compressionkit E2A2 codec, redistributed from danjacobellis/E2A2 with the author's permission

.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,72 @@
+---
+license: other
+license_name: redistributed-with-permission
+license_link: https://huggingface.co/danjacobellis/E2A2/tree/aafe5975f62631d55b418d997a63d81abae11494/spatial_audio
+---
+
+# E2A2 `spatial_audio` codec for compressionkit
+
+E2A2 (*Encoding Efficient Asymmetric Autoencoders*) `spatial_audio` is an audio codec for the seven-microphone array of Project Aria glasses at 48 kHz, the seven channels coded together. This repository holds its weights in the layout that [compressionkit](https://github.com/AmbiqAI/compressionkit) v0.3 reads, with an encoder half and a decoder half that load separately.
+
+## Source and permission
+
+The weights are redistributed from their original public release, [`danjacobellis/E2A2`, subfolder `spatial_audio/`, revision `aafe5975`](https://huggingface.co/danjacobellis/E2A2/tree/aafe5975f62631d55b418d997a63d81abae11494/spatial_audio), with the permission of the author, Dan Jacobellis. Ambiq did not train this codec. Its description, development and evaluation are documented at the original source and on the codec's page, <https://danjacobellis.net/compressors/E2A2_spatial_audio>. The entropy coder is GGDLPC (<https://danjacobellis.net/GGDLPC/>).
+
+Every tensor here equals the tensor of the same name at that revision; this repository divides them into the encoder and decoder halves listed below and adds the files compressionkit uses.
+
+## Files
+
+- `encoder/config.json`, `encoder/weights.safetensors` (the `merged.{op}.encoders.*` tensors) and `encoder/entropy.json` (the GGDLPC tables): everything needed to encode.
+- `decoder/config.json`, `decoder/weights.safetensors` (the `merged.{op}.decoder.*` tensors): everything needed to decode. Each `config.json` is the full source configuration, so each half constructs alone.
+- `rd.json`: the rate–distortion curve of the original release, byte-identical to it. The validation clips were coded as stored (no normalization), so the input scaling is fixed. Rates are bits per frame, the seven channels summed.
+- `web/`: the decoder of each operating point that the compressionkit browser demo offers, as LiteRT files (section below). The compressionkit loaders do not download `web/`.
+
+| file | bytes | SHA-256 |
+|---|---|---|
+| `decoder/config.json` | 1,536 | `6681cc25f7b190e8cccc9e489aa58020f59ca6c5d7813fe489089078ef1858d5` |
+| `decoder/weights.safetensors` | 2,450,820,696 | `28d762dfc52da141c488735effddff82d61984ebb19a1c61a84a9dc0cb3144b4` |
+| `encoder/config.json` | 1,536 | `6681cc25f7b190e8cccc9e489aa58020f59ca6c5d7813fe489089078ef1858d5` |
+| `encoder/entropy.json` | 17,826 | `092f7746670780573eb349c5418dafbf72c2dfe2ed85678c527a5e0aacfa5607` |
+| `encoder/weights.safetensors` | 513,704 | `65e860cf924da8fc4cae4f2ee48b7f82af1928a493ac7e0fb50e440a472c1296` |
+| `rd.json` | 578 | `9796ac3c08304da1698571a86444b19a98d10b5b1ac9b4cffa4e9d78893b9ee1` |
+
+## Input, output and operating points
+
+Input: float32 `(7, L)` samples at 48 kHz in the scale of the recordings it was developed on (no input map is published). Output: a reconstruction in the same form. Operating points (cumulative latent channels): `[16, 32, 56, 80]`. The stream of a lower operating point is a literal bit-prefix of the stream of the highest, and the rates are entropy-coded bits. compressionkit converts nothing: each codec states its input type and range.
+
+## Using it with compressionkit
+
+compressionkit v0.3 is in pre-release; these loaders are part of it.
+
+```python
+from compressionkit.e2a2 import hub
+
+# both halves
+config, weights, coder, ops = hub.load_split_from_hub('spatial_audio')
+model = hub.load_codec(config=config, weights=weights, n_ch=ops[-1])
+
+# one half (encode-only shown; half='decoder' for decode-only)
+cfg, w, coder, ops = hub.load_split_from_hub('spatial_audio', half='encoder')
+encoder = hub.build_partial(cfg, w, ops[-1])
+```
+
+## Status and limitations
+
+- A reference codec for evaluation and development with compressionkit, redistributed as released; not a production model, and not qualified for any product or regulated use.
+- Its quality on data unlike the data of its original release is not characterized here; measure it on your own data (`compressionkit.calibrate` fits an input map on your training split).
+- Maintainer for compressionkit: Dan Jacobellis (Ambiq). Questions about the codec itself: the original source.
+
+<!-- web/: hub/e2a2_web.py -->
+## Browser decoders (`web/`)
+
+For each operating point the compressionkit browser demo offers, its decoder as a LiteRT `.tflite` window decoder, which runs in a web browser on LiteRT.js (CPU), with a `.json` beside it that gives its shapes and windowing. A window decodes `chunk` output samples (pixels along each side for the image codec) with `margin` of context on each side; the latent grid is cut into overlapping windows and the outputs joined. The PSNR column is the largest difference between the PSNR of the joined output, decoded in LiteRT.js in Chromium, and that of the full-size PyTorch decode, over the example recordings of the demo. The compressionkit loaders do not download `web/`.
+
+| file | operating point | chunk | margin | bytes | PSNR difference (dB, rows) | decoder weights (SHA-256) |
+|---|---|---|---|---|---|---|
+| `web/spatial_audio_op16_chunk16384_m4096.tflite` | 16 | 16384 samples | 4096 | 410,460,028 | 3.3e-09 (1) | `28d762dfc52d` |
+| `web/spatial_audio_op32_chunk16384_m4096.tflite` | 32 | 16384 samples | 4096 | 410,656,636 | 3.9e-10 (1) | `28d762dfc52d` |
+| `web/spatial_audio_op56_chunk16384_m4096.tflite` | 56 | 16384 samples | 4096 | 814,609,900 | 8.0e-09 (1) | `28d762dfc52d` |
+| `web/spatial_audio_op80_chunk16384_m4096.tflite` | 80 | 16384 samples | 4096 | 815,789,548 | 3.4e-07 (1) | `28d762dfc52d` |
+
+Weights fp32, converted from `decoder/weights.safetensors` with litert-torch 0.10.0. The last column gives the start of the SHA-256 of the `decoder/weights.safetensors` it was converted from, which is this repository's.
+<!-- /web/ -->

