3b6fd99989
WAI is a container + capability-dispatch standard for media. It does not re-implement codecs; it dispatches to SOTA standard libraries (PNG, FLAC, zstd as the mandatory floor; AVIF, JPEG-XL, Opus, AV1, XZ as the recommended modern set) and adds an envelope that lets a neural shared-prior path coexist with the model-free floor. The capability a file requires is named, not supplied. Two paths open every conforming file: the neural condition (sink regenerates from a shared ambient prior) and the zeroth condition (registered SOTA codec menu). Apache-2.0. Adds: - wai/SPEC.md — WAI v1.0 spec (envelope, manifest, capability dispatch, registered capabilities including v1.1 multi-rendition extension) - wai/wai-rs/ — Rust reference implementation: lib + cdylib + staticlib, C FFI, 11 capability + envelope round-trip tests passing - wai/wai-web/ — wasm-bindgen wrapper for browser sinks - wai/wai-neural/ — Python tooling for neural-condition models - wai/tools/ — codec export + encode helpers (Python) - wai/corpus/CORPUS.md — third-party test corpus reacquisition guide Updates README.md to list four standards instead of three. Known consumer: CommunicationOS uses WAI for media transport over MoQT/MLS sessions. Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
67 lines
2.4 KiB
Python
67 lines
2.4 KiB
Python
"""Extract per-channel CDF tables + medians from bmshj2018-factorized.
|
|
|
|
The EntropyBottleneck's prior is *factorized* — every spatial position
|
|
in channel `c` shares the same CDF. After `net.update()`, three tensors
|
|
parameterize the rANS coder:
|
|
|
|
_quantized_cdf int32 [C, max_cdf_length] row-major per-channel CDF
|
|
_cdf_length int32 [C] active length per channel
|
|
_offset int32 [C] integer offset per channel
|
|
_get_medians() float [C, 1, 1] per-channel median for de-quantization
|
|
|
|
These four arrays are everything a JS / Rust decoder needs to reproduce
|
|
the int32 symbols, then dequantize back to float32 `y_hat` for g_s.
|
|
|
|
Output is a compact JSON sidecar shipped at
|
|
`wai-web/demo/models/bmshj2018/cdfs_q<N>.json`.
|
|
"""
|
|
from __future__ import annotations
|
|
|
|
import argparse
|
|
import json
|
|
from pathlib import Path
|
|
|
|
import torch
|
|
from compressai.zoo import models
|
|
|
|
|
|
def main() -> int:
|
|
ap = argparse.ArgumentParser()
|
|
ap.add_argument("--quality", type=int, default=3)
|
|
ap.add_argument("--out", required=True)
|
|
args = ap.parse_args()
|
|
out = Path(args.out); out.parent.mkdir(parents=True, exist_ok=True)
|
|
|
|
print(f"loading bmshj2018-factorized (quality={args.quality})…")
|
|
net = models["bmshj2018-factorized"](quality=args.quality, pretrained=True).eval()
|
|
net.update(force=True)
|
|
eb = net.entropy_bottleneck
|
|
|
|
cdf = eb._quantized_cdf.cpu().int().tolist() # list[list[int]] shape [C, max_len]
|
|
cdf_len = eb._cdf_length.cpu().int().tolist() # list[int] len C
|
|
offset = eb._offset.cpu().int().tolist() # list[int] len C
|
|
medians = eb._get_medians().detach().reshape(-1).cpu().float().tolist()
|
|
|
|
C = len(cdf_len)
|
|
blob = {
|
|
"model": "bmshj2018-factorized",
|
|
"quality": args.quality,
|
|
"precision": 16,
|
|
"bypass_precision": 4,
|
|
"channels": C,
|
|
"cdf_length": cdf_len,
|
|
"offset": offset,
|
|
"medians": medians,
|
|
"cdf": cdf, # large; we trim each row to cdf_length below
|
|
}
|
|
# Save bytes: trim each CDF row to its actual length.
|
|
blob["cdf"] = [row[:cdf_len[i]] for i, row in enumerate(cdf)]
|
|
|
|
out.write_text(json.dumps(blob, separators=(",", ":")))
|
|
print(f" channels={C} cdf_max_len={max(cdf_len)}")
|
|
print(f" wrote {out} ({out.stat().st_size:,} B)")
|
|
return 0
|
|
|
|
|
|
if __name__ == "__main__":
|
|
raise SystemExit(main())
|