Blame for scripts/rht_int4_convert.py

#!/usr/bin/env python3
"""Convert a supported Hugging Face model to uzu with RHT + INT4 Group32.

This bypasses MLX quantization and applies lalamo's HybridSpec directly:
  BF16 Linear -> RHT (Hadamard block 32) -> asymmetric INT4 group 32 -> HybridMatrix

The resulting lalamo model is saved, its HybridSpec metadata is rewritten to
uzu's flat form, and encoding.json is added.

Example (0.8B, local CPU):
  JAX_PLATFORM_NAME=cpu python scripts/rht_int4_convert.py \
    --spec-name Qwen3.5-0.8B \
    --lalamo-path /tmp/lalamo-direct \
    --output-dir /tmp/qwen35_0.8b_rht_int4_lalamo \
    --bf16-dir /tmp/qwen35_0.8b_bf16_lalamo
"""

from __future__ import annotations

import argparse
import json
import os
import sys
import time
from dataclasses import replace
from pathlib import Path


def parse_args() -> argparse.Namespace:
    parser = argparse.ArgumentParser()
    parser.add_argument("--spec-name", default="Qwen3.5-0.8B", help="Built-in lalamo model spec name")
    parser.add_argument("--hf-repo", default=None, help="Override origin with this Hugging Face repo")
    parser.add_argument("--hf-dir", default=None, help="Use a local HF snapshot directory instead of downloading")
    parser.add_argument("--vendor", default=None)
    parser.add_argument("--name", default=None)
    parser.add_argument("--output-dir", required=True)
    parser.add_argument("--bf16-dir", default=None, help="Also save the unquantized model here")
    parser.add_argument(
        "--load-from",
        default=None,
        help="Load an existing lalamo model directory instead of importing from Hugging Face",
    )
    parser.add_argument("--lalamo-path", default=os.environ.get("LALAMO_PATH", "/tmp/lalamo-direct"))
    parser.add_argument(
        "--parallel",
        choices=("replicated", "tensor", "fsdp", "data"),
        default="replicated",
        help="JAX device sharding: tensor/fsdp use all visible GPUs as one memory pool",
    )
    parser.add_argument("--bits", type=int, default=4, choices=(4, 6, 8))
    parser.add_argument(
        "--layer-bits-json",
        type=Path,
        default=None,
        help='Per-layer bit overrides: {"default": 4, "rules": [{"match": "mlp/down_projection", "bits": 6}]}',
    )
    parser.add_argument(
        "--symmetric-bits",
        default="8",
        help=(
            "Comma-separated bit widths exported as symmetric int (no zero points). "
            "uzu only compiles the zero-point gemm for 4 bits, so int8 must be symmetric. "
            "Pass an empty string to force asymmetric everywhere."
        ),
    )
    parser.add_argument(
        "--embedding-bits",
        type=int,
        default=None,
        choices=(4, 8),
        help="Override --bits for the tied/untied embedding and readout only",
    )
    parser.add_argument(
        "--layer-filter",
        default="",
        help="Only quantize layers whose path contains one of these comma-separated substrings; others stay full precision",
    )
    parser.add_argument("--group-size", type=int, default=32)
    parser.add_argument("--block-size", type=int, default=32, choices=(32, 64, 128))
    parser.add_argument(
        "--no-rht",
        action="store_true",
        help="Disable RHT and quantize with plain INT4 Group (isolation baseline)",
    )
    parser.add_argument(
        "--quantize-embedding",
        action="store_true",
        help="Also quantize the tied token embedding with RHT output-only (matches Mirai-M)",
    )
    parser.add_argument("--dtype", default="bfloat16")
    parser.add_argument("--skip-incompatible", action="store_true", help="Leave non-divisible Linears full precision")
    parser.add_argument("--dry-run", action="store_true")
    parser.add_argument("--hf-token", default=None)
    parser.add_argument(
        "--preconditioners",
        default=None,
        help="NPZ from scripts/yaqa_calibration.py; keys are '<path>/input_block' and optional '/output_block'",
    )
    parser.add_argument(
        "--calibration-json",
        type=Path,
        default=None,
        help="calibration.json paired with --preconditioners; embedded into provenance.json",
    )
    parser.add_argument(
        "--progress-jsonl",
        type=Path,
        default=None,
        help="default: <output-dir>/convert_progress.jsonl",
    )
    return parser.parse_args()


