Upload folder using huggingface_hub

README.md

@@ -26,8 +26,13 @@ the rest is the song less them. The network between its STFT and iSTFT, as
 | File | Size | SHA-256 |
 |---|---|---|
 | `mel-roformer.fp16.onnx` | 464.1 MB (464,050,097 bytes) | `adcbbb826f12453fbb86652b503242d565a9c81cdef1bca2fe0e54752eba91ef` |
+| `mel-roformer.engine.bin` | 466.1 MB (466,082,640 bytes) | `140fedcd9d03a8f918b235688c69f4dfdd86b08f9d430277659c095688a026ff` |
+| `mel-roformer.engine.json` | 16 KB (16,282 bytes) | `3588f4da4a862d939d3e9e314d04c9b4c79706a54838fead43a73a14ca78f10b` |
 
-Float16 weights, float32 compute (onnxruntime casts them at load). The float32 export it is made from is 927.5 MB.
+The ONNX file: float16 weights, float32 compute (onnxruntime casts them at load); the float32 export it is made from is
+927.5 MB. The engine files: the same tensors (weights float16, tables float32, indices int32) for the package's own WebGPU
+engine, `roformer.js`, which a browser whose GPU offers subgroup matrices (Chromium on Metal) runs instead of onnxruntime:
+the `.json` names each tensor, its type, shape and offset in the `.bin`. A browser downloads one or the other.
 
 ## Source
 
@@ -92,7 +97,9 @@ GPU runs it in few, large operations (826 of them); none of it changes the funct
 segments, 89 to 92 dB SNR. This file against the float32 graph: 48 dB SNR on a segment of noise-floored tones.
 
 On onnxruntime 1.30's WebGPU (Metal, Apple M4 Max), 2.4 s a segment for the float32 graph, 3.0 for this one; on its CPU,
-9 s. SCNet-large's LSTMs take 19 of its 26 s a segment on the same WebGPU.
+9 s. SCNet-large's LSTMs take 19 of its 26 s a segment on the same WebGPU. The engine, in Chromium on Metal under the same
+load as onnxruntime-web's WebGPU there: 1.5 to 1.7 s a segment against 2.6 to 3.3, 47.8 dB from onnxruntime's CPU on the
+float32 graph (the float16 weights' rounding, as this ONNX file's).
 
 ## Quality
 

mel-roformer.engine.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:140fedcd9d03a8f918b235688c69f4dfdd86b08f9d430277659c095688a026ff
+size 466082640

mel-roformer.engine.json

@@ -0,0 +1 @@
+{"model": "mel-roformer", "frames": 801, "freqs": 1025, "channels": 2, "dim": 384, "heads": 8, "depth": 6, "bands": 60, "groups": [[0, 18, 28], [18, 24, 44], [24, 29, 64], [29, 34, 88], [34, 39, 124], [39, 43, 164], [43, 48, 228], [48, 52, 304], [52, 56, 396], [56, 60, 520]], "widths": [28, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 28, 28, 28, 36, 36, 36, 40, 40, 44, 52, 52, 52, 60, 64, 68, 76, 80, 80, 88, 96, 104, 112, 116, 124, 132, 144, 156, 164, 176, 188, 200, 216, 228, 244, 264, 284, 304, 320, 344, 372, 396, 420, 452, 488, 520], "tensors": {"split_w.0": {"type": "float16", "shape": [18, 28, 384], "offset": 0}, "split_w.1": {"type": "float16", "shape": [6, 44, 384], "offset": 387072}, "split_w.2": {"type": "float16", "shape": [5, 64, 384], "offset": 589824}, "split_w.3": {"type": "float16", "shape": [5, 88, 384], "offset": 835584}, "split_w.4": {"type": "float16", "shape": [5, 124, 384], "offset": 