759, mean_B 0.001975, std_B 0.875768
dit_step10_xt: cos 0.999999, max_err 0.018023, mean_err 0.001001, mean_A 0.007031, std_A 0.919302, mean_B 0.007025, std_B 0.919308
dit_step15_xt: cos 0.999998, max_err 0.026060, mean_err 0.001444, mean_A 0.012251, std_A 1.079009, mean_B 0.012237, std_B 1.079004
dit_step20_xt: cos 0.999998, max_err 0.034425, mean_err 0.001955, mean_A 0.017328, std_A 1.312273, mean_B 0.017310, std_B 1.312272
dit_step25_xt: cos 0.999997, max_err 0.049076, mean_err 0.002528, mean_A 0.022298, std_A 1.586107, mean_B 0.022269, std_B 1.586108