# uzu config strictness:
#   * every struct uses `#[serde(deny_unknown_fields)]`
#   * every field is required, even `Option<T>` (missing != null)
# lalamo emits a newer token_codec schema that violates both rules.
_UZU_TOKEN_CODEC_REQUIRED = ("output_parser_regex", "end_of_thinking_tag")
_UZU_TOKEN_CODEC_UNSUPPORTED = ("response_format", "reasoning_effort")


def normalize_config_for_uzu(output_dir: Path) -> None:
    path = output_dir / "config.json"
    config = json.loads(path.read_text())

    # Keep the untouched lalamo form for lalamo-side evaluation (PPL / logits).
    lalamo_path = output_dir / "config.lalamo.json"
    if not lalamo_path.exists():
        lalamo_path.write_text(json.dumps(config, indent=2) + "\n")

    token_codec = config.get("token_codec_config")
    if token_codec is None:
        return

    removed = [key for key in _UZU_TOKEN_CODEC_UNSUPPORTED if token_codec.pop(key, None) is not None]
    added = [key for key in _UZU_TOKEN_CODEC_REQUIRED if key not in token_codec]
    for key in added:
        token_codec[key] = None

    if removed or added:
        path.write_text(json.dumps(config, indent=2) + "\n")
    print(f"normalize config.json: removed={removed or '[]'} added={added or '[]'}")


def find_spec(qwen_models, name: str):
    for spec in qwen_models:
        if spec.name == name:
            return spec
    raise ValueError(f"model spec {name!r} not found")


def should_quantize(path: str, filters: list[str]) -> bool:
    """Empty filters means everything; otherwise any substring match."""
    if not filters:
        return True
    return any(item in path for item in filters)


def _normalize_path(path: str) -> str:
    """Drop the leading dots lalamo uses per component.

    Site paths look like ``.decoder/.transformer/.layers/[0]/.mlp/.down_projection``
    while the lane spec writes ``mlp/down_projection``, so a naive substring
    test never matched and every layer stayed at the default bits (measured:
    uniform4 and down-out6 produced identical PPL).
    """
    return path.replace(".", "")


def resolve_bits(path: str, rules: list[dict], default: int) -> int:
    """First rule whose ``match`` substring appears in the layer path wins."""
    normalized = _normalize_path(path)
    for rule in rules:
        match = _normalize_path(str(rule.get("match", "")))
        if match and match in normalized:
            return int(rule["bits"])
    return default


def load_layer_bits(spec_path: Path | None, default: int) -> tuple[dict, list[dict]]:
    if spec_path is None:
        return {"default": default, "rules": []}, []
    payload = json.loads(spec_path.read_text(encoding="utf-8"))
    rules = payload.get("rules", [])
    for rule in rules:
        bits = int(rule["bits"])
        if bits not in (4, 6, 8):
            raise SystemExit(f"unsupported bits in layer-bits-json: {bits}")
        if not rule.get("match"):
            raise SystemExit("layer-bits-json rule needs a non-empty 'match'")
    summary = {"default": int(payload.get("default", default)), "rules": rules}
    return summary, rules


def main() -> None:
    args = parse_args()
    os.environ.setdefault("JAX_PLATFORM_NAME", "cpu")
    if args.hf_token:
        os.environ["HF_TOKEN"] = args.hf_token

    progress_path = args.progress_jsonl or (Path(args.output_dir) / "convert_progress.jsonl")
    progress_path.parent.mkdir(parents=True, exist_ok=True)
    run_id = time.strftime("%Y%m%d-%H%M%S")
    progress_path.write_text("", encoding="utf-8")

    def progress(record: dict) -> None:
        record = {"run_id": run_id, **record}
        with open(progress_path, "a", encoding="utf-8") as file:
            file.write(json.dumps(record) + "\n")

    sys.path.insert(0, str(Path(__file__).resolve().parent))
    sys.path.insert(0, args.lalamo_path)

    import jax
    import jax.numpy as jnp
    import equinox as eqx
    import numpy as np

    from lalamo.compressed.hybrid import HybridSpec
    from lalamo.compressed.int import IntSpec
    from lalamo.compressed.incoherence import HadamardSpec, IncoherenceProcessingMode
    from lalamo.model_import.common import import_model
    from lalamo.model_import.model_specs.qwen import QWEN_MODELS
    from lalamo.modules.linear import Linear
    from lalamo.preconditioner import Preconditioner
    from lalamo.utils.sharding import ShardingConfig
    from lalamo.utils.surgery import map_nodes_of_type_with_path, select_nodes_of_type
    from lalamo.weight_matrix import Layout
    from rewrite_lalamo_hybrid_spec import rewrite