decoder/config.json

@@ -0,0 +1,81 @@
+{
+  "codec": "E2A2",
+  "format": "e2a2_hub1",
+  "source_checkpoint": "e2a2s2_g4.pth",
+  "modality": "audio",
+  "dim": 1,
+  "input_channels": 7,
+  "ps": [
+    128,
+    128,
+    64,
+    32
+  ],
+  "group_sizes": [
+    16,
+    16,
+    24,
+    24
+  ],
+  "cumulative_channels": [
+    16,
+    32,
+    56,
+    80
+  ],
+  "n_groups": 4,
+  "affine": true,
+  "linear_input": false,
+  "decoder_configs": [
+    {
+      "decoder_ps": 128,
+      "decoder_dim": 1024,
+      "decoder_kernel_size": 3,
+      "decoder_depth": 12,
+      "decoder_mlp_ratio": 4,
+      "decoder_layerscale": true,
+      "decoder_layerscale_init": 1e-06
+    },
+    {
+      "decoder_ps": 128,
+      "decoder_dim": 1024,
+      "decoder_kernel_size": 3,
+      "decoder_depth": 12,
+      "decoder_mlp_ratio": 4,
+      "decoder_layerscale": true,
+      "decoder_layerscale_init": 1e-06
+    },
+    {
+      "decoder_ps": 128,
+      "decoder_dim": 1024,
+      "decoder_kernel_size": 3,
+      "decoder_depth": 24,
+      "decoder_mlp_ratio": 4,
+      "decoder_layerscale": true,
+      "decoder_layerscale_init": 1e-06
+    },
+    {
+      "decoder_ps": 128,
+      "decoder_dim": 1024,
+      "decoder_kernel_size": 3,
+      "decoder_depth": 24,
+      "decoder_mlp_ratio": 4,
+      "decoder_layerscale": true,
+      "decoder_layerscale_init": 1e-06
+    }
+  ],
+  "macroregion": 16384,
+  "stream_mode": "all_detail",
+  "compander": "softsign",
+  "compander_bits": 6,
+  "alphabet": [
+    -31,
+    31
+  ],
+  "input_range": [
+    -1,
+    1
+  ],
+  "sample_rate": 48000,
+  "val_normalize": "none"
+}

decoder/weights.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28d762dfc52da141c488735effddff82d61984ebb19a1c61a84a9dc0cb3144b4
+size 2450820696

encoder/config.json

@@ -0,0 +1,81 @@
+{
+  "codec": "E2A2",
+  "format": "e2a2_hub1",
+  "source_checkpoint": "e2a2s2_g4.pth",
+  "modality": "audio",
+  "dim": 1,
+  "input_channels": 7,
+  "ps": [
+    128,
+    128,
+    64,
+    32
+  ],
+  "group_sizes": [
+    16,
+    16,
+    24,
+    24
+  ],
+  "cumulative_channels": [
+    16,
+    32,
+    56,
+    80
+  ],
+  "n_groups": 4,
+  "affine": true,
+  "linear_input": false,
+  "decoder_configs": [
+    {
+      "decoder_ps": 128,
+      "decoder_dim": 1024,
+      "decoder_kernel_size": 3,
+      "decoder_depth": 12,
+      "decoder_mlp_ratio": 4,
+      "decoder_layerscale": true,
+      "decoder_layerscale_init": 1e-06
+    },
+    {
+      "decoder_ps": 128,
+      "decoder_dim": 1024,
+      "decoder_kernel_size": 3,
+      "decoder_depth": 12,
+      "decoder_mlp_ratio": 4,
+      "decoder_layerscale": true,
+      "decoder_layerscale_init": 1e-06
+    },
+    {
+      "decoder_ps": 128,
+      "decoder_dim": 1024,
+      "decoder_kernel_size": 3,
+      "decoder_depth": 24,
+      "decoder_mlp_ratio": 4,
+      "decoder_layerscale": true,
+      "decoder_layerscale_init": 1e-06
+    },
+    {
+      "decoder_ps": 128,
+      "decoder_dim": 1024,
+      "decoder_kernel_size": 3,
+      "decoder_depth": 24,
+      "decoder_mlp_ratio": 4,
+      "decoder_layerscale": true,
+      "decoder_layerscale_init": 1e-06
+    }
+  ],
+  "macroregion": 16384,
+  "stream_mode": "all_detail",
+  "compander": "softsign",
+  "compander_bits": 6,
+  "alphabet": [
+    -31,
+    31
+  ],
+  "input_range": [
+    -1,
+    1
+  ],
+  "sample_rate": 48000,
+  "val_normalize": "none"
+}

