From 3ede71fc912e36c43196d095bfe66a2efe234bda Mon Sep 17 00:00:00 2001
From: Hamilakis Nicolas <nicolas.hamilakis@ens.psl.eu>
Date: Thu, 23 Jul 2026 15:01:11 +0200
Subject: [PATCH 1/3] Migrated towards a python module instead of standalone
 scripts

---
 .gitmodules                                   |   3 -
 VTC-2                                         |   1 -
 check_sys_dependencies.sh                     |   7 -
 pyproject.toml                                |  19 +-
 scripts/run.sh => run.sh                      |   2 +-
 scripts/infer.py                              | 328 ------------------
 src/vtc_inference/__init__.py                 |   4 +
 src/vtc_inference/__main__.py                 |   4 +
 src/vtc_inference/cli.py                      | 103 ++++++
 .../vtc_inference/convert_audios.py           |   6 +-
 src/vtc_inference/infer.py                    | 160 +++++++++
 src/vtc_inference/ressources.py               | 161 +++++++++
 src/vtc_inference/utils.py                    | 167 +++++++++
 thresholds/f1.toml                            |  15 -
 thresholds/hp.toml                            |  15 -
 uv.lock                                       | 224 +++++++++++-
 16 files changed, 834 insertions(+), 385 deletions(-)
 delete mode 160000 VTC-2
 rename scripts/run.sh => run.sh (91%)
 mode change 100644 => 100755
 delete mode 100644 scripts/infer.py
 create mode 100644 src/vtc_inference/__init__.py
 create mode 100644 src/vtc_inference/__main__.py
 create mode 100644 src/vtc_inference/cli.py
 rename scripts/convert.py => src/vtc_inference/convert_audios.py (97%)
 create mode 100644 src/vtc_inference/infer.py
 create mode 100644 src/vtc_inference/ressources.py
 create mode 100644 src/vtc_inference/utils.py
 delete mode 100644 thresholds/f1.toml
 delete mode 100644 thresholds/hp.toml

diff --git a/.gitmodules b/.gitmodules
index 731f895..e69de29 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -1,3 +0,0 @@
-[submodule "VTC-2"]
-	path = VTC-2
-	url = https://huggingface.co/coml/VTC-2
diff --git a/VTC-2 b/VTC-2
deleted file mode 160000
index 6b1a955..0000000
--- a/VTC-2
+++ /dev/null
@@ -1 +0,0 @@
-Subproject commit 6b1a95508302edc14c50f670cd9a30d66fa4f88a
diff --git a/check_sys_dependencies.sh b/check_sys_dependencies.sh
index abcfa47..71d73c7 100755
--- a/check_sys_dependencies.sh
+++ b/check_sys_dependencies.sh
@@ -13,13 +13,6 @@ else
     missing_deps+=("uv")
 fi
 
-# Check for git-lfs
-if command -v git-lfs &> /dev/null; then
-    echo "✓ git-lfs is installed ($(git-lfs --version))"
-else
-    echo "✗ git-lfs is NOT installed"
-    missing_deps+=("git-lfs")
-fi
 
 # Check for ffmpeg
 if command -v ffmpeg &> /dev/null; then
diff --git a/pyproject.toml b/pyproject.toml
index 09115da..a6bad66 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -5,18 +5,33 @@ description = "The inference code for the Voice Type Classifier model"
 readme = "README.md"
 requires-python = ">=3.13"
 dependencies = [
+    "huggingface-hub>=0.35.0",
     "pyannote-audio>=3.4.0",
     "segma",
     "tqdm>=4.67.1",
 ]
 
+[project.scripts]
+vtc-infer = "vtc_inference.cli:vtc_infer_cli"
+convert-audios = "vtc_inference.convert_audios:convert_audios_cli"
+
+[project.urls]
+documentation = "https://docs.cognitive-ml.fr/fastabx"
+
 [tool.uv.sources]
 segma = { git = "https://github.com/MarvinLvn/segma", rev = "e6cebc5b232ac1a30554caa7ec35dd22057d07f0" }
 
 [dependency-groups]
+dev = [
+    "ipython>=9.15.0",
+]
 doc = [
     "zensical>=0.0.37",
 ]
 
-[project.urls]
-documentation = "https://docs.cognitive-ml.fr/fastabx"
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/vtc_inference"]
diff --git a/scripts/run.sh b/run.sh
old mode 100644
new mode 100755
similarity index 91%
rename from scripts/run.sh
rename to run.sh
index 20737be..16ef56c
--- a/scripts/run.sh
+++ b/run.sh
@@ -6,6 +6,6 @@ output=out/$(date +%Y%m%d_%H%M)_inference_output
 