    if args.parallel == "tensor":
        sharding = ShardingConfig.tensor_parallel()
    elif args.parallel == "fsdp":
        sharding = ShardingConfig.fully_sharded_data_parallel()
    elif args.parallel == "data":
        sharding = ShardingConfig.data_parallel()
    else:
        sharding = ShardingConfig.replicated()

    if args.load_from:
        from lalamo.model import Model

        print(f"load lalamo model <- {args.load_from}")
        model = Model.load(args.load_from, sharding, dtype=jnp.dtype(args.dtype))
        print("load done")
    else:
        spec = find_spec(QWEN_MODELS, args.spec_name)
        if args.hf_dir:
            from pathlib import Path as _Path

            from lalamo.model_import.origins import LocalOrigin

            weight_files = tuple(sorted(path.name for path in _Path(args.hf_dir).glob("*.safetensors")))
            if not weight_files:
                raise SystemExit(f"no safetensors files in {args.hf_dir}")
            spec = replace(
                spec,
                vendor=args.vendor or spec.vendor,
                name=args.name or spec.name,
                origin=LocalOrigin(root=args.hf_dir, weight_files=weight_files),
            )
        elif args.hf_repo:
            from lalamo.model_import.origins import HuggingFaceOrigin

            spec = replace(
                spec,
                vendor=args.vendor or spec.vendor,
                name=args.name or spec.name,
                origin=HuggingFaceOrigin(repo=args.hf_repo),
            )

        print(f"import {spec.vendor}/{spec.name} <- {spec.origin.description}")
        results = import_model(
            spec,
            sharding_config=sharding,
            dtype=jnp.dtype(args.dtype),
        )
        model = results.model
        print("import done")

    if args.bf16_dir:
        bf16_dir = Path(args.bf16_dir)
        bf16_dir.mkdir(parents=True, exist_ok=True)
        print(f"save BF16 baseline -> {bf16_dir}")
        model.save(bf16_dir)
        normalize_config_for_uzu(bf16_dir)
        (bf16_dir / "COMPLETE").write_text("ok\n", encoding="utf-8")
        print("BF16 baseline saved")

    linear_list = select_nodes_of_type(Linear, model)
    required_divisor = args.group_size if args.no_rht else max(args.block_size, args.group_size)
    bad = []
    for path, linear in linear_list:
        shape = tuple(int(x) for x in linear.weights.shape)
        if len(shape) != 2 or any(dimension % required_divisor for dimension in shape):
            bad.append(("/".join(path), shape))

    print(f"Linear layers: {len(linear_list)}, incompatible: {len(bad)}")
    for path, shape in bad:
        print(f"  {path}: {shape}")
    if bad and not args.skip_incompatible:
        raise SystemExit("incompatible Linear shapes; use --skip-incompatible to leave them full precision")

    symmetric_bits = {
        int(item) for item in str(args.symmetric_bits).split(",") if item.strip()
    }
    if not symmetric_bits.issubset({4, 8}):
        raise SystemExit("--symmetric-bits accepts only 4 and 8")

    layer_bits_summary, layer_bits_rules = load_layer_bits(args.layer_bits_json, args.bits)
    if layer_bits_rules:
        print(f"layer bits overrides: {json.dumps(layer_bits_summary, ensure_ascii=False)}")

    def build_hybrid_spec(bits: int) -> HybridSpec:
        return HybridSpec(
            quantization_spec=IntSpec(
                bits=bits,
                group_size=args.group_size,
                is_symmetric=bits in symmetric_bits,
                layout=Layout.OUTPUT_INPUT,
            ),
            adapter_spec=None,
            incoherence_spec=None
            if args.no_rht
            else HadamardSpec(
                block_size=args.block_size,
                processing_mode=IncoherenceProcessingMode.INPUT_OUTPUT,
            ),
        )

    hybrid_spec = build_hybrid_spec(args.bits)
    layer_bits_used: dict[str, int] = {}
    quantize_key = jax.random.key(0)
    counts = {"quantized": 0, "skipped": 0, "yaqa": 0, "filtered": 0}
    layer_filters = [item for item in (args.layer_filter or "").split(",") if item]
    if layer_filters:
        print(f"layer filter: {layer_filters} (other layers stay full precision)")