encoder/entropy.json

@@ -0,0 +1 @@
+{"spec": "e2a2 stream layout v1 over [GGDLPC v1: unified coder (one learned causal FIR predictor per channel with arithmetic-mean fallback; one stateless pooled statistic: mean of the causal template abs differences and the last 16 coded abs residuals in scan order; two per-channel affine laws -> analytic folded GGD conditionals on the tau=2^(1/4) scale grid; canonical Huffman tables under the derived quotient split, <= 63 windowed hi values + escape at any bit depth; derived zero-run mode with elementary Golomb codes and exclusion tables, always available); inputs in [-N, N], N <= 65536; one call = one byte-padded bitstream, channel prefixes are literal bit prefixes]: two_stream = global[first scale group, whole plane, one pad] + per-macroregion detail streams (scale groups >= 1, coarse-to-fine, channel-major, one pad per region); all_detail = NO global stream, every channel per macroregion, each region decodable from its own bytes alone; one GGDLPC channel record per latent channel (global index); op-point streams are literal bit-prefixes", "version": 5, "stream_mode": "all_detail", "rank": 1, "macroregion": 16384, "N": 31, "scale_groups": [[128, 0, 32], [64, 32, 56], [32, 56, 80]], "channels": [{"channel": 0, "mu": 0.0005708670910493827, "a": 0.9017839311968358, "b": 0.0956995194041974, "u": 0.9121752076531893, "v": 0.20170438466836413, "wq": [-11, 0], "bq": 0, "n_samples": 6635520}, {"channel": 1, "mu": 0.0006474247685185185, "a": 0.9000958047565498, "b": 0.0625, "u": 1.0177904504107034, "v": 0.20673688190050246, "wq": [27, -8], "bq": 0, "n_samples": 6635520}, {"channel": 2, "mu": 0.008614547164351852, "a": 0.9027381136555681, "b": 0.12989438289956107, "u": 1.0123520715491947, "v": 0.28729721068330993, "wq": [-3, -1], "bq": 2, "n_samples": 6635520}, {"channel": 3, "mu": -0.013762598861882715, "a": 0.883406981827356, "b": 0.0690563803551493, "u": 1.0046127446311997, "v": 0.20829629798108376, "wq": [1, 0], "bq": -4, "n_samples": 6635520}, {"channel": 4, "mu": -0.005915436921296297, "a": 0.9029219744911943, "b": 0.11947185022601119, "u": 1.041006630589943, "v": 0.28461826910505683, "wq": [-8, -1], "bq": -2, "n_samples": 6635520}, {"channel": 5, "mu": 0.012428867669753086, "a": 0.9022576681316609, "b": 0.08653790694784817, "u": 0.9653389593427275, "v": 0.20332482227229173, "wq": [41, -13], "bq": 3, "n_samples": 6635520}, {"channel": 6, "mu": -0.01597071518132716, "a": 0.8811534124632248, "b": 0.07137113992335331, "u": 1.0162746220125527, "v": 0.21267515726617103, "wq": [-4, 0], "bq": -4, "n_samples": 6635520}, {"channel": 7, "mu": -0.016415593653549384, "a": 0.9056088196213486, "b": 0.0791645301899263, "u": 0.8332296187309466, "v": 0.17385373409123112, "wq": [-15, -2], "bq": -5, "n_samples": 6635520}, {"channel": 8, "mu": -0.017731843171296297, "a": 0.9051031985740728, "b": 0.07712130388760255, "u": 0.8152351181644432, "v": 0.16060257996371796, "wq": [-18, -4], "bq": -5, "n_samples": 6635520}, {"channel": 9, "mu": -0.019217182677469136, "a": 0.9008073376215292, "b": 0.13166249417908607, "u": 1.0025452922969091, "v": 0.2672102911173681, "wq": [-2, -2], "bq": -5, "n_samples": 6635520}, {"channel": 10, "mu": -0.001075122974537037, "a": 0.8913750479672228, "b": 0.06559722757709947, "u": 0.9358189270505027, "v": 0.17289027697985943, "wq": [-24, -8], "bq": -1, "n_samples": 6635520}, {"channel": 11, "mu": -0.008342676986882716, "a": 0.8844468757955307, "b": 0.06560239396956605, "u": 0.9672061603973354, "v": 0.17519294222788603, "wq": [-1, -2], "bq": -2, "n_samples": 6635520}, {"channel": 12, "mu": 0.020107241030092593, "a": 0.9016484788947234, "b": 0.13345052517654002, "u": 1.0416785384702674, "v": 0.2936750016535386, "wq": [2, 0], "bq": 5, "n_samples": 6635520}, {"channel": 13, "mu": 0.028784631799768517, "a": 0.9019688242613666, "b": 0.1086271473646514, "u": 0.9327420840385914, "v": 0.2249384742790021, "wq": [8, 0], "bq": 7, "n_samples": 6635520}, {"channel": 14, "mu": 0.013662229938271604, "a": 0.9045345353132669, "b": 0.09424045558000756, "u": 0.9350991044950185, "v": 0.2116681310524888, "wq": [10, 0], "bq": 4, "n_samples": 6635520}, {"channel": 15, "mu": -0.008098084249614198, "a": 0.8975112522795058, "b": 0.10002378613041828, "u": 0.922008856536483, "v": 0.20246070696647564, "wq": [-19, -2], "bq": -2, "n_samples": 6635520}, {"channel": 16, "mu": 0.0019912531346450617, "a": 0.893283001231819, "b": 0.12479329785409486, "u": 1.0841928813254051, "v": 