7f59912916
The model code was spread across eight crates that had grown into each other:
ggml and cuda and mlx each owned part of a tensor runtime, llama and tts and
voice2 each owned part of a model, and libs/diffusion owned everything else.
They are now one tree with an explicit shape:
libs/ai/cuda — kernels and launch surface
libs/ai/metal — Metal shaders and the shim
libs/ai/llm — the language-model runtime (sessions, lanes, contexts,
the CUDA and Metal executors, the compiled Metal path)
libs/ai/models/ — common, flux, h3, music, paint, speech, stems, vision
libs/diffusion is not deleted but demoted: what remains is the VALIDATOR
crate — several dozen `*_validate.rs` oracles that check a native
implementation against a reference, which is where they belong now that the
implementations live next door.
The functional work inside the move is mostly in the LLM runtime: N lanes that
draft while one verify batch serves all of them, per-slot prefill over a shared
folded attention arena, speculation that survives batching, and a scheduler
that reports rather than publishes. And in the CUDA build: a machine without
usable CUDA must still LINK (and say so), the default kernel arch is the
building machine's GPU, `NO_CUDA` forces the stub even where the toolkit
exists, and kernels compile in parallel with progress.
libs/video_flow is new here: classical optical flow estimation and the `mkfl`
motion-field payload — a flow field measured from a clip without a model,
which is what drives free-rate bounce-looping playback and the uprez/tween
enhance pipe.
44 lines
1.4 KiB
Python
44 lines
1.4 KiB
Python
#!/usr/bin/env python3
|
|
"""Emit `src/g2p/vocab.rs` from Kokoro's `config.json`.
|
|
|
|
python3 gen_vocab.py config.json ../src/g2p/vocab.rs
|
|
"""
|
|
|
|
import json
|
|
import sys
|
|
|
|
|
|
def main():
|
|
src, dst = sys.argv[1], sys.argv[2]
|
|
vocab = json.load(open(src))["vocab"]
|
|
pairs = sorted(vocab.items(), key=lambda kv: kv[0])
|
|
|
|
lines = [
|
|
"// Generated by tools/gen_vocab.py from Kokoro's config.json. Do not edit.",
|
|
"",
|
|
"/// Phoneme symbol -> token id, sorted by symbol for binary search.",
|
|
f"pub const VOCAB: [(char, u16); {len(pairs)}] = [",
|
|
]
|
|
for symbol, token in pairs:
|
|
# Always escape: the vocab contains `'`, `"`, a combining tilde and IPA.
|
|
escaped = f"\\u{{{ord(symbol):04x}}}"
|
|
shown = symbol if symbol.strip() and symbol != "'" else repr(symbol)
|
|
lines.append(f" ('{escaped}', {token}), // {shown}")
|
|
lines.append("];")
|
|
lines.append("")
|
|
lines.append("/// Look up a phoneme symbol's token id.")
|
|
lines.append("pub fn token(symbol: char) -> Option<u16> {")
|
|
lines.append(" VOCAB")
|
|
lines.append(" .binary_search_by_key(&symbol, |(sym, _)| *sym)")
|
|
lines.append(" .ok()")
|
|
lines.append(" .map(|index| VOCAB[index].1)")
|
|
lines.append("}")
|
|
lines.append("")
|
|
|
|
with open(dst, "w") as out:
|
|
out.write("\n".join(lines))
|
|
print(f"{dst}: {len(pairs)} symbols")
|
|
|
|
|
|
if __name__ == "__main__":
|
|
main()
|