1173504}, "split_w.5": {"type": "float16", "shape": [4, 164, 384], "offset": 1649664}, "split_w.6": {"type": "float16", "shape": [5, 228, 384], "offset": 2153472}, "split_w.7": {"type": "float16", "shape": [4, 304, 384], "offset": 3028992}, "split_w.8": {"type": "float16", "shape": [4, 396, 384], "offset": 3962880}, "split_w.9": {"type": "float16", "shape": [4, 520, 384], "offset": 5179392}, "split_b.0": {"type": "float32", "shape": [18, 1, 384], "offset": 6776832}, "split_b.1": {"type": "float32", "shape": [6, 1, 384], "offset": 6804480}, "split_b.2": {"type": "float32", "shape": [5, 1, 384], "offset": 6813696}, "split_b.3": {"type": "float32", "shape": [5, 1, 384], "offset": 6821376}, "split_b.4": {"type": "float32", "shape": [5, 1, 384], "offset": 6829056}, "split_b.5": {"type": "float32", "shape": [4, 1, 384], "offset": 6836736}, "split_b.6": {"type": "float32", "shape": [5, 1, 384], "offset": 6842880}, "split_b.7": {"type": "float32", "shape": [4, 1, 384], "offset": 6850560}, "split_b.8": {"type": "float32", "shape": [4, 1, 384], "offset": 6856704}, "split_b.9": {"type": "float32", "shape": [4, 1, 384], "offset": 6862848}, "blocks.0.0.out": {"type": "float32", "shape": [384], "offset": 6868992}, "blocks.0.0.layers.0.0.qkvg": {"type": "float16", "shape": [384, 1544], "offset": 6870528}, "blocks.0.0.layers.0.0.gb": {"type": "float32", "shape": [8], "offset": 8056320}, "blocks.0.0.layers.0.0.wo": {"type": "float16", "shape": [512, 384], "offset": 8056352}, "blocks.0.0.layers.0.1.w1": {"type": "float16", "shape": [384, 1536], "offset": 8449568}, "blocks.0.0.layers.0.1.b1": {"type": "float32", "shape": [1536], "offset": 9629216}, "blocks.0.0.layers.0.1.w2": {"type": "float16", "shape": [1536, 384], "offset": 9635360}, "blocks.0.0.layers.0.1.b2": {"type": "float32", "shape": [384], "offset": 10815008}, "blocks.0.1.out": {"type": "float32", "shape": [384], "offset": 10816544}, "blocks.0.1.layers.0.0.qkvg": {"type": "float16", "shape": [384, 1544], "offset": 10818080}, "blocks.0.1.layers.0.0.gb": {"type": "float32", "shape": [8], "offset": 12003872}, "blocks.0.1.layers.0.0.wo": {"type": "float16", "shape": [512, 384], "offset": 12003904}, "blocks.0.1.layers.0.1.w1": {"type": "float16", "shape": [384, 1536], "offset": 12397120}, "blocks.0.1.layers.0.1.b1": {"type": "float32", "shape": [1536], "offset": 13576768}, "blocks.0.1.layers.0.1.w2": {"type": "float16", "shape": [1536, 384], "offset": 13582912}, "blocks.0.1.layers.0.1.b2": {"type": "float32", "shape": [384], "offset": 14762560}, "blocks.1.0.out": {"type": "float32", "shape": [384], "offset": 14764096}, "blocks.1.0.layers.0.0.qkvg": {"type": "float16", "shape": [384, 1544], "offset": 14765632}, "blocks.1.0.layers.0.0.gb": {"type": "float32", "shape": [8], "offset": 15951424}, "blocks.1.0.layers.0.0.wo": {"type": "float16", "shape": [512, 384], "offset": 15951456}, "blocks.1.0.layers.0.1.w1": {"type": "float16", "shape": [384, 1536], "offset": 16344672}, "blocks.1.0.layers.0.1.b1": {"type": "float32", "shape": [1536], "offset": 17524320}, "blocks.1.0.layers.0.1.w2": {"type": "float16", "shape": [1536, 