 mkdir -p $output
 
-uv run scripts/infer.py \
+uv run vtc-infer \
     --wavs $audios_path \
     --output $output
diff --git a/scripts/infer.py b/scripts/infer.py
deleted file mode 100644
index f343df8..0000000
--- a/scripts/infer.py
+++ /dev/null
@@ -1,328 +0,0 @@
-import argparse
-import copy
-import logging
-import shutil
-from pathlib import Path
-from typing import Literal
-
-import polars as pl
-from pyannote.core import Annotation, Segment
-from segma.inference import get_list_of_files_to_process, run_inference_on_audios
-from segma.utils.io import get_audio_info
-
-logging.basicConfig(
-    level=logging.INFO,
-    format="[%(asctime)s] [%(levelname)s] - %(message)s",
-    datefmt="%Y.%m.%d %H:%M:%S",
-)
-logger = logging.getLogger("inference")
-
-
-def load_aa(path: Path):
-    data = pl.read_csv(
-        source=path,
-        has_header=False,
-        new_columns=("uid", "start_time_s", "duration_s", "label"),
-        schema={
-            "uid": pl.String(),
-            "start_time_s": pl.Float64(),
-            "duration_s": pl.Float64(),
-            "label": pl.String(),
-        },
-        separator=" ",
-    )
-    return data
-
-
-def load_rttm(path: Path | str) -> pl.DataFrame:
-    try:
-        data = pl.read_csv(
-            source=path,
-            has_header=False,
-            columns=[1, 3, 4, 7],
-            new_columns=("uid", "start_time_s", "duration_s", "label"),
-            schema_overrides={
-                "uid": pl.String(),
-                "start_time_s": pl.Float64(),
-                "duration_s": pl.Float64(),
-                "label": pl.String(),
-            },
-            separator=" ",
-        )
-    except pl.exceptions.NoDataError:
-        data = pl.DataFrame(
-            None,
-            {
-                "uid": pl.String(),
-                "start_time_s": pl.Float64(),
-                "duration_s": pl.Float64(),
-                "label": pl.String(),
-            },
-        )
-
-    return data
-
-
-def load_one_uri(uri_df: pl.DataFrame):
-    for uids, turns in uri_df.group_by("uid"):
-        uid = uids[0]
-        annotation = Annotation(uri=uid)
-        for i, turn in enumerate(turns.iter_rows(named=True)):
-            segment = Segment(
-                turn["start_time_s"], turn["start_time_s"] + turn["duration_s"]
-            )
-            annotation[segment, i] = turn["label"]
-        yield uid, annotation
-
-
-def process_annot(
-    annotation: Annotation,
-    min_duration_off_s: float = 0.1,
-    min_duration_on_s: float = 0.1,
-) -> Annotation:
-    """Create a new `Annotation` with the `min_duration_off` and `min_duration_off` rules applied.
-
-    Args:
-        annotation (Annotation): input annotation
-        min_duration_off_s (float, optional): Remove speech segments shorter than that many seconds. Defaults to 0.1.
-        min_duration_on_s (float, optional): Fill same-speaker gaps shorter than that many seconds. Defaults to 0.1.
-
-    Returns:
-        Annotation: Processed annotation.
-    """
-    active = copy.deepcopy(annotation)
-    # NOTE - Fill regions shorter than that many seconds.
-    if min_duration_off_s > 0.0:
-        active = active.support(collar=min_duration_off_s)
-    # NOTE - remove regions shorter than that many seconds.
-    if min_duration_on_s > 0:
-        for segment, track in list(active.itertracks()):
-            if segment.duration < min_duration_on_s:
-                del active[segment, track]
-    return active
-
-
-def merge_segments(
-    file_uris_to_merge: list[str],
-    output: str,
-    min_duration_on_s: float = 0.1,
-    min_duration_off_s: float = 0.1,
-    write_empty: bool = True,
-):
-    output = Path(output)
-    raw_output_p = output / "raw_rttm"
-
-    # NOTE - merge RTTMs
-    uri_to_annot: dict[str, Annotation] = {}
-    uri_to_proc_annot: dict[str, Annotation] = {}
-    merged_out_p = output / "rttm"
-    merged_out_p.mkdir(exist_ok=True, parents=True)
-
-    for file_uri in file_uris_to_merge:
-        file = raw_output_p / f"{file_uri}.rttm"
-        if not file.exists():
-            continue
-
-        match file.suffix:
-            case ".aa":
-                data = load_aa(file)
-            case ".rttm":
-                data = load_rttm(file)
-            case _:
-                raise ValueError(
-                    f"File not found error, extension is not supported: {file}"
-                )
-
-        # NOTE - process, should handle the case where a single rttm contains multiple URIS
-        for uri, annot in load_one_uri(data):
-            uri_to_annot[uri] = annot
-            uri_to_proc_annot[uri] = process_annot(
-                annotation=annot,
-                min_duration_off_s=min_duration_off_s,
-                min_duration_on_s=min_duration_on_s,
-            )
-
-        for uri, annot in uri_to_proc_annot.items():
-            (merged_out_p / f"{uri}.rttm").write_text(annot.to_rttm())
-
-    # NOTE - Writting missing rttm files
-    if write_empty:
-        for uri in set(file_uris_to_merge) - set(uri_to_proc_annot.keys()):
-            (merged_out_p / f"{uri}.rttm").touch()
-
-
-def check_audio_files(audio_files_to_process: list[Path]) -> None:
-    """Fails if the audios are not sampled at 16_000 Hz and contain more than one channel."""
-
-    for wav_p in audio_files_to_process:
-        info = get_audio_info(wav_p)
-        # NOTE - check that the audio is valid
-        if not info.sample_rate == 16_000:
-            raise ValueError(
-                f"file `{wav_p}` is not samlped at 16 000 hz. Please convert your audio files."
-            )
-        if not info.n_channels == 1:
-            raise ValueError(
-                f"file `{wav_p}` has more than one channel. You can average your channels or use another channel reduction technique."
-            )
-
-
-def main(
-    output: str,
-    uris: Path | None = None,
-    config: str = "VTC-2/model/config.toml",
-    wavs: str = "data/debug/wav",
-    checkpoint: str = "VTC-2/model/best.ckpt",
-    save_probs: bool = False,
-    high_precision: bool = False,
-    thresholds: None | Path = Path("thresholds/f1.toml"),
-    min_duration_on_s: float = 0.1,
-    min_duration_off_s: float = 0.1,
-    batch_size: int = 128,
-    stride_pct: float = 0.25,
-    write_empty: bool = True,
-    write_csv: bool = True,
-    recursive_search: bool = False,
-    device: Literal["gpu", "cuda", "cpu", "mps"] = "gpu",
-    keep_raw: bool = False,
-    *args,
-    **kwargs,
-):
-    if high_precision:
-        thresholds = Path("thresholds/hp.toml")
-
-    output = Path(output)
-    output.mkdir(parents=True, exist_ok=True)
-
-    if thresholds:
-        shutil.copy(str(thresholds), dst=output)
-        logger.info(f"Using thresholds: {thresholds}")
-
-    logger.info("Running inference on audio files.")
-    processed_files = run_inference_on_audios(
-        config=config,
-        uris=uris,
-        wavs=wavs,
-        checkpoint=checkpoint,
-        output=output,
-        thresholds=thresholds,
-        batch_size=batch_size,
-        device=device,
-        recursive=recursive_search,
-        save_probs=save_probs,
-        stride_pct=stride_pct,
-        logger=logger,
-    )
-
-    logger.info("Merging detected speech segments.")
-    merge_segments(
-        file_uris_to_merge=[f.stem for f in processed_files],
-        output=output,
-        min_duration_on_s=min_duration_on_s,
-        min_duration_off_s=min_duration_off_s,
-        write_empty=write_empty,
-    )
-
-    if not keep_raw:
-        # NOTE - remove <output>/raw_rttm
-        shutil.rmtree(str(Path(output / "raw_rttm").absolute()))
-
-    # NOTE - write RTTMs to `csv` files
-    if write_csv:
-        if keep_raw:
-            # NOTE - Raw RTTMs
-            raw_rttm_file_p = sorted(list((output / "raw_rttm").glob("*.rttm")))
-            raw_rttm_file_dfs = []
-            for rttm_file in raw_rttm_file_p:
-                raw_rttm_file_dfs.append(load_rttm(rttm_file))
-            pl.concat(raw_rttm_file_dfs).write_csv(output / "raw_rttm.csv")
-
-        # NOTE - merged RTTMs
-        rttm_file_p = sorted(list((output / "rttm").glob("*.rttm")))
-        rttm_file_dfs = []
-        for rttm_file in rttm_file_p:
-            rttm_file_dfs.append(load_rttm(rttm_file))
-        pl.concat(rttm_file_dfs).write_csv(output / "rttm.csv")
-
-    logger.info(f"Inference finished, files can be found here: '{output.absolute()}/'")
-
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser()
-    parser.add_argument(
-        "--config",
-        type=str,
-        default="VTC-2/model/config.toml",
-        help="Config file to be loaded and used for inference.",
-    )
-    parser.add_argument(
-        "--uris", help="Path to a file containing the list of uris to use."
-    )
-    parser.add_argument(
-        "--wavs",
-        default="data/debug/wav",
-        help="Folder containing the audio files to run inference on.",
-    )
-    parser.add_argument(
-        "--checkpoint",
-        default="VTC-2/model/best.ckpt",
-        help="Path to a pretrained model checkpoint.",
-    )
-    parser.add_argument(
-        "--output",
-        required=True,
-        help="Output Path to the folder that will contain the final predictions.",
-    )
-    parser.add_argument("--save_probs", action="store_true",
-                        help="Save averaged per-frame probabilities to disk.")
-    parser.add_argument("--stride_pct", default=0.25, type=float,
-                        help="Sliding window stride as a fraction of chunk_duration_s (0.5 = 50%% overlap, 0.25 = 75%%).")
-    parser.add_argument(
-        "--thresholds",
-        type=Path,
-        default=Path("thresholds/f1.toml"),
-        help="If thresholds dict is given, perform predictions using thresholding.",
-    )
-    parser.add_argument(
-        "--high_precision",
-        action="store_true",
-        help="Loads the high precision thresholds, overwrites the `--thresholds` argument.",
-    )
-    parser.add_argument(
-        "--min_duration_on_s",
-        default=0.1,
-        type=float,
-        help="Remove speech segments shorter than that many seconds.",
-    )
-    parser.add_argument(
-        "--min_duration_off_s",
-        default=0.1,
-        type=float,
-        help="Fill same-speaker gaps shorter than that many seconds.",
-    )
-    parser.add_argument(
-        "--batch_size",
-        default=128,
-        type=int,
-        help="Batch size to use for the forward pass of the model.",
-    )
-    parser.add_argument(
-        "--recursive_search",
-        action="store_true",
-        help="Recursively search for `.wav` files. Might be slow. Defaults to False.",
-    )
-    parser.add_argument(
-        "--device",
-        default="cuda",
-        choices=["gpu", "cuda", "cpu", "mps"],
-        help="Size of the batch used for the forward pass in the model.",
-    )
-    parser.add_argument(
-        "--keep_raw",
-        action="store_true",
-        help="If active, the raw RTTM will be kept and saved to disk in the `<output>/raw_rttm/` folder and a `<output>/raw_rttm.csv` file will be created.",
-    )
-
-    args = parser.parse_args()
-
-    main(**vars(args))
diff --git a/src/vtc_inference/__init__.py b/src/vtc_inference/__init__.py
new file mode 100644
index 0000000..f937386
--- /dev/null
+++ b/src/vtc_inference/__init__.py
@@ -0,0 +1,4 @@
+from .infer import VtcInferenceParams, run_vtc_with_params, run_vtc
+
+
+__all__ = ["VtcInferenceParams", "run_vtc_with_params", "run_vtc"]
diff --git a/src/vtc_inference/__main__.py b/src/vtc_inference/__main__.py
new file mode 100644
index 0000000..d401cb0
--- /dev/null
+++ b/src/vtc_inference/__main__.py
@@ -0,0 +1,4 @@
+from .cli import vtc_infer_cli
+
+
+vtc_infer_cli()
diff --git a/src/vtc_inference/cli.py b/src/vtc_inference/cli.py
new file mode 100644
index 0000000..20a70c3
--- /dev/null
+++ b/src/vtc_inference/cli.py
@@ -0,0 +1,103 @@
+import argparse
+from .infer import prepare_vtc_inference, run_vtc_with_params
+
+
+def vtc_infer_cli() -> None:
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--config",
+        type=str,
+        default="2.1",
+        help="Local config.toml file/directory, or a known model version key (e.g. '2.1', '2.0').",
+    )
+    parser.add_argument(
+        "--uris", help="Path to a file containing the list of uris to use."
+    )
+    parser.add_argument(
+        "--wavs",
+        default="data/debug/wav",
+        help="Folder containing the audio files to run inference on.",
+    )
+    parser.add_argument(
+        "--checkpoint",
+        default="2.1",
+        type=str,
+        help="Local checkpoint file/directory, or a known model version key (e.g. '2.1', '2.0').",
+    )
+    parser.add_argument(
+        "--output",
+        required=True,
+        help="Output Path to the folder that will contain the final predictions.",
+    )
+    parser.add_argument(
+        "--save_probs",
+        action="store_true",
+        help="Save averaged per-frame probabilities to disk.",
+    )
+    parser.add_argument(
+        "--stride_pct",
+        default=0.25,
+        type=float,
+        help="Sliding window stride as a fraction of chunk_duration_s (0.5 = 50%% overlap, 0.25 = 75%%).",
+    )
+    parser.add_argument(
+        "--thresholds-location",
+        type=str,
+        default="2.1",
+        help=(
+            "Local directory/file containing the thresholds TOML, or a known "
+            "model version key. Defaults to wherever --checkpoint resolves to."
+        ),
+    )
+    parser.add_argument(
+        "--thresholds-name",
+        type=str,
+        default="f1",
+        help=(
+            "Thresholds file to use (without extension), e.g. 'f1' -> "
+            "'<location>/thresholds/f1.toml'. Ignored if --high-precision is set."
+        ),
+    )
+    parser.add_argument(
+        "--high-precision",
+        action="store_true",
+        help="Use the high-precision thresholds ('hp'), overriding custom thresholds.",
+    )
+    parser.add_argument(
+        "--min_duration_on_s",
+        default=0.1,
+        type=float,
+        help="Remove speech segments shorter than that many seconds.",
+    )
+    parser.add_argument(
+        "--min_duration_off_s",
+        default=0.1,
+        type=float,
+        help="Fill same-speaker gaps shorter than that many seconds.",
+    )
+    parser.add_argument(
+        "--batch_size",
+        default=128,
+        type=int,
+        help="Batch size to use for the forward pass of the model.",
+    )
+    parser.add_argument(
+        "--recursive_search",
+        action="store_true",
+        help="Recursively search for `.wav` files. Might be slow. Defaults to False.",
+    )
+    parser.add_argument(
+        "--device",
+        default="cuda",
+        choices=["gpu", "cuda", "cpu", "mps"],
+        help="Size of the batch used for the forward pass in the model.",
+    )
+    parser.add_argument(
+        "--keep_raw",
+        action="store_true",
+        help="If active, the raw RTTM will be kept and saved to disk in the `<output>/raw_rttm/` folder and a `<output>/raw_rttm.csv` file will be created.",
+    )
+
+    args = parser.parse_args()
+    params = prepare_vtc_inference(**vars(args))
+    run_vtc_with_params(params)
diff --git a/scripts/convert.py b/src/vtc_inference/convert_audios.py
similarity index 97%
rename from scripts/convert.py
rename to src/vtc_inference/convert_audios.py
index cec6038..3c63a48 100644
--- a/scripts/convert.py
+++ b/src/vtc_inference/convert_audios.py
@@ -53,7 +53,7 @@ def convert_audios(
         encoder.to_file(output / audio_p.name)
 