0.30993667545523423, "wq": [-5, 0], "bq": 1, "n_samples": 6635520}, {"channel": 17, "mu": -5.621262538580247e-05, "a": 0.8932123222229784, "b": 0.12968381211125707, "u": 1.0813217460203617, "v": 0.2988667294406793, "wq": [-3, -2], "bq": 0, "n_samples": 6635520}, {"channel": 18, "mu": 0.003959900655864198, "a": 0.8963487477884218, "b": 0.12822240837361706, "u": 1.0994057437289404, "v": 0.3066824710583574, "wq": [4, 0], "bq": 1, "n_samples": 6635520}, {"channel": 19, "mu": 0.00643506462191358, "a": 0.8917647541158426, "b": 0.12697035387624955, "u": 1.0751986183370517, "v": 0.3060002993887776, "wq": [-4, 0], "bq": 2, "n_samples": 6635520}, {"channel": 20, "mu": 0.0008303795331790123, "a": 0.8935171468959797, "b": 0.1265912747966758, "u": 1.0832459142427722, "v": 0.31472444170969627, "wq": [-3, 0], "bq": 0, "n_samples": 6635520}, {"channel": 21, "mu": 0.006588782793209877, "a": 0.8957557763862051, "b": 0.12379824202266451, "u": 1.08750756802675, "v": 0.31031738875372084, "wq": [5, -1], "bq": 2, "n_samples": 6635520}, {"channel": 22, "mu": 0.0008263105227623457, "a": 0.8968198756397274, "b": 0.124933600415919, "u": 1.0522012571342017, "v": 0.30055057357859133, "wq": [4, 0], "bq": 0, "n_samples": 6635520}, {"channel": 23, "mu": -0.004537549430941358, "a": 0.8955751900434158, "b": 0.12721974252848425, "u": 1.0568388459853966, "v": 0.30154875528822833, "wq": [5, 0], "bq": -1, "n_samples": 6635520}, {"channel": 24, "mu": 0.007460003134645061, "a": 0.8931192075006416, "b": 0.12657848035216915, "u": 1.076145855617438, "v": 0.2967901585165941, "wq": [0, 0], "bq": 2, "n_samples": 6635520}, {"channel": 25, "mu": -0.008667896412037038, "a": 0.894266021148531, "b": 0.13036831559406864, "u": 1.0951671803769971, "v": 0.3024504208725084, "wq": [2, 0], "bq": -2, "n_samples": 6635520}, {"channel": 26, "mu": 0.004537549430941358, "a": 0.8960595193451824, "b": 0.12152178727649243, "u": 1.079353177123968, "v": 0.29426749702270966, "wq": [1, -1], "bq": 1, "n_samples": 6635520}, {"channel": 27, "mu": -0.00047170380015432097, "a": 0.894006132480591, "b": 0.1250313110176981, "u": 1.0874586185719775, "v": 0.31529226612375527, "wq": [-1, 0], "bq": 0, "n_samples": 6635520}, {"channel": 28, "mu": 0.0008368598090277777, "a": 0.8937318612154285, "b": 0.12549239575197768, "u": 1.07496123839384, "v": 0.30893771372489426, "wq": [4, 0], "bq": 0, "n_samples": 6635520}, {"channel": 29, "mu": 0.007621557918595679, "a": 0.8952147553118671, "b": 0.12562129363421945, "u": 1.0880428269416675, "v": 0.31332876179477903, "wq": [7, 0], "bq": 2, "n_samples": 6635520}, {"channel": 30, "mu": -0.003152126736111111, "a": 0.8936993631937385, "b": 0.12522671642655336, "u": 1.0841332091000602, "v": 0.3098755522234649, "wq": [0, 0], "bq": -1, "n_samples": 6635520}, {"channel": 31, "mu": -0.003377881462191358, "a": 0.8944204644139522, "b": 0.1271427030370615, "u": 1.068129645630903, "v": 0.306100195115895, "wq": [3, 0], "bq": -1, "n_samples": 6635520}, {"channel": 32, "mu": -0.0015396683304398147, "a": 0.9190815232124077, "b": 0.10791368535300336, "u": 1.2422913910381124, "v": 0.3289416816734312, "wq": [3, -1], "bq": 0, "n_samples": 13271040}, {"channel": 33, "mu": -0.0014427656008873457, "a": 0.9120542607058544, "b": 0.09675840387931034, "u": 1.1613286527586177, "v": 0.305561936388014, "wq": [-1, -1], "bq": 0, "n_samples": 13271040}, {"channel": 34, "mu": -0.000955614631558642, "a": 0.9133731558065888, "b": 0.09865635294224184, "u": 1.1967014148069492, "v": 0.3216095404057031, "wq": [-3, 0], "bq": 0, "n_samples": 13271040}, {"channel": 35, "mu": -0.0019222306616512345, "a": 0.9175128600401602, "b": 0.09549066097928809, "u": 1.1892381799310545, "v": 0.3046718112215645, "wq": [8, 0], "bq": 0, "n_samples": 13271040}, {"channel": 36, "mu": -0.0015629521122685185, "a": 0.9157957608516474, "b": 0.10895195626597323, "u": 1.1705356327749712, "v": 0.32888427194756054, "wq": [-8, 0], "bq": 0, "n_samples": 13271040}, {"channel": 37, "mu": -0.0011656207802854938, "a": 0.9157927268454161, "b": 0.09795459355653048, "u": 1.1619039511705402, "v": 0.3073402113634364, "wq": [16, -1], "bq": 0, "n_samples": 13271040}, {"channel": 38, "mu": -0.001116868007330247, "a": 0.9186568769092013, "b": 0.10736350270962203, "u": 1.2315117073170931, "v": 0.3226947587164761, "wq": [9, 0], "bq": 0, "n_samples": 13271040}, {"channel": 39, "mu": 0.0007100423177083333, "a": 0.9194247418745797, "b": 0.10900399041638789, "u": 1.243738735496971, "v": 0.33354784823162525, "wq": [2, -4], "bq": 