384], "offset": 17530464}, "blocks.1.0.layers.0.1.b2": {"type": "float32", "shape": [384], "offset": 18710112}, "blocks.1.1.out": {"type": "float32", "shape": [384], "offset": 18711648}, "blocks.1.1.layers.0.0.qkvg": {"type": "float16", "shape": [384, 1544], "offset": 18713184}, "blocks.1.1.layers.0.0.gb": {"type": "float32", "shape": [8], "offset": 19898976}, "blocks.1.1.layers.0.0.wo": {"type": "float16", "shape": [512, 384], "offset": 19899008}, "blocks.1.1.layers.0.1.w1": {"type": "float16", "shape": [384, 1536], "offset": 20292224}, "blocks.1.1.layers.0.1.b1": {"type": "float32", "shape": [1536], "offset": 21471872}, "blocks.1.1.layers.0.1.w2": {"type": "float16", "shape": [1536, 384], "offset": 21478016}, "blocks.1.1.layers.0.1.b2": {"type": "float32", "shape": [384], "offset": 22657664}, "blocks.2.0.out": {"type": "float32", "shape": [384], "offset": 22659200}, "blocks.2.0.layers.0.0.qkvg": {"type": "float16", "shape": [384, 1544], "offset": 22660736}, "blocks.2.0.layers.0.0.gb": {"type": "float32", "shape": [8], "offset": 23846528}, "blocks.2.0.layers.0.0.wo": {"type": "float16", "shape": [512, 384], "offset": 23846560}, "blocks.2.0.layers.0.1.w1": {"type": "float16", "shape": [384, 1536], "offset": 24239776}, "blocks.2.0.layers.0.1.b1": {"type": "float32", "shape": [1536], "offset": 25419424}, "blocks.2.0.layers.0.1.w2": {"type": "float16", "shape": [1536, 384], "offset": 25425568}, "blocks.2.0.layers.0.1.b2": {"type": "float32", "shape": [384], "offset": 26605216}, "blocks.2.1.out": {"type": "float32", "shape": [384], "offset": 26606752}, "blocks.2.1.layers.0.0.qkvg": {"type": "float16", "shape": [384, 1544], "offset": 26608288}, "blocks.2.1.layers.0.0.gb": {"type": "float32", "shape": [8], "offset": 27794080}, "blocks.2.1.layers.0.0.wo": {"type": "float16", "shape": [512, 384], "offset": 27794112}, "blocks.2.1.layers.0.1.w1": {"type": "float16", "shape": [384, 1536], "offset": 28187328}, "blocks.2.1.layers.0.1.b1": {"type": "float32", "shape": [1536], "offset": 29366976}, "blocks.2.1.layers.0.1.w2": {"type": "float16", "shape": [1536, 384], "offset": 29373120}, "blocks.2.1.layers.0.1.b2": {"type": "float32", "shape": [384], "offset": 30552768}, "blocks.3.0.out": {"type": "float32", "shape": [384], "offset": 30554304}, "blocks.3.0.layers.0.0.qkvg": {"type": "float16", "shape": [384, 1544], "offset": 30555840}, "blocks.3.0.layers.0.0.gb": {"type": "float32", "shape": [8], "offset": 31741632}, "blocks.3.0.layers.0.0.wo": {"type": "float16", "shape": [512, 384], "offset": 31741664}, "blocks.3.0.layers.0.1.w1": {"type": "float16", "shape": [384, 1536], "offset": 32134880}, "blocks.3.0.layers.0.1.b1": {"type": "float32", "shape": [1536], "offset": 33314528}, "blocks.3.0.layers.0.1.w2": {"type": "float16", "shape": [1536, 384], "offset": 33320672}, "blocks.3.0.layers.0.1.b2": {"type": "float32", "shape": [384], "offset": 34500320}, "blocks.3.1.out": {"type": "float32", "shape": [384], "offset": 34501856}, "blocks.3.1.layers.0.0.qkvg": {"type": "float16", "shape": [384, 1544], "offset": 