 
-if __name__ == "__main__":
+def convert_audios_cli() -> None:
     import argparse
 
     parser = argparse.ArgumentParser()
@@ -75,3 +75,7 @@ def convert_audios(
         output=args.output,
         allow_upsampling=args.allow_upsampling,
     )
+
+
+if __name__ == "__main__":
+    convert_audios_cli()
diff --git a/src/vtc_inference/infer.py b/src/vtc_inference/infer.py
new file mode 100644
index 0000000..ffd2c91
--- /dev/null
+++ b/src/vtc_inference/infer.py
@@ -0,0 +1,160 @@
+import logging
+import shutil
+import typing as t
+from pathlib import Path
+
+from segma.inference import run_inference_on_audios
+
+from .ressources import (
+    VersionStr,
+    resolve_model_config_path,
+    resolve_model_path,
+    resolve_thresholds_path,
+)
+from .utils import merge_segments, write_rttm_csvs
+
+logging.basicConfig(
+    level=logging.INFO,
+    format="[%(asctime)s] [%(levelname)s] - %(message)s",
+    datefmt="%Y.%m.%d %H:%M:%S",
+)
+logger = logging.getLogger("inference")
+
+
+class VtcInferenceParams(t.NamedTuple):
+    """Resolved, ready-to-run parameters for `run_vtc_inference`."""
+
+    output: Path
+    config_path: Path
+    checkpoint_path: Path
+    thresholds_path: Path | None
+    uris: Path | None
+    wavs: str
+    save_probs: bool
+    min_duration_on_s: float
+    min_duration_off_s: float
+    batch_size: int
+    stride_pct: float
+    write_empty: bool
+    write_csv: bool
+    recursive_search: bool
+    device: t.Literal["gpu", "cuda", "cpu", "mps"]
+    keep_raw: bool
+
+
+def prepare_vtc_inference(
+    output: str,
+    uris: Path | None = None,
+    config: Path | VersionStr = "2.1",
+    wavs: str = "data/debug/wav",
+    checkpoint: Path | VersionStr = "2.1",
+    save_probs: bool = False,
+    high_precision: bool = False,
+    thresholds: str | None = "f1",
+    thresholds_location: Path | VersionStr = "2.1",
+    min_duration_on_s: float = 0.1,
+    min_duration_off_s: float = 0.1,
+    batch_size: int = 128,
+    stride_pct: float = 0.25,
+    write_empty: bool = True,
+    write_csv: bool = True,
+    recursive_search: bool = False,
+    device: t.Literal["gpu", "cuda", "cpu", "mps"] = "gpu",
+    keep_raw: bool = False,
+) -> VtcInferenceParams:
+    """Resolve raw arguments into concrete filesystem paths.
+
+    Resolves `config` via `resolve_model_config_path`, `checkpoint` via
+    `resolve_model_path`, and the thresholds file (name forced to `"hp"`
+    when `high_precision` is set) via `resolve_thresholds_path` against
+    `checkpoint`. Creates `output` and copies the resolved thresholds file
+    into it.
+
+    Raises:
+        ModelNotFoundError: if `config`, `checkpoint`, or `thresholds`
+            cannot be resolved.
+    """
+
+    output_dir = Path(output)
+    output_dir.mkdir(parents=True, exist_ok=True)
+
+    config_path = resolve_model_config_path(config)
+    checkpoint_path = resolve_model_path(checkpoint)
+
+    # Note: this overrides a custom threshold
+    thresholds_path: Path | None = None
+    if high_precision:
+        thresholds_path = resolve_thresholds_path("hp", "2.1")
+    else:
+        location = thresholds_location if thresholds_location is not None else checkpoint
+        thresholds_name = thresholds if thresholds else "f1"
+        thresholds_path = resolve_thresholds_path(thresholds_name, location)
+
+
+    return VtcInferenceParams(
+        output=output_dir,
+        config_path=config_path,
+        checkpoint_path=checkpoint_path,
+        thresholds_path=thresholds_path,
+        uris=uris,
+        wavs=wavs,
+        save_probs=save_probs,
+        min_duration_on_s=min_duration_on_s,
+        min_duration_off_s=min_duration_off_s,
+        batch_size=batch_size,
+        stride_pct=stride_pct,
+        write_empty=write_empty,
+        write_csv=write_csv,
+        recursive_search=recursive_search,
+        device=device,
+        keep_raw=keep_raw,
+    )
+
+
+def run_vtc_with_params(params: VtcInferenceParams) -> None:
+    """Run VTC inference and write merged (and optionally raw) RTTM/CSV outputs.
+
+    Raises:
+        OSError: if raw RTTM cleanup fails.
+    """
+    logger.info("Running inference on audio files.")
+    processed_files = run_inference_on_audios(
+        config=params.config_path,
+        uris=params.uris,
+        wavs=params.wavs,  # pyright: ignore[reportArgumentType]
+        checkpoint=params.checkpoint_path,
+        output=params.output,
+        thresholds=params.thresholds_path,
+        batch_size=params.batch_size,
+        device=params.device,
+        recursive=params.recursive_search,
+        save_probs=params.save_probs,
+        stride_pct=params.stride_pct,
+        logger=logger,
+    )
+
+    logger.info("Merging detected speech segments.")
+    merge_segments(
+        file_uris_to_merge=[f.stem for f in processed_files],
+        output=params.output,  # pyright: ignore[reportArgumentType]
+        min_duration_on_s=params.min_duration_on_s,
+        min_duration_off_s=params.min_duration_off_s,
+        write_empty=params.write_empty,
+    )
+
+    if not params.keep_raw:
+        # NOTE - remove <output>/raw_rttm
+        shutil.rmtree(str(Path(params.output / "raw_rttm").absolute()))
+
+    if params.write_csv:
+        write_rttm_csvs(params.output, keep_raw=params.keep_raw)
+
+    logger.info(
+        f"Inference finished, files can be found here: '{params.output.absolute()}/'"
+    )
+
+
+def run_vtc(*args: t.Any, **kwargs: t.Any) -> None:
+    """entry point combining argument resolution and inference."""
+    params = prepare_vtc_inference(*args, **kwargs)
+    run_vtc_with_params(params)
diff --git a/src/vtc_inference/ressources.py b/src/vtc_inference/ressources.py
new file mode 100644
index 0000000..0f3fa92
--- /dev/null
+++ b/src/vtc_inference/ressources.py
@@ -0,0 +1,161 @@
+import typing as t
+from pathlib import Path
+
+from huggingface_hub import snapshot_download
+
+VersionStr: t.TypeAlias = str
+REPO_ID = "coml/VTC-2"
+MODEL_TO_REVISION = {
+    "2.1": "91167f6",
+    "2.0": "91e67b5",
+}
+_WEIGHTS_SUFFIXES = ("*.pt", "*.ckpt")
+
+
+class ModelResolutionError(Exception):
+    """Base error for model-artifact resolution failures."""
+
+
+class ModelNotFoundError(ModelResolutionError):
+    """Raised when a model artifact cannot be resolved from any source."""
+
+
+def _resolve_named_file(directory_or_file: Path, filename: str) -> Path | None:
+    """Return `filename`, if `directory_or_file` is that file or a directory containing it."""
+    resolved = directory_or_file
+    if resolved.is_file() and resolved.name == filename:
+        return directory_or_file
+    elif resolved.is_dir():
+        matches = list(directory_or_file.rglob(filename))
+        if len(matches) != 1:
+            return None
+        return matches[0]
+    return None
+
+
+def _resolve_pt_path(directory_or_file: Path) -> Path | None:
+    """Return the model weights file inside `directory_or_file`, if present.
+
+    Accepts a direct path to a `.pt`/`.ckpt` file, or a directory to search
+    for one. If multiple candidates are found, a file whose stem is `best`
+    is preferred; otherwise the first match (sorted) is used.
+    """
+    resolved = directory_or_file
+    if resolved.is_file() and resolved.suffix in (".pt", ".ckpt"):
+        return directory_or_file
+    elif resolved.is_dir():
+        matches = sorted(
+            match
+            for pattern in _WEIGHTS_SUFFIXES
+            for match in directory_or_file.rglob(pattern)
+        )
+        if not matches:
+            return None
+        for match in matches:
+            if match.stem == "best":
+                return match
+        return matches[0]
+    return None
+
+
+def _download_repo(model_version: VersionStr) -> Path:
+    """Fetch (or reuse the cached copy of) the HF Hub snapshot for `model_version`.
+
+    Raises:
+        ModelNotFoundError: if `model_version` is not a known MODEL_TO_REVISION key.
+    """
+    revision = MODEL_TO_REVISION.get(model_version)
+    if revision is None:
+        raise ModelNotFoundError(f"Unknown model version: {model_version!r}")
+    return Path(snapshot_download(repo_id="coml/VTC-2", revision=revision))
+
+
+def resolve_model_path(model: Path | VersionStr) -> Path:
+    """Resolve a model reference to its `.pt | .ckpt` weights file.
+
+    Tries, in order: `model` as a local directory containing a single `.pt | .ckpt`
+    file; `model` as a string naming such a directory; `model` as a
+    MODEL_TO_REVISION key, in which case the HF Hub repo is fetched and
+    `model/best.ckpt` is used.
+
+    Raises:
+        ModelNotFoundError: if none of the above resolve to an existing file.
+    """
+    if isinstance(model, Path):
+        pt_file = _resolve_pt_path(model)
+        if pt_file is None:
+            raise ModelNotFoundError(f"No single .pt or .ckpt file found in {model}")
+        return pt_file
+
+    local_dir = Path(model)
+    if local_dir.is_dir() or local_dir.is_file():
+        pt_file = _resolve_pt_path(local_dir)
+        if pt_file is not None:
+            return pt_file
+
+    repo_dir = _download_repo(model)
+    pt_file = repo_dir / "model" / "best.ckpt"
+    if not pt_file.is_file():
+        raise ModelNotFoundError(f"{pt_file} not found in downloaded repo")
+    return pt_file
+
+
+def resolve_model_config_path(model: Path | VersionStr) -> Path:
+    """Resolve a model reference to its `config.toml`.