0, "n_samples": 13271040}, {"channel": 40, "mu": 0.001866771556712963, "a": 0.9181852289564695, "b": 0.10878881796078374, "u": 1.251894804203897, "v": 0.33408482720545357, "wq": [-1, -1], "bq": 0, "n_samples": 13271040}, {"channel": 41, "mu": 0.0016789189091435184, "a": 0.9144394891285378, "b": 0.09414733317225474, "u": 1.1647065252307416, "v": 0.29217256450456675, "wq": [-4, -1], "bq": 0, "n_samples": 13271040}, {"channel": 42, "mu": -0.0028339150511188273, "a": 0.9153415951947353, "b": 0.11006300389140697, "u": 1.2223894775708308, "v": 0.31747100878055745, "wq": [7, 0], "bq": -1, "n_samples": 13271040}, {"channel": 43, "mu": -0.0005576804832175926, "a": 0.917311164490449, "b": 0.09678419179342318, "u": 1.1968817003379357, "v": 0.3202227341917857, "wq": [12, -2], "bq": 0, "n_samples": 13271040}, {"channel": 44, "mu": 0.0007856957706404321, "a": 0.9197676902831594, "b": 0.10737726341446957, "u": 1.2140566991512924, "v": 0.31781991334531023, "wq": [4, 0], "bq": 0, "n_samples": 13271040}, {"channel": 45, "mu": -0.0035508897569444444, "a": 0.9156349506865017, "b": 0.10987985392507134, "u": 1.113412947668896, "v": 0.3080155051615178, "wq": [-18, -1], "bq": -1, "n_samples": 13271040}, {"channel": 46, "mu": -0.0009269808545524691, "a": 0.9195521503055968, "b": 0.103267781376935, "u": 1.192832734556208, "v": 0.29828545156150726, "wq": [29, -3], "bq": 0, "n_samples": 13271040}, {"channel": 47, "mu": -0.0005729015962577161, "a": 0.9187724888227724, "b": 0.09947471103309431, "u": 1.2252656087174003, "v": 0.3384653253112798, "wq": [13, 0], "bq": 0, "n_samples": 13271040}, {"channel": 48, "mu": 0.0012257517119984567, "a": 0.920724471657152, "b": 0.1075341896748716, "u": 1.248101185652461, "v": 0.3311712591517346, "wq": [22, -5], "bq": 0, "n_samples": 13271040}, {"channel": 49, "mu": 0.0010852201485339506, "a": 0.9187753193810293, "b": 0.10083503925880317, "u": 1.2059434071500523, "v": 0.3303177619347161, "wq": [20, -3], "bq": 0, "n_samples": 13271040}, {"channel": 50, "mu": -0.0022671923225308643, "a": 0.9166497852117818, "b": 0.10142476078911183, "u": 1.1729261862147995, "v": 0.3174773673246936, "wq": [14, -1], "bq": -1, "n_samples": 13271040}, {"channel": 51, "mu": 0.001975203149112654, "a": 0.9219614493500042, "b": 0.10809824705331973, "u": 1.2599916793253507, "v": 0.34163202257421776, "wq": [10, -1], "bq": 1, "n_samples": 13271040}, {"channel": 52, "mu": -0.0009046766493055556, "a": 0.9187461809769248, "b": 0.10393228894790936, "u": 1.2216397865596564, "v": 0.3122362361303447, "wq": [12, -2], "bq": 0, "n_samples": 13271040}, {"channel": 53, "mu": -0.0003352412471064815, "a": 0.9196127119269202, "b": 0.10113438295620247, "u": 1.2437370159427232, "v": 0.3513079936765372, "wq": [7, -1], "bq": 0, "n_samples": 13271040}, {"channel": 54, "mu": -0.004091239269868827, "a": 0.9168832981373632, "b": 0.09809305918818954, "u": 1.1972930694044142, "v": 0.32125833882543925, "wq": [10, -4], "bq": -1, "n_samples": 13271040}, {"channel": 55, "mu": 0.00022507655767746914, "a": 0.9135350299993062, "b": 0.09831628317005055, "u": 1.187983158080647, "v": 0.31940604964569347, "wq": [-2, 0], "bq": 0, "n_samples": 13271040}, {"channel": 56, "mu": -0.00047132703993055556, "a": 0.9304574513781986, "b": 0.0625, "u": 0.981639852675042, "v": 0.18144680841300803, "wq": [19, 9], "bq": 0, "n_samples": 26542080}, {"channel": 57, "mu": -0.00025465223524305557, "a": 0.9259653799369353, "b": 0.0625, "u": 1.0380610615085113, "v": 0.2061933779077613, "wq": [-2, 7], "bq": 0, "n_samples": 26542080}, {"channel": 58, "mu": 0.00016569914641203704, "a": 0.9278805758936008, "b": 0.0625, "u": 1.0446699149328802, "v": 0.20737466295873894, "wq": [9, 10], "bq": 0, "n_samples": 26542080}, {"channel": 59, "mu": 0.0014611514998070989, "a": 0.9290097986159063, "b": 0.0625, "u": 0.9873311454104807, "v": 0.18110690846880173, "wq": [12, 10], "bq": 0, "n_samples": 26542080}, {"channel": 60, "mu": 0.0008014066719714507, "a": 0.9272755624760476, "b": 0.0625, "u": 1.045643672200729, "v": 0.21027082564983615, "wq": [7, 9], "bq": 0, "n_samples": 26542080}, {"channel": 61, "mu": -0.0009562927999614198, "a": 0.9256489336558488, "b": 0.0625, "u": 1.0537962361539248, "v": 0.21092859567817465, "wq": [-3, 10], "bq": 0, "n_samples": 26542080}, {"channel": 62, "mu": -0.0009296935281635802, "a": 0.9258648748779614, "b": 0.0625, "u": 1.038751317300205, "v": 0.20750639560703818, "wq": [3, 8], "bq": 0, "n_samples": 26542080}, {"channel": 63, "mu": 