Add files using upload-large-folder tool

.gitattributes

@@ -58,3 +58,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+metadata.tsv filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,51 @@
+# MoVE Dataset — output-200
+
+Bilingual (EN↔ZH) expressive speech-to-speech translation dataset.
+5 emotion categories: `angry`, `happy`, `sad`, `laugh`, `crying`.
+
+## Directory Structure
+
+```
+output-200/
+├── en/{emotion}/*.flac       # English TTS audio (FLAC, lossless)
+├── zh/{emotion}/*.flac       # Chinese TTS audio (FLAC, lossless)
+├── metadata/
+│   ├── metadata_shard_*.jsonl   # S2ST conversation format (per shard)
+│   └── entries_shard_*.jsonl    # Per-entry detail (text, ASR, duration, WER)
+├── metadata.tsv              # Generated by make_metadata_tsv.py
+├── make_metadata_tsv.py      # Step 1: shards → metadata.tsv
+└── make_kimi_train.py        # Step 2: metadata.tsv → Kimi training format
+```
+
+## metadata.tsv
+
+Columns: `zh_path`, `en_path`, `zh_text`, `en_text`, `category`
+
+All paths are relative to this directory and use `.flac` extension.
+
+```bash
+python make_metadata_tsv.py
+```
+
+## Kimi-Audio Training Format
+
+> ⚠️ **WARNING**: `make_kimi_train.py` converts all `.flac` files to `.wav` **in-place**
+> and deletes the original `.flac` files. WAV files are approximately **2–3× larger**
+> than FLAC. Run this only when you intend to use the data for local Kimi-Audio training
+> and no longer need the FLAC files.
+
+Produces `metadata_kimi_train.jsonl` in Kimi-Audio conversation format:
+
+```json
+{"task_type": "s-s", "conversation": [
+  {"role": "user", "message_type": "text", "content": "Translate the given English speech into Chinese while preserving its expressiveness."},
+  {"role": "user", "message_type": "audio", "content": "en/angry/angry_000001_en.wav"},
+  {"role": "assistant", "message_type": "audio-text", "content": ["zh/angry/angry_000001_zh.wav", "..."]}
+]}
+```
+
+Each pair generates two entries (EN→ZH and ZH→EN).
+
+```bash
+python make_kimi_train.py
+```

make_kimi_train.py

@@ -0,0 +1,107 @@
+#!/usr/bin/env python3
+"""
+Step 2: metadata.tsv → FLAC-to-WAV conversion + metadata_kimi_train.jsonl
+
+WARNING: This script converts all .flac files to .wav IN-PLACE and deletes the .flac.
+         WAV files are approximately 2-3x larger than FLAC.
+         Purpose: produce training data in Kimi-Audio conversation format.
+
+Output: metadata_kimi_train.jsonl  (same directory as this script)
+        Each pair produces two entries:
+          1. EN→ZH: user gives EN audio, assistant responds with ZH audio + ZH text
+          2. ZH→EN: user gives ZH audio, assistant responds with EN audio + EN text
+"""
+
+import csv
+import json
+import os
+import soundfile as sf
+from tqdm import tqdm
+
+BASE_DIR     = os.path.dirname(__file__)
+INPUT_TSV    = os.path.join(BASE_DIR, "metadata.tsv")
+OUTPUT_JSONL = os.path.join(BASE_DIR, "metadata_kimi_train.jsonl")
+
+
+def flac_to_wav(flac_path: str) -> str:
+    """Convert a .flac file to .wav in-place (deletes .flac). Returns .wav path."""
+    wav_path = flac_path[:-5] + ".wav"
+    data, sr = sf.read(flac_path)
+    sf.write(wav_path, data, sr)
+    os.remove(flac_path)
+    return wav_path
+
+
+def flac_rel_to_wav_rel(p: str) -> str:
+    return p[:-5] + ".wav" if p.endswith(".flac") else p
+
+
+def main():
+    with open(INPUT_TSV, "r", encoding="utf-8") as f:
+        rows = list(csv.DictReader(f, delimiter="\t"))
+    print(f"Loaded {len(rows)} pairs from {INPUT_TSV}")
+    print("Converting FLAC → WAV (in-place) and writing metadata_kimi_train.jsonl ...")
+
+    ok = skip = fail = 0
+    out_f = open(OUTPUT_JSONL, "w", encoding="utf-8")
+
+    for row in tqdm(rows, unit="pairs"):
+        zh_flac = os.path.join(BASE_DIR, row["zh_path"])
+        en_flac = os.path.join(BASE_DIR, row["en_path"])
+        zh_text = row["zh_text"]
+        en_text = row["en_text"]
+
+        # Convert FLAC → WAV for both files
+        converted = {}
+        for key, flac_abs in [("zh", zh_flac), ("en", en_flac)]:
+            wav_abs = flac_abs[:-5] + ".wav"
+            if os.path.exists(wav_abs):
+                converted[key] = wav_abs
+            elif os.path.exists(flac_abs):
+                try:
+                    converted[key] = flac_to_wav(flac_abs)
+                except Exception as e:
+                    print(f"\n  FAILED converting {flac_abs}: {e}")
+                    fail += 1
+                    break
+            else:
+                print(f"\n  MISSING: {flac_abs}")
+                fail += 1
+                break
+        else:
+            # Both converted successfully — write two conversation entries
+            en_rel = flac_rel_to_wav_rel(row["en_path"])
+            zh_rel = flac_rel_to_wav_rel(row["zh_path"])
+
+            # EN → ZH
+            out_f.write(json.dumps({
+                "task_type": "s-s",
+                "conversation": [
+                    {"role": "user",      "message_type": "text",       "content": "Translate the given English speech into Chinese while preserving its expressiveness."},
+                    {"role": "user",      "message_type": "audio",      "content": en_rel},
+                    {"role": "assistant", "message_type": "audio-text", "content": [zh_rel, zh_text]},
+                ]
+            }, ensure_ascii=False) + "\n")
+
+            # ZH → EN
+            out_f.write(json.dumps({
+                "task_type": "s-s",
+                "conversation": [
+                    {"role": "user",      "message_type": "text",       "content": "Translate the given Chinese speech into English while preserving its expressiveness."},
+                    {"role": "user",      "message_type": "audio",      "content": zh_rel},
+                    {"role": "assistant", "message_type": "audio-text", "content": [en_rel, en_text]},
+                ]
+            }, ensure_ascii=False) + "\n")
+
+            ok += 1
+            continue
+
+        skip += 1
+
+    out_f.close()
+    print(f"\nDone: {ok} pairs converted, {skip} skipped/missing, {fail} errors")
+    print(f"Output: {OUTPUT_JSONL}  ({ok * 2} total conversation entries)")
+
+
+if __name__ == "__main__":
+    main()

metadata.tsv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63f4b45f6ba7ad2bc0384a01865dc97451835c9312feb2d430713f9e30ab0adc
+size 183396500