+
+    Tries, in order: `model` as a direct path to `config.toml` or a local
+    directory containing it; `model` as a string naming the same; `model`
+    as a MODEL_TO_REVISION key, in which case the HF Hub repo is fetched
+    and `model/config.toml` is used.
+
+    Raises:
+        ModelNotFoundError: if none of the above resolve to an existing file.
+    """
+    if isinstance(model, Path):
+        config_file = _resolve_named_file(model, "config.toml")
+        if config_file is None:
+            raise ModelNotFoundError(f"No config.toml found in {model}")
+        return config_file
+
+    local_path = Path(model)
+    if local_path.is_dir() or local_path.is_file():
+        config_file = _resolve_named_file(local_path, "config.toml")
+        if config_file is not None:
+            return config_file
+
+    repo_dir = _download_repo(model)
+    config_file = repo_dir / "model" / "config.toml"
+    if not config_file.is_file():
+        raise ModelNotFoundError(f"{config_file} not found in downloaded repo")
+    return config_file
+
+
+def resolve_thresholds_path(name: str, location: Path | VersionStr) -> Path:
+    """Resolve a thresholds file named `name` relative to `location`.
+
+    Tries, in order: `location` as a direct path to `{name}.toml` or a local
+    directory containing it; `location` as a string naming the same;
+    `location` as a MODEL_TO_REVISION key, in which case the HF Hub repo is
+    fetched and `thresholds/{name}.toml` is used.
+
+    Raises:
+        ModelNotFoundError: if none of the above resolve to an existing file.
+    """
+    filename = f"{name}.toml"
+    if isinstance(location, Path):
+        thresholds_file = _resolve_named_file(location, filename)
+        if thresholds_file is None:
+            raise ModelNotFoundError(f"No {filename} found in {location}")
+        return thresholds_file
+
+    local_path = Path(location)
+    if local_path.is_dir() or local_path.is_file():
+        thresholds_file = _resolve_named_file(local_path, filename)
+        if thresholds_file is not None:
+            return thresholds_file
+
+    repo_dir = _download_repo(location)
+    thresholds_file = repo_dir / "thresholds" / filename
+    if not thresholds_file.is_file():
+        raise ModelNotFoundError(f"{thresholds_file} not found in downloaded repo")
+    return thresholds_file
diff --git a/src/vtc_inference/utils.py b/src/vtc_inference/utils.py
new file mode 100644
index 0000000..3a4d084
--- /dev/null
+++ b/src/vtc_inference/utils.py
@@ -0,0 +1,167 @@
+import copy
+from pathlib import Path
+
+import polars as pl
+from pyannote.core import Annotation, Segment
+from segma.utils.io import get_audio_info
+
+
+def load_aa(path: Path):
+    data = pl.read_csv(
+        source=path,
+        has_header=False,
+        new_columns=("uid", "start_time_s", "duration_s", "label"),
+        schema={
+            "uid": pl.String(),
+            "start_time_s": pl.Float64(),
+            "duration_s": pl.Float64(),
+            "label": pl.String(),
+        },
+        separator=" ",
+    )
+    return data
+
+
+def load_rttm(path: Path | str) -> pl.DataFrame:
+    try:
+        data = pl.read_csv(
+            source=path,
+            has_header=False,
+            columns=[1, 3, 4, 7],
+            new_columns=("uid", "start_time_s", "duration_s", "label"),
+            schema_overrides={
+                "uid": pl.String(),
+                "start_time_s": pl.Float64(),
+                "duration_s": pl.Float64(),
+                "label": pl.String(),
+            },
+            separator=" ",
+        )
+    except pl.exceptions.NoDataError:
+        data = pl.DataFrame(
+            None,
+            {
+                "uid": pl.String(),
+                "start_time_s": pl.Float64(),
+                "duration_s": pl.Float64(),
+                "label": pl.String(),
+            },
+        )
+
+    return data
+
+
+def load_one_uri(uri_df: pl.DataFrame):
+    for uids, turns in uri_df.group_by("uid"):
+        uid = uids[0]
+        annotation = Annotation(uri=uid)
+        for i, turn in enumerate(turns.iter_rows(named=True)):
+            segment = Segment(
+                turn["start_time_s"], turn["start_time_s"] + turn["duration_s"]
+            )
+            annotation[segment, i] = turn["label"]
+        yield uid, annotation
+
+
+def process_annot(
+    annotation: Annotation,
+    min_duration_off_s: float = 0.1,
+    min_duration_on_s: float = 0.1,
+) -> Annotation:
+    """Create a new `Annotation` with the `min_duration_off` and `min_duration_off` rules applied.
+
+    Args:
+        annotation (Annotation): input annotation
+        min_duration_off_s (float, optional): Remove speech segments shorter than that many seconds. Defaults to 0.1.
+        min_duration_on_s (float, optional): Fill same-speaker gaps shorter than that many seconds. Defaults to 0.1.
+
+    Returns:
+        Annotation: Processed annotation.
+    """
+    active = copy.deepcopy(annotation)
+    # NOTE - Fill regions shorter than that many seconds.
+    if min_duration_off_s > 0.0:
+        active = active.support(collar=min_duration_off_s)
+    # NOTE - remove regions shorter than that many seconds.
+    if min_duration_on_s > 0:
+        for segment, track in list(active.itertracks()):
+            if segment.duration < min_duration_on_s:
+                del active[segment, track]
+    return active
+
+
+def merge_segments(
+    file_uris_to_merge: list[str],
+    output: str,
+    min_duration_on_s: float = 0.1,
+    min_duration_off_s: float = 0.1,
+    write_empty: bool = True,
+):
+    output = Path(output)
+    raw_output_p = output / "raw_rttm"
+
+    # NOTE - merge RTTMs
+    uri_to_annot: dict[str, Annotation] = {}
+    uri_to_proc_annot: dict[str, Annotation] = {}
+    merged_out_p = output / "rttm"
+    merged_out_p.mkdir(exist_ok=True, parents=True)
+
+    for file_uri in file_uris_to_merge:
+        file = raw_output_p / f"{file_uri}.rttm"
+        if not file.exists():
+            continue
+
+        match file.suffix:
+            case ".aa":
+                data = load_aa(file)
+            case ".rttm":
+                data = load_rttm(file)
+            case _:
+                raise ValueError(
+                    f"File not found error, extension is not supported: {file}"
+                )
+
+        # NOTE - process, should handle the case where a single rttm contains multiple URIS
+        for uri, annot in load_one_uri(data):
+            uri_to_annot[uri] = annot
+            uri_to_proc_annot[uri] = process_annot(
+                annotation=annot,
+                min_duration_off_s=min_duration_off_s,
+                min_duration_on_s=min_duration_on_s,
+            )
+
+        for uri, annot in uri_to_proc_annot.items():
+            (merged_out_p / f"{uri}.rttm").write_text(annot.to_rttm())
+
+    # NOTE - Writting missing rttm files
+    if write_empty:
+        for uri in set(file_uris_to_merge) - set(uri_to_proc_annot.keys()):
+            (merged_out_p / f"{uri}.rttm").touch()
+
+
+def check_audio_files(audio_files_to_process: list[Path]) -> None:
+    """Fails if the audios are not sampled at 16_000 Hz and contain more than one channel."""
+
+    for wav_p in audio_files_to_process:
+        info = get_audio_info(wav_p)
+        # NOTE - check that the audio is valid
+        if not info.sample_rate == 16_000:
+            raise ValueError(
+                f"file `{wav_p}` is not samlped at 16 000 hz. Please convert your audio files."
+            )
+        if not info.n_channels == 1:
+            raise ValueError(
+                f"file `{wav_p}` has more than one channel. You can average your channels or use another channel reduction technique."
+            )
+
+
+def write_rttm_csvs(output: Path, *, keep_raw: bool) -> None:
+    """Load merged (and optionally raw) RTTM files and write them out as CSVs."""
+    if keep_raw:
+        raw_rttm_files = sorted((output / "raw_rttm").glob("*.rttm"))
+        raw_rttm_dfs = [load_rttm(f) for f in raw_rttm_files]
+        pl.concat(raw_rttm_dfs).write_csv(output / "raw_rttm.csv")
+
+    rttm_files = sorted((output / "rttm").glob("*.rttm"))
+    rttm_dfs = [load_rttm(f) for f in rttm_files]
+    pl.concat(rttm_dfs).write_csv(output / "rttm.csv")
diff --git a/thresholds/f1.toml b/thresholds/f1.toml
deleted file mode 100644
index 1af4bf7..0000000
--- a/thresholds/f1.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-[KCHI]
-lower_bound = 0.437
-upper_bound = 1.0
-
-[OCH]
-lower_bound = 0.378
-upper_bound = 1.0
-
-[MAL]
-lower_bound = 0.429
-upper_bound = 1.0
-
-[FEM]
-lower_bound = 0.377
-upper_bound = 1.0
diff --git a/thresholds/hp.toml b/thresholds/hp.toml
deleted file mode 100644
index 06cbeed..0000000
--- a/thresholds/hp.toml
+++ /dev/null
@@ -1,15 +0,0 @@
-[KCHI]
-lower_bound = 0.8
-upper_bound = 1.0
-
-[OCH]
-lower_bound = 0.52
-upper_bound = 1.0
-
-[MAL]
-lower_bound = 0.75
-upper_bound = 1.0
-
-[FEM]
-lower_bound = 0.8
-upper_bound = 1.0
diff --git a/uv.lock b/uv.lock
index cf1d2bb..3e2fb8f 100644
--- a/uv.lock
+++ b/uv.lock
@@ -104,6 +104,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/c5/7c/83ff6046176a675e6a1e8aeefed8892cd97fe7c46af93cc540d1b24b8323/asteroid_filterbanks-0.4.0-py3-none-any.whl", hash = "sha256:4932ac8b6acc6e08fb87cbe8ece84215b5a74eee284fe83acf3540a72a02eaf5", size = 29912, upload-time = "2021-04-09T20:03:05.817Z" },
 ]
 