34503392}, "blocks.3.1.layers.0.0.gb": {"type": "float32", "shape": [8], "offset": 35689184}, "blocks.3.1.layers.0.0.wo": {"type": "float16", "shape": [512, 384], "offset": 35689216}, "blocks.3.1.layers.0.1.w1": {"type": "float16", "shape": [384, 1536], "offset": 36082432}, "blocks.3.1.layers.0.1.b1": {"type": "float32", "shape": [1536], "offset": 37262080}, "blocks.3.1.layers.0.1.w2": {"type": "float16", "shape": [1536, 384], "offset": 37268224}, "blocks.3.1.layers.0.1.b2": {"type": "float32", "shape": [384], "offset": 38447872}, "blocks.4.0.out": {"type": "float32", "shape": [384], "offset": 38449408}, "blocks.4.0.layers.0.0.qkvg": {"type": "float16", "shape": [384, 1544], "offset": 38450944}, "blocks.4.0.layers.0.0.gb": {"type": "float32", "shape": [8], "offset": 39636736}, "blocks.4.0.layers.0.0.wo": {"type": "float16", "shape": [512, 384], "offset": 39636768}, "blocks.4.0.layers.0.1.w1": {"type": "float16", "shape": [384, 1536], "offset": 40029984}, "blocks.4.0.layers.0.1.b1": {"type": "float32", "shape": [1536], "offset": 41209632}, "blocks.4.0.layers.0.1.w2": {"type": "float16", "shape": [1536, 384], "offset": 41215776}, "blocks.4.0.layers.0.1.b2": {"type": "float32", "shape": [384], "offset": 42395424}, "blocks.4.1.out": {"type": "float32", "shape": [384], "offset": 42396960}, "blocks.4.1.layers.0.0.qkvg": {"type": "float16", "shape": [384, 1544], "offset": 42398496}, "blocks.4.1.layers.0.0.gb": {"type": "float32", "shape": [8], "offset": 43584288}, "blocks.4.1.layers.0.0.wo": {"type": "float16", "shape": [512, 384], "offset": 43584320}, "blocks.4.1.layers.0.1.w1": {"type": "float16", "shape": [384, 1536], "offset": 43977536}, "blocks.4.1.layers.0.1.b1": {"type": "float32", "shape": [1536], "offset": 45157184}, "blocks.4.1.layers.0.1.w2": {"type": "float16", "shape": [1536, 384], "offset": 45163328}, "blocks.4.1.layers.0.1.b2": {"type": "float32", "shape": [384], "offset": 46342976}, "blocks.5.0.out": {"type": "float32", "shape": [384], "offset": 46344512}, "blocks.5.0.layers.0.0.qkvg": {"type": "float16", "shape": [384, 1544], "offset": 46346048}, "blocks.5.0.layers.0.0.gb": {"type": "float32", "shape": [8], "offset": 47531840}, "blocks.5.0.layers.0.0.wo": {"type": "float16", "shape": [512, 384], "offset": 47531872}, "blocks.5.0.layers.0.1.w1": {"type": "float16", "shape": [384, 1536], "offset": 47925088}, "blocks.5.0.layers.0.1.b1": {"type": "float32", "shape": [1536], "offset": 49104736}, "blocks.5.0.layers.0.1.w2": {"type": "float16", "shape": [1536, 384], "offset": 49110880}, "blocks.5.0.layers.0.1.b2": {"type": "float32", "shape": [384], "offset": 50290528}, "blocks.5.1.out": {"type": "float32", "shape": [384], "offset": 50292064}, "blocks.5.1.layers.0.0.qkvg": {"type": "float16", "shape": [384, 1544], "offset": 50293600}, "blocks.5.1.layers.0.0.gb": {"type": "float32", "shape": [8], "offset": 51479392}, "blocks.5.1.layers.0.0.wo": {"type": "float16", "shape": [512, 384], "offset": 51479424}, "blocks.5.1.layers.0.1.w1": {"type": "float16", "shape": [384, 1536], "offset": 