0.001507229275173611, "a": 0.9278768109762393, "b": 0.0625, "u": 1.0108876179981883, "v": 0.200180463070821, "wq": [7, 7], "bq": 0, "n_samples": 26542080}, {"channel": 64, "mu": 0.0020052686149691357, "a": 0.9303262250849217, "b": 0.0625, "u": 0.9908005475918162, "v": 0.18491811475814984, "wq": [16, 9], "bq": 0, "n_samples": 26542080}, {"channel": 65, "mu": 0.0012848653911072532, "a": 0.9300916252234339, "b": 0.0625, "u": 1.0358850275432645, "v": 0.2102015599049638, "wq": [11, 4], "bq": 0, "n_samples": 26542080}, {"channel": 66, "mu": -0.0014217423804012345, "a": 0.9284328182515208, "b": 0.0625, "u": 1.0390051282991404, "v": 0.2084356370201183, "wq": [9, 7], "bq": 0, "n_samples": 26542080}, {"channel": 67, "mu": -0.0011259102527006173, "a": 0.9278613026586803, "b": 0.0625, "u": 1.1113545021076372, "v": 0.24300497288946993, "wq": [9, 9], "bq": 0, "n_samples": 26542080}, {"channel": 68, "mu": -0.0007314799744405864, "a": 0.9330427402212317, "b": 0.0625, "u": 1.0042483039034211, "v": 0.21343635270678035, "wq": [12, 5], "bq": 0, "n_samples": 26542080}, {"channel": 69, "mu": -0.0009152636115933642, "a": 0.928672587704032, "b": 0.0625, "u": 1.1189668258840846, "v": 0.24536485423674761, "wq": [22, 4], "bq": 0, "n_samples": 26542080}, {"channel": 70, "mu": -0.0008722375940393519, "a": 0.9296460116701226, "b": 0.0625, "u": 0.99478033836847, "v": 0.1840242564264549, "wq": [15, 11], "bq": 0, "n_samples": 26542080}, {"channel": 71, "mu": 0.00031613950376157407, "a": 0.9292915619822829, "b": 0.0625, "u": 1.0048328153830333, "v": 0.18724430574439163, "wq": [12, 8], "bq": 0, "n_samples": 26542080}, {"channel": 72, "mu": -0.0005227171344521605, "a": 0.9307023294797813, "b": 0.0625, "u": 1.0143700013149402, "v": 0.2029689923098109, "wq": [5, 4], "bq": 0, "n_samples": 26542080}, {"channel": 73, "mu": 0.0005608829451195988, "a": 0.9319836787832358, "b": 0.0625, "u": 1.0118573841819418, "v": 0.20349468709537513, "wq": [15, 8], "bq": 0, "n_samples": 26542080}, {"channel": 74, "mu": 0.00021818184558256172, "a": 0.9320249648422837, "b": 0.0625, "u": 0.9943636938074081, "v": 0.19817041226433926, "wq": [6, 6], "bq": 0, "n_samples": 26542080}, {"channel": 75, "mu": -0.0005164252387152777, "a": 0.9312516572211441, "b": 0.0625, "u": 0.9991513374766231, "v": 0.19827831929355083, "wq": [21, 11], "bq": 0, "n_samples": 26542080}, {"channel": 76, "mu": 0.0006590666594328703, "a": 0.9313956623329576, "b": 0.0625, "u": 1.0120794790609589, "v": 0.21619198925093583, "wq": [9, 7], "bq": 0, "n_samples": 26542080}, {"channel": 77, "mu": -0.002216894832658179, "a": 0.9342802946169676, "b": 0.06305343944372177, "u": 1.0263065909066862, "v": 0.22166194930652033, "wq": [27, 7], "bq": 0, "n_samples": 26542080}, {"channel": 78, "mu": 0.002515138225790895, "a": 0.9244629682647789, "b": 0.0644499343980233, "u": 1.1238510070159928, "v": 0.2104780399683044, "wq": [-6, 12], "bq": 1, "n_samples": 26542080}, {"channel": 79, "mu": 0.0006598955319251543, "a": 0.9265609050209617, "b": 0.0625, "u": 1.0656381532113557, "v": 0.21415801555051936, "wq": [4, 6], "bq": 0, "n_samples": 26542080}], "groups": [{"channels": [0, 16], "ps": 128, "stream": "detail", "slicing": "tile f=128", "rank": 1, "stream_mode": "all_detail", "dataset": "cache/index:train", "seed": 0, "ggdlpc_fit": {"N": 31, "rank": 1, "K_beta": 8, "order1": 2, "dense": true, "dataset": "cache/index:train", "seed": 0, "passes": ["predictor moments", "dense h(S, r) histogram"]}}, {"channels": [16, 32], "ps": 128, "stream": "detail", "slicing": "tile f=128", "rank": 1, "stream_mode": "all_detail", "dataset": "cache/index:train", "seed": 0, "ggdlpc_fit": {"N": 31, "rank": 1, "K_beta": 8, "order1": 2, "dense": true, "dataset": "cache/index:train", "seed": 0, "passes": ["predictor moments", "dense h(S, r) histogram"]}}, {"channels": [32, 56], "ps": 64, "stream": "detail", "slicing": "tile f=256", "rank": 1, "stream_mode": "all_detail", "dataset": "cache/index:train", "seed": 0, "ggdlpc_fit": {"N": 31, "rank": 1, "K_beta": 8, "order1": 2, "dense": true, "dataset": "cache/index:train", "seed": 0, "passes": ["predictor moments", "dense h(S, r) histogram"]}}, {"channels": [56, 80], "ps": 32, "stream": "detail", "slicing": "tile f=512", "rank": 1, "stream_mode": "all_detail", "dataset": "cache/index:train", "seed": 0, "ggdlpc_fit": {"N": 31, "rank": 1, "K_beta": 8, "order1": 2, "dense": true, "dataset": "cache/index:train", "seed": 0, "passes": ["predictor moments", "dense h(S, r) histogram"]}}]}

