File size: 6,778 Bytes
4af42ec
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
# MultiShotMaster: validates outputs, checks prompts/tokens, measures the visible cuts, writes per-episode metadata,
# timing table, and builds the release dir (~/msm_run/release) in the same layout as ERCBench-ShotStream-400.
import csv, datetime as dt, glob, hashlib, json, os, re, shutil, subprocess, sys
import av, numpy as np
W = os.path.expanduser("~/msm_run")
idx = json.load(open(os.path.expanduser("~/shotstream_run/inputs/prompts_index.json")))
DXB = dt.timezone(dt.timedelta(hours=4))
F = 75  # frames per shot (configured shot_groups [75k, 75(k+1)])

# ---- timing + token records, per worker log (later runs of an episode override earlier ones)
timing, tok_of, tok_problems = {}, {}, []
for logp in sorted(glob.glob(f"{W}/logs/worker_gpu*.log")):
    gpu = int(re.search(r"gpu(\d+)", logp).group(1))
    tokp = f"{W}/logs/tokens_gpu{gpu}.jsonl"
    tok = [json.loads(l) for l in open(tokp, encoding="utf-8")] if os.path.exists(tokp) else []
    pos, st = 0, {}
    for line in open(logp, errors="replace"):
        m = re.search(r"(\d{4}-\d\d-\d\dT[\d:.]+Z) \[EPISODE\] (start|saved) (\S+)", line)
        if not m: continue
        ts = dt.datetime.strptime(m.group(1), "%Y-%m-%dT%H:%M:%S.%fZ").replace(tzinfo=dt.timezone.utc)
        eid = m.group(3); n = idx[eid]["shot_count"]
        if m.group(2) == "start":
            st[eid] = ts
            recs = tok[pos:pos + 2 * n]; pos += 2 * n   # n positive prompts, then n negative prompts
            if len(recs) < 2 * n: continue              # interrupted before encoding finished
            probs = []
            for k, s in enumerate(idx[eid]["shots"]):
                r = recs[k]
                if r["raw"] != "Story:  Now: " + s["prompt"]: probs.append((eid, k + 1, "text mismatch / misaligned"))
                if hashlib.sha256(r["raw"][len("Story:  Now: "):].encode()).hexdigest() != s["sha256"]: probs.append((eid, k + 1, "sha mismatch"))
                if r["n_tokens"] > r["seq_len"]: probs.append((eid, k + 1, f"{r['n_tokens']} tokens > {r['seq_len']}"))
            tok_of[eid] = ([recs[k]["n_tokens"] for k in range(n)], recs[n]["n_tokens"], probs)
        else:
            mem = re.search(r"peak_mem_gb=([\d.]+)", line)
            timing[eid] = {"seconds": round((ts - st[eid]).total_seconds(), 2), "gpu": gpu, "start": st[eid], "end": ts,
                           "peak_mem_gb": float(mem.group(1)) if mem else None, "log": os.path.basename(logp)}

def probe(p):
    out = subprocess.run(["ffprobe", "-v", "error", "-count_frames", "-select_streams", "v:0", "-show_entries",
        "stream=width,height,r_frame_rate,nb_read_frames", "-of", "json", p], capture_output=True, text=True)
    s = json.loads(out.stdout)["streams"][0]
    dec = subprocess.run(["ffmpeg", "-v", "error", "-xerror", "-i", p, "-f", "null", "-"], capture_output=True, text=True)
    return s, dec.returncode == 0 and not dec.stderr.strip()

def measure_cuts(p, n):
    """Per configured cut 75k: frame index with the strongest consecutive-frame change in [75k-2, 75k+8]."""
    fr = [f.to_ndarray(format="gray").astype(np.float32) for f in av.open(p).decode(video=0)]
    d = np.array([np.abs(fr[i] - fr[i - 1]).mean() for i in range(1, len(fr))])   # d[i-1]: change into frame i
    med = float(np.median(d)); out = []
    for k in range(1, n):
        c = F * k; lo, hi = c - 2, c + 8
        i = lo + int(np.argmax(d[lo - 1:hi]))
        out.append({"configured": c, "measured": i, "offset": i - c, "strength": round(float(d[i - 1]) / med, 1)})
    return out