51872640}, "blocks.5.1.layers.0.1.b1": {"type": "float32", "shape": [1536], "offset": 53052288}, "blocks.5.1.layers.0.1.w2": {"type": "float16", "shape": [1536, 384], "offset": 53058432}, "blocks.5.1.layers.0.1.b2": {"type": "float32", "shape": [384], "offset": 54238080}, "mw.0": {"type": "float16", "shape": [60, 384, 1536], "offset": 54239616}, "mw.1": {"type": "float16", "shape": [60, 1536, 1536], "offset": 125018496}, "mb.0": {"type": "float32", "shape": [60, 1, 1536], "offset": 408134016}, "mb.1": {"type": "float32", "shape": [60, 1, 1536], "offset": 408502656}, "m2_w.0": {"type": "float16", "shape": [18, 1536, 56], "offset": 408871296}, "m2_w.1": {"type": "float16", "shape": [6, 1536, 88], "offset": 411967872}, "m2_w.2": {"type": "float16", "shape": [5, 1536, 128], "offset": 413589888}, "m2_w.3": {"type": "float16", "shape": [5, 1536, 176], "offset": 415555968}, "m2_w.4": {"type": "float16", "shape": [5, 1536, 248], "offset": 418259328}, "m2_w.5": {"type": "float16", "shape": [4, 1536, 328], "offset": 422068608}, "m2_w.6": {"type": "float16", "shape": [5, 1536, 456], "offset": 426099072}, "m2_w.7": {"type": "float16", "shape": [4, 1536, 608], "offset": 433103232}, "m2_w.8": {"type": "float16", "shape": [4, 1536, 792], "offset": 440574336}, "m2_w.9": {"type": "float16", "shape": [4, 1536, 1040], "offset": 450306432}, "m2_b.0": {"type": "float32", "shape": [18, 1, 56], "offset": 463085952}, "m2_b.1": {"type": "float32", "shape": [6, 1, 88], "offset": 463089984}, "m2_b.2": {"type": "float32", "shape": [5, 1, 128], "offset": 463092096}, "m2_b.3": {"type": "float32", "shape": [5, 1, 176], "offset": 463094656}, "m2_b.4": {"type": "float32", "shape": [5, 1, 248], "offset": 463098176}, "m2_b.5": {"type": "float32", "shape": [4, 1, 328], "offset": 463103136}, "m2_b.6": {"type": "float32", "shape": [5, 1, 456], "offset": 463108384}, "m2_b.7": {"type": "float32", "shape": [4, 1, 608], "offset": 463117504}, "m2_b.8": {"type": "float32", "shape": [4, 1, 792], "offset": 463127232}, "m2_b.9": {"type": "float32", "shape": [4, 1, 1040], "offset": 463139904}, "si0": {"type": "int32", "shape": [18, 28], "offset": 463156544}, "si1": {"type": "int32", "shape": [6, 44], "offset": 463158560}, "si2": {"type": "int32", "shape": [5, 64], "offset": 463159616}, "si3": {"type": "int32", "shape": [5, 88], "offset": 463160896}, "si4": {"type": "int32", "shape": [5, 124], "offset": 463162656}, "si5": {"type": "int32", "shape": [4, 164], "offset": 463165136}, "si6": {"type": "int32", "shape": [5, 228], "offset": 463167760}, "si7": {"type": "int32", "shape": [4, 304], "offset": 463172320}, "si8": {"type": "int32", "shape": [4, 396], "offset": 463177184}, "si9": {"type": "int32", "shape": [4, 520], "offset": 463183520}, "i1": {"type": "int32", "shape": [4100], "offset": 463191840}, "i2": {"type": "int32", "shape": [4100], "offset": 463208240}, "wt": {"type": "float32", "shape": [4100], "offset": 463224640}, "blocks.0.0.layers.0.0.cos": {"type": "float32", "shape": [801, 64], "offset": 463241040}, "blocks.0.0.layers.0.0.sin": {"type": "float32", "shape": [801, 