+[[package]]
+name = "asttokens"
+version = "3.0.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/25/1e/faf0f247f6f881b98fc4d6d07e14085cb89d13665084e6d6ac1dc2c03d0b/asttokens-3.0.2.tar.gz", hash = "sha256:3ecdbd8f2cc195f53ccada3a613538bb5f9ef6f6869129f13e03c30a677b8fe2", size = 63136, upload-time = "2026-07-12T03:31:49.084Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/d4/2b/04b8a15f3a1c77bc79ddf5c73875327f34b4fa75982df2b76e45e402d364/asttokens-3.0.2-py3-none-any.whl", hash = "sha256:9da13157f5b28becde0bd374fc677dcd3c290614264eff096f167c469cd9f933", size = 28702, upload-time = "2026-07-12T03:31:47.542Z" },
+]
+
 [[package]]
 name = "attrs"
 version = "25.3.0"
@@ -295,6 +304,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/e7/05/c19819d5e3d95294a6f5947fb9b9629efb316b96de511b418c53d245aae6/cycler-0.12.1-py3-none-any.whl", hash = "sha256:85cef7cff222d8644161529808465972e51340599459b8ac3ccbac5a854e0d30", size = 8321, upload-time = "2023-10-07T05:32:16.783Z" },
 ]
 
+[[package]]
+name = "decorator"
+version = "5.3.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/60/8b/32f9823da46cde7df2087faa08cd98d01b908f8dcab982cdba9c84e85355/decorator-5.3.1.tar.gz", hash = "sha256:4cbcdd55a6efadb9dbea26b858f4fb3264567b52d69ca0d25b721b553f60ea82", size = 58084, upload-time = "2026-05-18T06:03:28.057Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/05/7f/798705f5296a58ca505d600456748d1be48078eac8a7050d8a98bc9edb89/decorator-5.3.1-py3-none-any.whl", hash = "sha256:f47fe6fdbd2edd623ecfe36875d37aba411624e2670dd395dddae1358689bb3c", size = 10365, upload-time = "2026-05-18T06:03:26.517Z" },
+]
+
 [[package]]
 name = "deepmerge"
 version = "2.0"
@@ -319,6 +337,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/87/62/9773de14fe6c45c23649e98b83231fffd7b9892b6cf863251dc2afa73643/einops-0.8.1-py3-none-any.whl", hash = "sha256:919387eb55330f5757c6bea9165c5ff5cfe63a642682ea788a6d472576d81737", size = 64359, upload-time = "2025-02-09T03:17:01.998Z" },
 ]
 
+[[package]]
+name = "executing"
+version = "2.2.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/cc/28/c14e053b6762b1044f34a13aab6859bbf40456d37d23aa286ac24cfd9a5d/executing-2.2.1.tar.gz", hash = "sha256:3632cc370565f6648cc328b32435bd120a1e4ebb20c77e3fdde9a13cd1e533c4", size = 1129488, upload-time = "2025-09-01T09:48:10.866Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/c1/ea/53f2148663b321f21b5a606bd5f191517cf40b7072c0497d3c92c4a13b1e/executing-2.2.1-py2.py3-none-any.whl", hash = "sha256:760643d3452b4d777d295bb167ccc74c64a81df23fb5e08eff250c425a4b2017", size = 28317, upload-time = "2025-09-01T09:48:08.5Z" },
+]
+
 [[package]]
 name = "filelock"
 version = "3.19.1"
@@ -451,7 +478,6 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/49/e8/58c7f85958bda41dafea50497cbd59738c5c43dbbea5ee83d651234398f4/greenlet-3.2.4-cp313-cp313-macosx_11_0_universal2.whl", hash = "sha256:1a921e542453fe531144e91e1feedf12e07351b1cf6c9e8a3325ea600a715a31", size = 272814, upload-time = "2025-08-07T13:15:50.011Z" },
     { url = "https://files.pythonhosted.org/packages/62/dd/b9f59862e9e257a16e4e610480cfffd29e3fae018a68c2332090b53aac3d/greenlet-3.2.4-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:cd3c8e693bff0fff6ba55f140bf390fa92c994083f838fece0f63be121334945", size = 641073, upload-time = "2025-08-07T13:42:57.23Z" },
     { url = "https://files.pythonhosted.org/packages/f7/0b/bc13f787394920b23073ca3b6c4a7a21396301ed75a655bcb47196b50e6e/greenlet-3.2.4-cp313-cp313-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:710638eb93b1fa52823aa91bf75326f9ecdfd5e0466f00789246a5280f4ba0fc", size = 655191, upload-time = "2025-08-07T13:45:29.752Z" },
-    { url = "https://files.pythonhosted.org/packages/f2/d6/6adde57d1345a8d0f14d31e4ab9c23cfe8e2cd39c3baf7674b4b0338d266/greenlet-3.2.4-cp313-cp313-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:c5111ccdc9c88f423426df3fd1811bfc40ed66264d35aa373420a34377efc98a", size = 649516, upload-time = "2025-08-07T13:53:16.314Z" },
     { url = "https://files.pythonhosted.org/packages/7f/3b/3a3328a788d4a473889a2d403199932be55b1b0060f4ddd96ee7cdfcad10/greenlet-3.2.4-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:d76383238584e9711e20ebe14db6c88ddcedc1829a9ad31a584389463b5aa504", size = 652169, upload-time = "2025-08-07T13:18:32.861Z" },
     { url = "https://files.pythonhosted.org/packages/ee/43/3cecdc0349359e1a527cbf2e3e28e5f8f06d3343aaf82ca13437a9aa290f/greenlet-3.2.4-cp313-cp313-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:23768528f2911bcd7e475210822ffb5254ed10d71f4028387e5a99b4c6699671", size = 610497, upload-time = "2025-08-07T13:18:31.636Z" },
     { url = "https://files.pythonhosted.org/packages/b8/19/06b6cf5d604e2c382a6f31cafafd6f33d5dea706f4db7bdab184bad2b21d/greenlet-3.2.4-cp313-cp313-musllinux_1_1_aarch64.whl", hash = "sha256:00fadb3fedccc447f517ee0d3fd8fe49eae949e1cd0f6a611818f4f6fb7dc83b", size = 1121662, upload-time = "2025-08-07T13:42:41.117Z" },
@@ -462,7 +488,6 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/22/5c/85273fd7cc388285632b0498dbbab97596e04b154933dfe0f3e68156c68c/greenlet-3.2.4-cp314-cp314-macosx_11_0_universal2.whl", hash = "sha256:49a30d5fda2507ae77be16479bdb62a660fa51b1eb4928b524975b3bde77b3c0", size = 273586, upload-time = "2025-08-07T13:16:08.004Z" },
     { url = "https://files.pythonhosted.org/packages/d1/75/10aeeaa3da9332c2e761e4c50d4c3556c21113ee3f0afa2cf5769946f7a3/greenlet-3.2.4-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:299fd615cd8fc86267b47597123e3f43ad79c9d8a22bebdce535e53550763e2f", size = 686346, upload-time = "2025-08-07T13:42:59.944Z" },
     { url = "https://files.pythonhosted.org/packages/c0/aa/687d6b12ffb505a4447567d1f3abea23bd20e73a5bed63871178e0831b7a/greenlet-3.2.4-cp314-cp314-manylinux2014_ppc64le.manylinux_2_17_ppc64le.whl", hash = "sha256:c17b6b34111ea72fc5a4e4beec9711d2226285f0386ea83477cbb97c30a3f3a5", size = 699218, upload-time = "2025-08-07T13:45:30.969Z" },
-    { url = "https://files.pythonhosted.org/packages/dc/8b/29aae55436521f1d6f8ff4e12fb676f3400de7fcf27fccd1d4d17fd8fecd/greenlet-3.2.4-cp314-cp314-manylinux2014_s390x.manylinux_2_17_s390x.whl", hash = "sha256:b4a1870c51720687af7fa3e7cda6d08d801dae660f75a76f3845b642b4da6ee1", size = 694659, upload-time = "2025-08-07T13:53:17.759Z" },
     { url = "https://files.pythonhosted.org/packages/92/2e/ea25914b1ebfde93b6fc4ff46d6864564fba59024e928bdc7de475affc25/greenlet-3.2.4-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:061dc4cf2c34852b052a8620d40f36324554bc192be474b9e9770e8c042fd735", size = 695355, upload-time = "2025-08-07T13:18:34.517Z" },
     { url = "https://files.pythonhosted.org/packages/72/60/fc56c62046ec17f6b0d3060564562c64c862948c9d4bc8aa807cf5bd74f4/greenlet-3.2.4-cp314-cp314-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:44358b9bf66c8576a9f57a590d5f5d6e72fa4228b763d0e43fee6d3b06d3a337", size = 657512, upload-time = "2025-08-07T13:18:33.969Z" },
     { url = "https://files.pythonhosted.org/packages/23/6e/74407aed965a4ab6ddd93a7ded3180b730d281c77b765788419484cdfeef/greenlet-3.2.4-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:2917bdf657f5859fbf3386b12d68ede4cf1f04c90c3a6bc1f013dd68a22e2269", size = 1612508, upload-time = "2025-11-04T12:42:23.427Z" },
@@ -532,6 +557,52 @@ version = "0.2.7"
 source = { registry = "https://pypi.org/simple" }
 sdist = { url = "https://files.pythonhosted.org/packages/72/84/9f71369fc868dc963ddf51d1bfd8853a9793a37a21c9081a433f6e81d56a/interlap-0.2.7.tar.gz", hash = "sha256:31e4f30c54b067c4939049f5d8131ae5e2fa682ec71aa56f89c0e5b900806ec9", size = 6066, upload-time = "2020-10-02T17:54:06.464Z" }
 