    calibration = None
    if args.preconditioners:
        calibration = np.load(args.preconditioners)
        print(f"preconditioners: {args.preconditioners} ({len(calibration.files)} arrays)")

    def preconditioner_for(path) -> Preconditioner | None:
        if calibration is None:
            return None
        key = "/".join(path)
        input_key = f"{key}/input_block"
        if input_key not in calibration:
            return None
        output_key = f"{key}/output_block"
        return Preconditioner.init(
            input_block=jnp.asarray(calibration[input_key], dtype=jnp.float32),
            output_block=jnp.asarray(calibration[output_key], dtype=jnp.float32)
            if output_key in calibration
            else None,
        )

    def quantize_linear(path, linear: Linear) -> Linear:
        started = time.time()
        key = "/".join(path)
        if not should_quantize(key, layer_filters):
            counts["filtered"] += 1
            return linear
        shape = tuple(int(x) for x in linear.weights.shape)
        if len(shape) != 2 or any(dimension % required_divisor for dimension in shape):
            counts["skipped"] += 1
            progress({"path": key, "shape": shape, "status": "skipped", "elapsed_s": time.time() - started})
            if not args.skip_incompatible:
                raise ValueError(f"incompatible Linear {path}: {shape}")
            return linear
        counts["quantized"] += 1
        preconditioner = preconditioner_for(path)
        if preconditioner is not None:
            counts["yaqa"] += 1
        progress({"path": key, "shape": shape, "status": "started"})
        bits = resolve_bits(key, layer_bits_rules, layer_bits_summary["default"])
        layer_bits_used[key] = bits
        spec = build_hybrid_spec(bits) if bits != args.bits else hybrid_spec
        try:
            new_weights = spec.compress(
                linear.weights.decompress(),
                key=quantize_key,
                preconditioner=preconditioner,
                sharding_config=linear.weights.sharding_config,
                is_sharded=linear.weights.is_sharded,
            )
        except Exception as error:
            progress(
                {
                    "path": key,
                    "shape": shape,
                    "status": "failed",
                    "error": repr(error),
                    "elapsed_s": time.time() - started,
                }
            )
            raise
        progress(
            {
                "path": key,
                "shape": shape,
                "status": "quantized",
                "yaqa": preconditioner is not None,
                "elapsed_s": time.time() - started,
            }
        )
        return eqx.tree_at(lambda x: x.weights, linear, new_weights)

    if args.dry_run:
        print("dry-run: no quantization performed")
        return

    if args.no_rht:
        print(
            f"quantize -> plain INT{args.bits} group {args.group_size} "
            f"(symmetric={args.bits in symmetric_bits}, layout=output_input)"
        )
        if args.embedding_bits is not None:
            print(
                f"embedding quantize -> plain INT{args.embedding_bits} group {args.group_size} "
                f"(symmetric={args.embedding_bits in symmetric_bits})"
            )
    else:
        print(
            f"quantize -> RHT block {args.block_size} + INT{args.bits} group {args.group_size} "
            f"(symmetric={args.bits in symmetric_bits}, layout=output_input)"
        )
        if args.embedding_bits is not None:
            print(
                f"embedding quantize -> RHT block {args.block_size} + INT{args.embedding_bits} group {args.group_size} "
                f"(symmetric={args.embedding_bits in symmetric_bits})"
            )
    quantized_model = map_nodes_of_type_with_path(Linear, quantize_linear, model)
    print(
        f"quantized: {counts['quantized']}, skipped: {counts['skipped']}, "
        f"yaqa: {counts['yaqa']}, filtered: {counts['filtered']}"
    )
    if layer_bits_used:
        distribution = {}
        for bits in layer_bits_used.values():
            distribution[bits] = distribution.get(bits, 0) + 1
        print(f"bits distribution: {dict(sorted(distribution.items()))}")

    def embedding_spec(layout: Layout, mode: IncoherenceProcessingMode | None) -> HybridSpec:
        embedding_bits = args.embedding_bits if args.embedding_bits is not None else args.bits
        return HybridSpec(
            quantization_spec=IntSpec(
                bits=embedding_bits,
                group_size=args.group_size,
                is_symmetric=embedding_bits in symmetric_bits,
                layout=layout,
            ),
            adapter_spec=None,
            incoherence_spec=None
            if args.no_rht or mode is None
            else HadamardSpec(block_size=args.block_size, processing_mode=mode),
        )