64], "offset": 463446096}, "blocks.0.0.layers.0.0.R": {"type": "float32", "shape": [64, 64], "offset": 463651152}, "blocks.0.1.layers.0.0.cos": {"type": "float32", "shape": [60, 64], "offset": 463667536}, "blocks.0.1.layers.0.0.sin": {"type": "float32", "shape": [60, 64], "offset": 463682896}, "blocks.0.1.layers.0.0.R": {"type": "float32", "shape": [64, 64], "offset": 463698256}, "blocks.1.0.layers.0.0.cos": {"type": "float32", "shape": [801, 64], "offset": 463714640}, "blocks.1.0.layers.0.0.sin": {"type": "float32", "shape": [801, 64], "offset": 463919696}, "blocks.1.0.layers.0.0.R": {"type": "float32", "shape": [64, 64], "offset": 464124752}, "blocks.1.1.layers.0.0.cos": {"type": "float32", "shape": [60, 64], "offset": 464141136}, "blocks.1.1.layers.0.0.sin": {"type": "float32", "shape": [60, 64], "offset": 464156496}, "blocks.1.1.layers.0.0.R": {"type": "float32", "shape": [64, 64], "offset": 464171856}, "blocks.2.0.layers.0.0.cos": {"type": "float32", "shape": [801, 64], "offset": 464188240}, "blocks.2.0.layers.0.0.sin": {"type": "float32", "shape": [801, 64], "offset": 464393296}, "blocks.2.0.layers.0.0.R": {"type": "float32", "shape": [64, 64], "offset": 464598352}, "blocks.2.1.layers.0.0.cos": {"type": "float32", "shape": [60, 64], "offset": 464614736}, "blocks.2.1.layers.0.0.sin": {"type": "float32", "shape": [60, 64], "offset": 464630096}, "blocks.2.1.layers.0.0.R": {"type": "float32", "shape": [64, 64], "offset": 464645456}, "blocks.3.0.layers.0.0.cos": {"type": "float32", "shape": [801, 64], "offset": 464661840}, "blocks.3.0.layers.0.0.sin": {"type": "float32", "shape": [801, 64], "offset": 464866896}, "blocks.3.0.layers.0.0.R": {"type": "float32", "shape": [64, 64], "offset": 465071952}, "blocks.3.1.layers.0.0.cos": {"type": "float32", "shape": [60, 64], "offset": 465088336}, "blocks.3.1.layers.0.0.sin": {"type": "float32", "shape": [60, 64], "offset": 465103696}, "blocks.3.1.layers.0.0.R": {"type": "float32", "shape": [64, 64], "offset": 465119056}, "blocks.4.0.layers.0.0.cos": {"type": "float32", "shape": [801, 64], "offset": 465135440}, "blocks.4.0.layers.0.0.sin": {"type": "float32", "shape": [801, 64], "offset": 465340496}, "blocks.4.0.layers.0.0.R": {"type": "float32", "shape": [64, 64], "offset": 465545552}, "blocks.4.1.layers.0.0.cos": {"type": "float32", "shape": [60, 64], "offset": 465561936}, "blocks.4.1.layers.0.0.sin": {"type": "float32", "shape": [60, 64], "offset": 465577296}, "blocks.4.1.layers.0.0.R": {"type": "float32", "shape": [64, 64], "offset": 465592656}, "blocks.5.0.layers.0.0.cos": {"type": "float32", "shape": [801, 64], "offset": 465609040}, "blocks.5.0.layers.0.0.sin": {"type": "float32", "shape": [801, 64], "offset": 465814096}, "blocks.5.0.layers.0.0.R": {"type": "float32", "shape": [64, 64], "offset": 466019152}, "blocks.5.1.layers.0.0.cos": {"type": "float32", "shape": [60, 64], "offset": 466035536}, "blocks.5.1.layers.0.0.sin": {"type": "float32", "shape": [60, 64], "offset": 466050896}, "blocks.5.1.layers.0.0.R": {"type": "float32", "shape": [64, 64], "offset": 466066256}}}