+[[package]]
+name = "ipython"
+version = "9.15.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "colorama", marker = "sys_platform == 'win32'" },
+    { name = "decorator" },
+    { name = "ipython-pygments-lexers" },
+    { name = "jedi" },
+    { name = "matplotlib-inline" },
+    { name = "pexpect", marker = "sys_platform != 'emscripten' and sys_platform != 'win32'" },
+    { name = "prompt-toolkit" },
+    { name = "psutil", marker = "sys_platform != 'cygwin' and sys_platform != 'emscripten'" },
+    { name = "pygments" },
+    { name = "stack-data" },
+    { name = "traitlets" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/53/59/165d3b4d75cc34add3122c4417ecb229085140ac573103c223cd01dde96f/ipython-9.15.0.tar.gz", hash = "sha256:da2819ce2aa83135257df830660b1176d986c3d2876db24df01974fa955b2756", size = 4442580, upload-time = "2026-06-26T11:03:35.913Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/40/3a/948263ca3b9d65bb2b1b0c521b3a49fad5d59ada58724bd87d2bd5ff3f36/ipython-9.15.0-py3-none-any.whl", hash = "sha256:515ad9c3cdf0c932a5a9f6245419e8aba706b7bd03c3e1d3a1c83d9351d6aa6e", size = 630895, upload-time = "2026-06-26T11:03:33.809Z" },
+]
+
+[[package]]
+name = "ipython-pygments-lexers"
+version = "1.1.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "pygments" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/ef/4c/5dd1d8af08107f88c7f741ead7a40854b8ac24ddf9ae850afbcf698aa552/ipython_pygments_lexers-1.1.1.tar.gz", hash = "sha256:09c0138009e56b6854f9535736f4171d855c8c08a563a0dcd8022f78355c7e81", size = 8393, upload-time = "2025-01-17T11:24:34.505Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/d9/33/1f075bf72b0b747cb3288d011319aaf64083cf2efef8354174e3ed4540e2/ipython_pygments_lexers-1.1.1-py3-none-any.whl", hash = "sha256:a9462224a505ade19a605f71f8fa63c2048833ce50abc86768a0d81d876dc81c", size = 8074, upload-time = "2025-01-17T11:24:33.271Z" },
+]
+
+[[package]]
+name = "jedi"
+version = "0.20.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "parso" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/46/b7/a3635f6a2d7cf5b5dd98064fc1d5fbbafcb25477bcea204a3a92145d158b/jedi-0.20.0.tar.gz", hash = "sha256:c3f4ccbd276696f4b19c54618d4fb18f9fc24b0aef02acf704b23f487daa1011", size = 3119416, upload-time = "2026-05-01T23:38:47.814Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/9a/93/242e2eab5fe682ffcb8b0084bde703a41d51e17ee0f3a31ff0d9d813620a/jedi-0.20.0-py2.py3-none-any.whl", hash = "sha256:7bdd9c2634f56713299976f4cbd59cb3fa92165cc5e05ea811fb253480728b67", size = 4884812, upload-time = "2026-05-01T23:38:43.919Z" },
+]
+
 [[package]]
 name = "jinja2"
 version = "3.1.6"
@@ -763,6 +834,18 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/e8/62/aeabeef1a842b6226a30d49dd13e8a7a1e81e9ec98212c0b5169f0a12d83/matplotlib-3.10.6-cp314-cp314t-win_arm64.whl", hash = "sha256:4dd83e029f5b4801eeb87c64efd80e732452781c16a9cf7415b7b63ec8f374d7", size = 8172588, upload-time = "2025-08-30T00:14:11.166Z" },
 ]
 
+[[package]]
+name = "matplotlib-inline"
+version = "0.2.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "traitlets" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/bd/c0/9f7c9a46090390368a4d7bcb76bb87a4a36c421e4c0792cdb53486ffac7a/matplotlib_inline-0.2.2.tar.gz", hash = "sha256:72f3fe8fce36b70d4a5b612f899090cd0401deddc4ea90e1572b9f4bfb058c79", size = 8150, upload-time = "2026-05-08T17:33:33.49Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/41/09/5b161152e2d90f7b87f781c2e1267494aef9c32498df793f73ad0a0a494a/matplotlib_inline-0.2.2-py3-none-any.whl", hash = "sha256:3c821cf1c209f59fb2d2d64abbf5b23b67bcb2210d663f9918dd851c6da1fcf6", size = 9534, upload-time = "2026-05-08T17:33:32.055Z" },
+]
+
 [[package]]
 name = "mdurl"
 version = "0.1.2"
@@ -924,7 +1007,7 @@ name = "nvidia-cudnn-cu12"
 version = "9.10.2.21"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "nvidia-cublas-cu12", marker = "sys_platform == 'linux'" },
+    { name = "nvidia-cublas-cu12" },
 ]
 wheels = [
     { url = "https://files.pythonhosted.org/packages/ba/51/e123d997aa098c61d029f76663dedbfb9bc8dcf8c60cbd6adbe42f76d049/nvidia_cudnn_cu12-9.10.2.21-py3-none-manylinux_2_27_x86_64.whl", hash = "sha256:949452be657fa16687d0930933f032835951ef0892b37d2d53824d1a84dc97a8", size = 706758467, upload-time = "2025-06-06T21:54:08.597Z" },
@@ -935,7 +1018,7 @@ name = "nvidia-cufft-cu12"
 version = "11.3.3.83"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "nvidia-nvjitlink-cu12", marker = "sys_platform == 'linux'" },
+    { name = "nvidia-nvjitlink-cu12" },
 ]
 wheels = [
     { url = "https://files.pythonhosted.org/packages/1f/13/ee4e00f30e676b66ae65b4f08cb5bcbb8392c03f54f2d5413ea99a5d1c80/nvidia_cufft_cu12-11.3.3.83-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:4d2dd21ec0b88cf61b62e6b43564355e5222e4a3fb394cac0db101f2dd0d4f74", size = 193118695, upload-time = "2025-03-07T01:45:27.821Z" },
@@ -962,9 +1045,9 @@ name = "nvidia-cusolver-cu12"
 version = "11.7.3.90"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "nvidia-cublas-cu12", marker = "sys_platform == 'linux'" },
-    { name = "nvidia-cusparse-cu12", marker = "sys_platform == 'linux'" },
-    { name = "nvidia-nvjitlink-cu12", marker = "sys_platform == 'linux'" },
+    { name = "nvidia-cublas-cu12" },
+    { name = "nvidia-cusparse-cu12" },
+    { name = "nvidia-nvjitlink-cu12" },
 ]
 wheels = [
     { url = "https://files.pythonhosted.org/packages/85/48/9a13d2975803e8cf2777d5ed57b87a0b6ca2cc795f9a4f59796a910bfb80/nvidia_cusolver_cu12-11.7.3.90-py3-none-manylinux_2_27_x86_64.whl", hash = "sha256:4376c11ad263152bd50ea295c05370360776f8c3427b30991df774f9fb26c450", size = 267506905, upload-time = "2025-03-07T01:47:16.273Z" },
@@ -975,7 +1058,7 @@ name = "nvidia-cusparse-cu12"
 version = "12.5.8.93"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "nvidia-nvjitlink-cu12", marker = "sys_platform == 'linux'" },
+    { name = "nvidia-nvjitlink-cu12" },
 ]
 wheels = [
     { url = "https://files.pythonhosted.org/packages/c2/f5/e1854cb2f2bcd4280c44736c93550cc300ff4b8c95ebe370d0aa7d2b473d/nvidia_cusparse_cu12-12.5.8.93-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:1ec05d76bbbd8b61b06a80e1eaf8cf4959c3d4ce8e711b65ebd0443bb0ebb13b", size = 288216466, upload-time = "2025-03-07T01:48:13.779Z" },
@@ -1088,6 +1171,27 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/cd/d7/612123674d7b17cf345aad0a10289b2a384bff404e0463a83c4a3a59d205/pandas-2.3.2-cp313-cp313t-musllinux_1_2_x86_64.whl", hash = "sha256:d2c3554bd31b731cd6490d94a28f3abb8dd770634a9e06eb6d2911b9827db370", size = 13186141, upload-time = "2025-08-21T10:28:05.377Z" },
 ]
 
+[[package]]
+name = "parso"
+version = "0.8.7"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/30/4b/90c937815137d43ce71ba043cd3566221e9df6b9c805f24b5d138c9d40a7/parso-0.8.7.tar.gz", hash = "sha256:eaaac4c9fdd5e9e8852dc778d2d7405897ec510f2a298071453e5e3a07914bb1", size = 401824, upload-time = "2026-05-01T23:13:02.138Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/99/5d/8268b644392ee874ee82a635cd0df1773de230bde356c38de28e298392cc/parso-0.8.7-py2.py3-none-any.whl", hash = "sha256:a8926eb2a1b915486941fdbd31e86a4baf88fe8c210f25f2f35ecec5b574ca1c", size = 107025, upload-time = "2026-05-01T23:12:58.867Z" },
+]
+
+[[package]]
+name = "pexpect"
+version = "4.9.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "ptyprocess" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/42/92/cc564bf6381ff43ce1f4d06852fc19a2f11d180f23dc32d9588bee2f149d/pexpect-4.9.0.tar.gz", hash = "sha256:ee7d41123f3c9911050ea2c2dac107568dc43b2d3b0c7557a33212c398ead30f", size = 166450, upload-time = "2023-11-25T09:07:26.339Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/9e/c3/059298687310d527a58bb01f3b1965787ee3b40dce76752eda8b44e9a2c5/pexpect-4.9.0-py2.py3-none-any.whl", hash = "sha256:7236d1e080e4936be2dc3e326cec0af72acf9212a7e1d060210e70a47e253523", size = 63772, upload-time = "2023-11-25T06:56:14.81Z" },
+]
+
 [[package]]
 name = "pillow"
 version = "11.3.0"
@@ -1175,6 +1279,18 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/74/c1/bb7e334135859c3a92ec399bc89293ea73f28e815e35b43929c8db6af030/primePy-1.3-py3-none-any.whl", hash = "sha256:5ed443718765be9bf7e2ff4c56cdff71b42140a15b39d054f9d99f0009e2317a", size = 4040, upload-time = "2018-05-29T17:18:17.53Z" },
 ]
 
+[[package]]
+name = "prompt-toolkit"
+version = "3.0.52"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "wcwidth" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/a1/96/06e01a7b38dce6fe1db213e061a4602dd6032a8a97ef6c1a862537732421/prompt_toolkit-3.0.52.tar.gz", hash = "sha256:28cde192929c8e7321de85de1ddbe736f1375148b02f2e17edd840042b1be855", size = 434198, upload-time = "2025-08-27T15:24:02.057Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/84/03/0d3ce49e2505ae70cf43bc5bb3033955d2fc9f932163e84dc0779cc47f48/prompt_toolkit-3.0.52-py3-none-any.whl", hash = "sha256:9aac639a3bbd33284347de5ad8d68ecc044b91a762dc39b7c21095fcd6a19955", size = 391431, upload-time = "2025-08-27T15:23:59.498Z" },
+]
+
 [[package]]
 name = "propcache"
 version = "0.3.2"
@@ -1230,6 +1346,52 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/97/b7/15cc7d93443d6c6a84626ae3258a91f4c6ac8c0edd5df35ea7658f71b79c/protobuf-6.32.1-py3-none-any.whl", hash = "sha256:2601b779fc7d32a866c6b4404f9d42a3f67c5b9f3f15b4db3cccabe06b95c346", size = 169289, upload-time = "2025-09-11T21:38:41.234Z" },
 ]
 