encoder/weights.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65e860cf924da8fc4cae4f2ee48b7f82af1928a493ac7e0fb50e440a472c1296
+size 513704

rd.json

@@ -0,0 +1,34 @@
+{
+  "source_checkpoint": "e2a2s2_g4.pth",
+  "rate_unit": "bits/sample",
+  "val_ds": "cache/index",
+  "n_val": 1215,
+  "rd_curve": [
+    {
+      "n_ch": 16,
+      "bpp": 0.2006698629329561,
+      "psnr": 31.76025830270721
+    },
+    {
+      "n_ch": 32,
+      "bpp": 0.46118514589048926,
+      "psnr": 33.027815629202536
+    },
+    {
+      "n_ch": 56,
+      "bpp": 1.2559980227623457,
+      "psnr": 35.112185056055175
+    },
+    {
+      "n_ch": 80,
+      "bpp": 2.3985202644533032,
+      "psnr": 36.8108549296502
+    }
+  ],
+  "verified_ops": [
+    16,
+    32,
+    56,
+    80
+  ]
+}

web/spatial_audio_op16_chunk16384_m4096.json

@@ -0,0 +1,54 @@
+{
+ "codec": "spatial_audio",
+ "repo": "Ambiq/compressionkit-e2a2-spatial-audio-v0.3",
+ "subfolder": null,
+ "revision": "04efcae6aeeacb929cca19dad9fef65ab4ccada6",
+ "op": 16,
+ "file": "spatial_audio_op16_chunk16384_m4096.tflite",
+ "bytes": 410460028,
+ "precision": "fp32",
+ "parameters": 102582151,
+ "plan": {
+  "nd": 1,
+  "decoder_ps": 128,
+  "receptive_cells": 13,
+  "chunk_cells": 128,
+  "in_channels": 16,
+  "out_channels": 7,
+  "margin_cells": 32,
+  "group_ps": [
+   128
+  ]
+ },
+ "overlap": 0.25,
+ "chunk": 16384,
+ "margin": 4096,
+ "window": 24576,
+ "inputs": [
+  {
+   "name": "args_0",
+   "role": "decoder grid window",
+   "shape": [
+    1,
+    16,
+    192
+   ]
+  },
+  {
+   "name": "args_1",
+   "role": "mask",
+   "shape": [
+    1,
+    1,
+    192
+   ]
+  }
+ ],
+ "output_shape": [
+  1,
+  7,
+  24576
+ ],
+ "litert_torch": "0.10.0",
+ "conversion_seconds": 4.1
+}