    def quantize_matrix(matrix, layout: Layout, mode: IncoherenceProcessingMode | None, label: str):
        started = time.time()
        embedding_dim = int(matrix.shape[-1])
        if not args.no_rht and embedding_dim % args.block_size:
            raise SystemExit(f"{label} dim {embedding_dim} is not divisible by block size {args.block_size}")
        new_matrix = embedding_spec(layout, mode).compress(
            matrix.decompress(),
            key=quantize_key,
            sharding_config=matrix.sharding_config,
            is_sharded=matrix.is_sharded,
        )
        shape = tuple(int(x) for x in matrix.shape)
        print(f"{label} quantized: shape={shape}")
        progress({"path": label, "shape": shape, "status": "quantized", "yaqa": False, "elapsed_s": time.time() - started})
        return new_matrix

    if args.quantize_embedding:
        embedding = getattr(quantized_model.decoder, "embedding", None)
        tied = getattr(embedding, "embedding", None)
        if tied is not None:
            matrix = quantize_matrix(
                tied,
                Layout.INPUT_OUTPUT,
                IncoherenceProcessingMode.OUTPUT,
                "tied embedding",
            )
            new_embedding = eqx.tree_at(lambda node: node.embedding, embedding, matrix)
            quantized_model = eqx.tree_at(
                lambda root: root.decoder.embedding,
                quantized_model,
                new_embedding,
            )
        elif hasattr(embedding, "input_embedding") and hasattr(embedding, "output_embedding"):
            input_embedding = quantize_matrix(
                embedding.input_embedding,
                Layout.INPUT_OUTPUT,
                IncoherenceProcessingMode.OUTPUT,
                "input_embedding",
            )
            output_embedding = quantize_matrix(
                embedding.output_embedding,
                Layout.OUTPUT_INPUT,
                IncoherenceProcessingMode.INPUT,
                "output_embedding (lm_head)",
            )
            embedding = eqx.tree_at(lambda node: node.input_embedding, embedding, input_embedding)
            embedding = eqx.tree_at(lambda node: node.output_embedding, embedding, output_embedding)
            quantized_model = eqx.tree_at(
                lambda root: root.decoder.embedding,
                quantized_model,
                embedding,
            )
        else:
            print("embedding: neither tied nor untied embedding found, skipped")

    output_dir = Path(args.output_dir)
    output_dir.mkdir(parents=True, exist_ok=True)
    print(f"save quantized model -> {output_dir}")
    quantized_model.save(output_dir)

    print("rewrite HybridSpec metadata for uzu")
    rewrite(output_dir / "model.safetensors")

    normalize_config_for_uzu(output_dir)

    encoding = {"type": "hanashi", "name": "qwen3.5"}
    (output_dir / "encoding.json").write_text(json.dumps(encoding, indent=2) + "\n")

    calibration_summary = None
    if args.calibration_json and args.calibration_json.exists():
        calibration_summary = json.loads(args.calibration_json.read_text())
    source = args.hf_repo or (str(args.hf_dir) if args.hf_dir else None) or args.load_from or args.spec_name
    provenance = {
        "kind": "yaqa" if args.preconditioners else "rtn",
        "source": source,
        "preconditioners": args.preconditioners,
        "calibration_json": str(args.calibration_json) if args.calibration_json else None,
        "calibration": calibration_summary,
        "layer_bits": layer_bits_summary if layer_bits_rules else None,
        "layer_bits_used": layer_bits_used if layer_bits_rules else None,
        "overlap_checked": (
            bool(calibration_summary.get("overlap_checked"))
            if calibration_summary is not None
            else args.preconditioners is None
        ),
    }
    (output_dir / "provenance.json").write_text(
        json.dumps(provenance, indent=2, ensure_ascii=False) + "\n",
        encoding="utf-8",
    )
    (output_dir / "COMPLETE").write_text("ok\n", encoding="utf-8")
    print("encoding.json / provenance.json written")

    for f in sorted(output_dir.iterdir()):
        if f.is_file():
            print(f"  {f.name}: {f.stat().st_size / 1e6:.1f} MB")


if __name__ == "__main__":
    main()