+[[package]]
+name = "psutil"
+version = "7.2.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/aa/c6/d1ddf4abb55e93cebc4f2ed8b5d6dbad109ecb8d63748dd2b20ab5e57ebe/psutil-7.2.2.tar.gz", hash = "sha256:0746f5f8d406af344fd547f1c8daa5f5c33dbc293bb8d6a16d80b4bb88f59372", size = 493740, upload-time = "2026-01-28T18:14:54.428Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/51/08/510cbdb69c25a96f4ae523f733cdc963ae654904e8db864c07585ef99875/psutil-7.2.2-cp313-cp313t-macosx_10_13_x86_64.whl", hash = "sha256:2edccc433cbfa046b980b0df0171cd25bcaeb3a68fe9022db0979e7aa74a826b", size = 130595, upload-time = "2026-01-28T18:14:57.293Z" },
+    { url = "https://files.pythonhosted.org/packages/d6/f5/97baea3fe7a5a9af7436301f85490905379b1c6f2dd51fe3ecf24b4c5fbf/psutil-7.2.2-cp313-cp313t-macosx_11_0_arm64.whl", hash = "sha256:e78c8603dcd9a04c7364f1a3e670cea95d51ee865e4efb3556a3a63adef958ea", size = 131082, upload-time = "2026-01-28T18:14:59.732Z" },
+    { url = "https://files.pythonhosted.org/packages/37/d6/246513fbf9fa174af531f28412297dd05241d97a75911ac8febefa1a53c6/psutil-7.2.2-cp313-cp313t-manylinux2010_x86_64.manylinux_2_12_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:1a571f2330c966c62aeda00dd24620425d4b0cc86881c89861fbc04549e5dc63", size = 181476, upload-time = "2026-01-28T18:15:01.884Z" },
+    { url = "https://files.pythonhosted.org/packages/b8/b5/9182c9af3836cca61696dabe4fd1304e17bc56cb62f17439e1154f225dd3/psutil-7.2.2-cp313-cp313t-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:917e891983ca3c1887b4ef36447b1e0873e70c933afc831c6b6da078ba474312", size = 184062, upload-time = "2026-01-28T18:15:04.436Z" },
+    { url = "https://files.pythonhosted.org/packages/16/ba/0756dca669f5a9300d0cbcbfae9a4c30e446dfc7440ffe43ded5724bfd93/psutil-7.2.2-cp313-cp313t-win_amd64.whl", hash = "sha256:ab486563df44c17f5173621c7b198955bd6b613fb87c71c161f827d3fb149a9b", size = 139893, upload-time = "2026-01-28T18:15:06.378Z" },
+    { url = "https://files.pythonhosted.org/packages/1c/61/8fa0e26f33623b49949346de05ec1ddaad02ed8ba64af45f40a147dbfa97/psutil-7.2.2-cp313-cp313t-win_arm64.whl", hash = "sha256:ae0aefdd8796a7737eccea863f80f81e468a1e4cf14d926bd9b6f5f2d5f90ca9", size = 135589, upload-time = "2026-01-28T18:15:08.03Z" },
+    { url = "https://files.pythonhosted.org/packages/81/69/ef179ab5ca24f32acc1dac0c247fd6a13b501fd5534dbae0e05a1c48b66d/psutil-7.2.2-cp314-cp314t-macosx_10_15_x86_64.whl", hash = "sha256:eed63d3b4d62449571547b60578c5b2c4bcccc5387148db46e0c2313dad0ee00", size = 130664, upload-time = "2026-01-28T18:15:09.469Z" },
+    { url = "https://files.pythonhosted.org/packages/7b/64/665248b557a236d3fa9efc378d60d95ef56dd0a490c2cd37dafc7660d4a9/psutil-7.2.2-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:7b6d09433a10592ce39b13d7be5a54fbac1d1228ed29abc880fb23df7cb694c9", size = 131087, upload-time = "2026-01-28T18:15:11.724Z" },
+    { url = "https://files.pythonhosted.org/packages/d5/2e/e6782744700d6759ebce3043dcfa661fb61e2fb752b91cdeae9af12c2178/psutil-7.2.2-cp314-cp314t-manylinux2010_x86_64.manylinux_2_12_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:1fa4ecf83bcdf6e6c8f4449aff98eefb5d0604bf88cb883d7da3d8d2d909546a", size = 182383, upload-time = "2026-01-28T18:15:13.445Z" },
+    { url = "https://files.pythonhosted.org/packages/57/49/0a41cefd10cb7505cdc04dab3eacf24c0c2cb158a998b8c7b1d27ee2c1f5/psutil-7.2.2-cp314-cp314t-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:e452c464a02e7dc7822a05d25db4cde564444a67e58539a00f929c51eddda0cf", size = 185210, upload-time = "2026-01-28T18:15:16.002Z" },
+    { url = "https://files.pythonhosted.org/packages/dd/2c/ff9bfb544f283ba5f83ba725a3c5fec6d6b10b8f27ac1dc641c473dc390d/psutil-7.2.2-cp314-cp314t-win_amd64.whl", hash = "sha256:c7663d4e37f13e884d13994247449e9f8f574bc4655d509c3b95e9ec9e2b9dc1", size = 141228, upload-time = "2026-01-28T18:15:18.385Z" },
+    { url = "https://files.pythonhosted.org/packages/f2/fc/f8d9c31db14fcec13748d373e668bc3bed94d9077dbc17fb0eebc073233c/psutil-7.2.2-cp314-cp314t-win_arm64.whl", hash = "sha256:11fe5a4f613759764e79c65cf11ebdf26e33d6dd34336f8a337aa2996d71c841", size = 136284, upload-time = "2026-01-28T18:15:19.912Z" },
+    { url = "https://files.pythonhosted.org/packages/e7/36/5ee6e05c9bd427237b11b3937ad82bb8ad2752d72c6969314590dd0c2f6e/psutil-7.2.2-cp36-abi3-macosx_10_9_x86_64.whl", hash = "sha256:ed0cace939114f62738d808fdcecd4c869222507e266e574799e9c0faa17d486", size = 129090, upload-time = "2026-01-28T18:15:22.168Z" },
+    { url = "https://files.pythonhosted.org/packages/80/c4/f5af4c1ca8c1eeb2e92ccca14ce8effdeec651d5ab6053c589b074eda6e1/psutil-7.2.2-cp36-abi3-macosx_11_0_arm64.whl", hash = "sha256:1a7b04c10f32cc88ab39cbf606e117fd74721c831c98a27dc04578deb0c16979", size = 129859, upload-time = "2026-01-28T18:15:23.795Z" },
+    { url = "https://files.pythonhosted.org/packages/b5/70/5d8df3b09e25bce090399cf48e452d25c935ab72dad19406c77f4e828045/psutil-7.2.2-cp36-abi3-manylinux2010_x86_64.manylinux_2_12_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:076a2d2f923fd4821644f5ba89f059523da90dc9014e85f8e45a5774ca5bc6f9", size = 155560, upload-time = "2026-01-28T18:15:25.976Z" },
+    { url = "https://files.pythonhosted.org/packages/63/65/37648c0c158dc222aba51c089eb3bdfa238e621674dc42d48706e639204f/psutil-7.2.2-cp36-abi3-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:b0726cecd84f9474419d67252add4ac0cd9811b04d61123054b9fb6f57df6e9e", size = 156997, upload-time = "2026-01-28T18:15:27.794Z" },
+    { url = "https://files.pythonhosted.org/packages/8e/13/125093eadae863ce03c6ffdbae9929430d116a246ef69866dad94da3bfbc/psutil-7.2.2-cp36-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:fd04ef36b4a6d599bbdb225dd1d3f51e00105f6d48a28f006da7f9822f2606d8", size = 148972, upload-time = "2026-01-28T18:15:29.342Z" },
+    { url = "https://files.pythonhosted.org/packages/04/78/0acd37ca84ce3ddffaa92ef0f571e073faa6d8ff1f0559ab1272188ea2be/psutil-7.2.2-cp36-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:b58fabe35e80b264a4e3bb23e6b96f9e45a3df7fb7eed419ac0e5947c61e47cc", size = 148266, upload-time = "2026-01-28T18:15:31.597Z" },
+    { url = "https://files.pythonhosted.org/packages/b4/90/e2159492b5426be0c1fef7acba807a03511f97c5f86b3caeda6ad92351a7/psutil-7.2.2-cp37-abi3-win_amd64.whl", hash = "sha256:eb7e81434c8d223ec4a219b5fc1c47d0417b12be7ea866e24fb5ad6e84b3d988", size = 137737, upload-time = "2026-01-28T18:15:33.849Z" },
+    { url = "https://files.pythonhosted.org/packages/8c/c7/7bb2e321574b10df20cbde462a94e2b71d05f9bbda251ef27d104668306a/psutil-7.2.2-cp37-abi3-win_arm64.whl", hash = "sha256:8c233660f575a5a89e6d4cb65d9f938126312bca76d8fe087b947b3a1aaac9ee", size = 134617, upload-time = "2026-01-28T18:15:36.514Z" },
+]
+
+[[package]]
+name = "ptyprocess"
+version = "0.7.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/20/e5/16ff212c1e452235a90aeb09066144d0c5a6a8c0834397e03f5224495c4e/ptyprocess-0.7.0.tar.gz", hash = "sha256:5c5d0a3b48ceee0b48485e0c26037c0acd7d29765ca3fbb5cb3831d347423220", size = 70762, upload-time = "2020-12-28T15:15:30.155Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/22/a6/858897256d0deac81a172289110f31629fc4cee19b6f01283303e18c8db3/ptyprocess-0.7.0-py2.py3-none-any.whl", hash = "sha256:4b41f3967fce3af57cc7e94b888626c18bf37a083e3651ca8feeb66d492fef35", size = 13993, upload-time = "2020-12-28T15:15:28.35Z" },
+]
+
+[[package]]
+name = "pure-eval"
+version = "0.2.3"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/cd/05/0a34433a064256a578f1783a10da6df098ceaa4a57bbeaa96a6c0352786b/pure_eval-0.2.3.tar.gz", hash = "sha256:5f4e983f40564c576c7c8635ae88db5956bb2229d7e9237d03b3c0b0190eaf42", size = 19752, upload-time = "2024-07-21T12:58:21.801Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/8e/37/efad0257dc6e593a18957422533ff0f87ede7c9c6ea010a2177d738fb82f/pure_eval-0.2.3-py3-none-any.whl", hash = "sha256:1db8e35b67b3d218d818ae653e27f06c3aa420901fa7b081ca98cbedc874e0d0", size = 11842, upload-time = "2024-07-21T12:58:20.04Z" },
+]
+
 [[package]]
 name = "pyannote-audio"
 version = "3.4.0"
@@ -1700,7 +1862,7 @@ wheels = [
 [[package]]
 name = "segma"
 version = "0.1.0"
-source = { git = "https://github.com/MarvinLvn/segma?rev=dad876e2268799ff12671a865167126eef66c34c#dad876e2268799ff12671a865167126eef66c34c" }
+source = { git = "https://github.com/MarvinLvn/segma?rev=e6cebc5b232ac1a30554caa7ec35dd22057d07f0#e6cebc5b232ac1a30554caa7ec35dd22057d07f0" }
 dependencies = [
     { name = "interlap" },
     { name = "lightning" },
@@ -1885,6 +2047,20 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/b8/d9/13bdde6521f322861fab67473cec4b1cc8999f3871953531cf61945fad92/sqlalchemy-2.0.43-py3-none-any.whl", hash = "sha256:1681c21dd2ccee222c2fe0bef671d1aef7c504087c9c4e800371cfcc8ac966fc", size = 1924759, upload-time = "2025-08-11T15:39:53.024Z" },
 ]
 