web/spatial_audio_op16_chunk16384_m4096.tflite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9eff2923517e1db0862dc4eb84a5416d999530f8e6d535a88b4d8fefadd530fd
+size 410460028

web/spatial_audio_op32_chunk16384_m4096.json

@@ -0,0 +1,54 @@
+{
+ "codec": "spatial_audio",
+ "repo": "Ambiq/compressionkit-e2a2-spatial-audio-v0.3",
+ "subfolder": null,
+ "revision": "04efcae6aeeacb929cca19dad9fef65ab4ccada6",
+ "op": 32,
+ "file": "spatial_audio_op32_chunk16384_m4096.tflite",
+ "bytes": 410656636,
+ "precision": "fp32",
+ "parameters": 102631303,
+ "plan": {
+  "nd": 1,
+  "decoder_ps": 128,
+  "receptive_cells": 13,
+  "chunk_cells": 128,
+  "in_channels": 32,
+  "out_channels": 7,
+  "margin_cells": 32,
+  "group_ps": [
+   128
+  ]
+ },
+ "overlap": 0.25,
+ "chunk": 16384,
+ "margin": 4096,
+ "window": 24576,
+ "inputs": [
+  {
+   "name": "args_0",
+   "role": "decoder grid window",
+   "shape": [
+    1,
+    32,
+    192
+   ]
+  },
+  {
+   "name": "args_1",
+   "role": "mask",
+   "shape": [
+    1,
+    1,
+    192
+   ]
+  }
+ ],
+ "output_shape": [
+  1,
+  7,
+  24576
+ ],
+ "litert_torch": "0.10.0",
+ "conversion_seconds": 3.9
+}

web/spatial_audio_op32_chunk16384_m4096.tflite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e70bc873f70da7cc1a2cedb918827038570dfed49162918788ba48cff0efade8
+size 410656636

web/spatial_audio_op56_chunk16384_m4096.json

@@ -0,0 +1,55 @@
+{
+ "codec": "spatial_audio",
+ "repo": "Ambiq/compressionkit-e2a2-spatial-audio-v0.3",
+ "subfolder": null,
+ "revision": "04efcae6aeeacb929cca19dad9fef65ab4ccada6",
+ "op": 56,
+ "file": "spatial_audio_op56_chunk16384_m4096.tflite",
+ "bytes": 814609900,
+ "precision": "fp32",
+ "parameters": 203589511,
+ "plan": {
+  "nd": 1,
+  "decoder_ps": 128,
+  "receptive_cells": 25,
+  "chunk_cells": 128,
+  "in_channels": 80,
+  "out_channels": 7,
+  "margin_cells": 32,
+  "group_ps": [
+   128,
+   64
+  ]
+ },
+ "overlap": 0.25,
+ "chunk": 16384,
+ "margin": 4096,
+ "window": 24576,
+ "inputs": [
+  {
+   "name": "args_0",
+   "role": "decoder grid window",
+   "shape": [
+    1,
+    80,
+    192
+   ]
+  },
+  {
+   "name": "args_1",
+   "role": "mask",
+   "shape": [
+    1,
+    1,
+    192
+   ]
+  }
+ ],
+ "output_shape": [
+  1,
+  7,
+  24576
+ ],
+ "litert_torch": "0.10.0",
+ "conversion_seconds": 7.3
+}

web/spatial_audio_op56_chunk16384_m4096.tflite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4ffba184af290c28aaf54ab5d169410d9c5e33b3462e99051c6afab1190af9f
+size 814609900

web/spatial_audio_op80_chunk16384_m4096.json

@@ -0,0 +1,56 @@
+{
+ "codec": "spatial_audio",
+ "repo": "Ambiq/compressionkit-e2a2-spatial-audio-v0.3",
+ "subfolder": null,
+ "revision": "7f1dbb0fa3ad41f7f2c7da9addadbc65f7d779e1",
+ "op": 80,
+ "file": "spatial_audio_op80_chunk16384_m4096.tflite",
+ "bytes": 815789548,
+ "precision": "fp32",
+ "parameters": 203884423,
+ "plan": {
+  "nd": 1,
+  "decoder_ps": 128,
+  "receptive_cells": 25,
+  "chunk_cells": 128,
+  "in_channels": 176,
+  "out_channels": 7,
+  "margin_cells": 32,
+  "group_ps": [
+   128,
+   64,
+   32
+  ]
+ },
+ "overlap": 0.25,
+ "chunk": 16384,
+ "margin": 4096,
+ "window": 24576,
+ "inputs": [
+  {
+   "name": "args_0",
+   "role": "decoder grid window",
+   "shape": [
+    1,
+    176,
+    192
+   ]
+  },
+  {
+   "name": "args_1",
+   "role": "mask",
+   "shape": [
+    1,
+    1,
+    192
+   ]
+  }
+ ],
+ "output_shape": [
+  1,
+  7,
+  24576
+ ],
+ "litert_torch": "0.10.0",
+ "conversion_seconds": 7.8
+}

web/spatial_audio_op80_chunk16384_m4096.tflite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e4b85c724565410f1ba73c45e54cfab14207659c2107743aff374beb92c95e8
+size 815789548