rel = f"{W}/release"; problems = []; rows = []
for eid, info in sorted(idx.items(), key=lambda x: x[1]["priority"]):
    n = info["shot_count"]; src = f"{W}/outputs/{eid}.mp4"; nf = F * n
    if not os.path.exists(src): problems.append((eid, "missing")); continue
    s, ok = probe(src)
    if not (s["width"] == 832 and s["height"] == 480 and s["r_frame_rate"] == "15/1" and int(s["nb_read_frames"]) == nf and ok):
        problems.append((eid, s, ok)); continue
    if eid not in timing: problems.append((eid, "no timing")); continue
    if eid not in tok_of: problems.append((eid, "no token record")); continue
    tok_problems += tok_of[eid][2]
    cuts = measure_cuts(src, n)
    d = f"{rel}/episodes/{eid}"; os.makedirs(d, exist_ok=True)
    shutil.copy2(src, f"{d}/episode.mp4")
    vsha = hashlib.sha256(open(src, "rb").read()).hexdigest()
    t = timing[eid]
    md = {"episode_id": eid, "priority": info["priority"], "shot_count": n,
          "shot_frame_ranges": [[F * k, F * (k + 1)] for k in range(n)],
          "cut_frames": [F * k for k in range(1, n)],
          "measured_cut_frames": [c["measured"] for c in cuts],
          "measured_cuts": cuts,
          "fps": 15, "resolution": {"width": 832, "height": 480}, "num_frames": nf,
          "seed": 42, "seed_scope": "set before every episode (code default)", "global_caption": "",
          "shots": [{"shot": p["shot"], "prompt": p["prompt"], "prompt_sha256": p["sha256"],
                     "text_encoder_input": "Story:  Now: " + p["prompt"], "umt5_tokens": tok_of[eid][0][i]}
                    for i, p in enumerate(info["shots"])],
          "negative_prompt_umt5_tokens": tok_of[eid][1],
          "video_sha256": vsha, "generation_seconds": t["seconds"], "peak_mem_gb": t["peak_mem_gb"], "gpu_index": t["gpu"],
          "generated_at_dubai": t["end"].astimezone(DXB).isoformat(), "source_log": t["log"]}
    json.dump(md, open(f"{d}/metadata.json", "w"), ensure_ascii=False, indent=2)
    rows.append([eid, info["priority"], n, t["gpu"], t["seconds"], round(t["seconds"] / n, 2),
                 t["start"].astimezone(DXB).isoformat(), t["end"].astimezone(DXB).isoformat(),
                 " ".join(str(c["offset"]) for c in cuts), vsha])

print("ok episodes:", len(rows), "problems:", len(problems), "| token problems:", len(tok_problems))
for p in problems[:20]: print("PROBLEM", p)
for p in tok_problems[:20]: print("TOKEN PROBLEM", p)
offs = [int(o) for r in rows for o in r[8].split()]
if offs: print("measured cut offsets (frames after configured cut):", {o: offs.count(o) for o in sorted(set(offs))})
allt = [x for e in tok_of.values() for x in e[0]]
if allt: print("tokens per shot: max", max(allt), "min", min(allt))
os.makedirs(f"{rel}/runs", exist_ok=True)
with open(f"{rel}/runs/timing.csv", "w", newline="") as f:
    w = csv.writer(f); w.writerow(["episode_id", "priority", "shot_count", "gpu", "seconds", "seconds_per_shot", "start_dubai", "end_dubai", "cut_offsets", "video_sha256"])
    w.writerows(rows)
if rows:
    tot_s = sum(r[4] for r in rows); tot_shots = sum(r[2] for r in rows)
    print("sum gen seconds", round(tot_s, 1), "shots", tot_shots, "avg s/shot", round(tot_s / tot_shots, 3))
sys.exit(1 if problems or tok_problems else 0)