+[[package]]
+name = "stack-data"
+version = "0.6.3"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "asttokens" },
+    { name = "executing" },
+    { name = "pure-eval" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/28/e3/55dcc2cfbc3ca9c29519eb6884dd1415ecb53b0e934862d3559ddcb7e20b/stack_data-0.6.3.tar.gz", hash = "sha256:836a778de4fec4dcd1dcd89ed8abff8a221f58308462e1c4aa2a3cf30148f0b9", size = 44707, upload-time = "2023-09-30T13:58:05.479Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/f1/7b/ce1eafaf1a76852e2ec9b22edecf1daa58175c090266e9f6c64afcd81d91/stack_data-0.6.3-py3-none-any.whl", hash = "sha256:d5558e0c25a4cb0853cddad3d77da9891a08cb85dd9f9f91b9f8cd66e511e695", size = 24521, upload-time = "2023-09-30T13:58:03.53Z" },
+]
+
 [[package]]
 name = "sympy"
 version = "1.14.0"
@@ -2143,6 +2319,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/d0/30/dc54f88dd4a2b5dc8a0279bdd7270e735851848b762aeb1c1184ed1f6b14/tqdm-4.67.1-py3-none-any.whl", hash = "sha256:26445eca388f82e72884e0d580d5464cd801a3ea01e63e5601bdff9ba6a48de2", size = 78540, upload-time = "2024-11-24T20:12:19.698Z" },
 ]
 
+[[package]]
+name = "traitlets"
+version = "5.15.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/57/a9/a2584b8313b89f94869ddb3c4074617a691de1812a614d2d50e32ca5a7a6/traitlets-5.15.1.tar.gz", hash = "sha256:7b1c07854fe25acb39e009bae49f11b79ff6cbb2f27999104e9110e7a6b53722", size = 163344, upload-time = "2026-06-03T12:26:06.181Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/96/8d/1080ee4c231f361b6ce4470d556c8c435b67c7e0753aaa641497ee92f88b/traitlets-5.15.1-py3-none-any.whl", hash = "sha256:770a53705f84b81ac107e83a1b3328ff2dae16094d8fc3cfc004e4b22dfd8e92", size = 85858, upload-time = "2026-06-03T12:26:04.395Z" },
+]
+
 [[package]]
 name = "transformers"
 version = "4.56.2"
@@ -2231,27 +2416,33 @@ wheels = [
 
 [[package]]
 name = "vtc"
-version = "2.1.0"
-source = { virtual = "." }
+version = "2.2.0"
+source = { editable = "." }
 dependencies = [
+    { name = "huggingface-hub" },
     { name = "pyannote-audio" },
     { name = "segma" },
     { name = "tqdm" },
 ]
 
 [package.dev-dependencies]
+dev = [
+    { name = "ipython" },
+]
 doc = [
     { name = "zensical" },
 ]
 
 [package.metadata]
 requires-dist = [
+    { name = "huggingface-hub", specifier = ">=0.35.0" },
     { name = "pyannote-audio", specifier = ">=3.4.0" },
-    { name = "segma", git = "https://github.com/MarvinLvn/segma?rev=dad876e2268799ff12671a865167126eef66c34c" },
+    { name = "segma", git = "https://github.com/MarvinLvn/segma?rev=e6cebc5b232ac1a30554caa7ec35dd22057d07f0" },
     { name = "tqdm", specifier = ">=4.67.1" },
 ]
 
 [package.metadata.requires-dev]
+dev = [{ name = "ipython", specifier = ">=9.15.0" }]
 doc = [{ name = "zensical", specifier = ">=0.0.37" }]
 
 [[package]]
@@ -2282,6 +2473,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/3c/c9/d9f0c7b8a743af589e694ce8fec8e6cffa46873179912d4ed4f992d08381/wandb-0.22.0-py3-none-win_amd64.whl", hash = "sha256:53ba0fa048b766c1aa44592f1e530fb7eead7749089a66c3892b35f153a8d8bd", size = 18742399, upload-time = "2025-09-18T19:13:19.26Z" },
 ]
 
+[[package]]
+name = "wcwidth"
+version = "0.8.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/34/74/c6428f875774288bec1396f5bfcbc2d925700a4dad61727fd5f2b12f249d/wcwidth-0.8.2.tar.gz", hash = "sha256:91fbef97204b96a3d4d421609b80340b760cf33e26da123ff243d76b1fda8dda", size = 1466253, upload-time = "2026-06-29T18:11:11.601Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/96/42/3e5985a0a7e57de470b320c6d6a1a67c844f6737a587f3d44dd13d1819e7/wcwidth-0.8.2-py3-none-any.whl", hash = "sha256:d63947694a0539a1d51e01eda7caf800c291020e6cdd7e28ad7b14dd33ad4f85", size = 323166, upload-time = "2026-06-29T18:11:09.888Z" },
+]
+
 [[package]]
 name = "yarl"
 version = "1.20.1"

From 714d079982a775a2c11814bd1f642419c7d29d82 Mon Sep 17 00:00:00 2001
From: Hamilakis Nicolas <nicolas.hamilakis@ens.psl.eu>
Date: Thu, 23 Jul 2026 15:04:49 +0200
Subject: [PATCH 2/3] Updated readme

---
 README.md | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/README.md b/README.md
index d582dd3..e989efd 100644
--- a/README.md
+++ b/README.md
@@ -30,8 +30,7 @@ As system dependencies, ensure that [uv](https://docs.astral.sh/uv/), [ffmpeg](h
 You can now clone the repo with:
 
 ```bash
-git lfs install
-git clone --recurse-submodules https://github.com/LAAC-LSCP/VTC.git
+git clone https://github.com/LAAC-LSCP/VTC.git
 cd VTC
 ```
 
@@ -48,7 +47,7 @@ If not, you can use the `scripts/convert.py` file to convert your audios to 16 0
 
 
 ```bash
-uv run scripts/infer.py \
+uv run vtc-infer \
     --wavs audios \        # path to the folder containing the audio files
     --output predictions \ # output folder
     --device cpu           # device to run the model on: ('cpu', 'cuda' or 'gpu', 'mps')
@@ -66,10 +65,10 @@ The model outputs are saved to `<output_folder>/` with the following structure:
 ```
 
 #### Helper script
-An example of a bash script is given to perform inference in `scripts/run.sh`. Simply set the correct variables in the script and run it:
+An example of a bash script is given to perform inference in `run.sh`. Simply set the correct variables in the script and run it:
 
 ```bash
-sh scripts/run.sh
+sh run.sh
 ````
 
 ## 3. Model Performance

From e8e819f781e4a25b1f24859d9af5d73f9e5e8df5 Mon Sep 17 00:00:00 2001
From: Hamilakis Nicolas <nicolas.hamilakis@ens.psl.eu>
Date: Thu, 23 Jul 2026 15:07:32 +0200
Subject: [PATCH 3/3] fix version typo

---
 src/vtc_inference/cli.py        | 6 +++---
 src/vtc_inference/ressources.py | 1 +
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/src/vtc_inference/cli.py b/src/vtc_inference/cli.py
index 20a70c3..f0dd5e9 100644
--- a/src/vtc_inference/cli.py
+++ b/src/vtc_inference/cli.py
@@ -7,7 +7,7 @@ def vtc_infer_cli() -> None:
     parser.add_argument(
         "--config",
         type=str,
-        default="2.1",
+        default="2.2",
         help="Local config.toml file/directory, or a known model version key (e.g. '2.1', '2.0').",
     )
     parser.add_argument(
@@ -20,7 +20,7 @@ def vtc_infer_cli() -> None:
     )
     parser.add_argument(
         "--checkpoint",
-        default="2.1",
+        default="2.2",
         type=str,
         help="Local checkpoint file/directory, or a known model version key (e.g. '2.1', '2.0').",
     )
@@ -43,7 +43,7 @@ def vtc_infer_cli() -> None:
     parser.add_argument(
         "--thresholds-location",
         type=str,
-        default="2.1",
+        default="2.2",
         help=(
             "Local directory/file containing the thresholds TOML, or a known "
             "model version key. Defaults to wherever --checkpoint resolves to."
diff --git a/src/vtc_inference/ressources.py b/src/vtc_inference/ressources.py
index 0f3fa92..ce092bb 100644
--- a/src/vtc_inference/ressources.py
+++ b/src/vtc_inference/ressources.py
@@ -6,6 +6,7 @@
 VersionStr: t.TypeAlias = str
 REPO_ID = "coml/VTC-2"
 MODEL_TO_REVISION = {
+    "2.2": "91167f6",
     "2.1": "91167f6",
     "2.0": "91e67b5",
 }