Claude
ab100f85a4
Backend only (window follows). Switched on in Settings > Modules. H3 repairs proven on a tiny H3 on CPU, UNTESTED on real weights. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MpnF7GwNGMbngU11wjym7P
75 lines
2.6 KiB
Python
75 lines
2.6 KiB
Python
"""Captions as tags: the small text helpers the tag editor and the trainer share. No ComfyUI, no I/O."""
|
|
|
|
from __future__ import annotations
|
|
|
|
MAX_CAPTION = 4000
|
|
|
|
|
|
def split_tags(text) -> list[str]:
|
|
""""a, b ,, a" -> ["a", "b"]: trimmed, no empties, first occurrence kept."""
|
|
seen, out = set(), []
|
|
for part in str(text or "").replace("\n", ",").split(","):
|
|
tag = " ".join(part.split())
|
|
if tag and tag.lower() not in seen:
|
|
seen.add(tag.lower())
|
|
out.append(tag)
|
|
return out
|
|
|
|
|
|
def join_tags(tags) -> str:
|
|
return ", ".join(split_tags(", ".join(str(t) for t in tags)))[:MAX_CAPTION]
|
|
|
|
|
|
def with_trigger(caption, trigger) -> str:
|
|
"""The trigger word first, once. An empty trigger changes nothing."""
|
|
trigger = " ".join(str(trigger or "").split())
|
|
if not trigger:
|
|
return str(caption or "")
|
|
rest = [t for t in split_tags(caption) if t.lower() != trigger.lower()]
|
|
return ", ".join([trigger, *rest])
|
|
|
|
|
|
def frequency(captions) -> list[tuple[str, int]]:
|
|
"""Every tag with the number of captions it is in, most common first, then A to Z."""
|
|
counts: dict[str, list] = {}
|
|
for caption in captions:
|
|
for tag in split_tags(caption):
|
|
counts.setdefault(tag.lower(), [tag, 0])[1] += 1
|
|
return sorted(((tag, n) for tag, n in counts.values()), key=lambda x: (-x[1], x[0].lower()))
|
|
|
|
|
|
def bulk(items, op, tag, new=None, ids=None) -> int:
|
|
"""Change captions in place; returns how many items changed.
|
|
|
|
`op` is add (at the end), prepend, remove or replace (tag -> new). `ids` limits it to those items.
|
|
"""
|
|
changed = 0
|
|
tag = " ".join(str(tag or "").split())
|
|
if not tag or op not in ("add", "prepend", "remove", "replace"):
|
|
return 0
|
|
if op == "replace":
|
|
new = " ".join(str(new or "").split())
|
|
if not new:
|
|
op = "remove"
|
|
for item in items:
|
|
if ids is not None and item.get("id") not in ids:
|
|
continue
|
|
tags = split_tags(item.get("caption"))
|
|
low = [t.lower() for t in tags]
|
|
if op == "add" and tag.lower() not in low:
|
|
tags.append(tag)
|
|
elif op == "prepend":
|
|
tags = [tag] + [t for t in tags if t.lower() != tag.lower()]
|
|
elif op == "remove" and tag.lower() in low:
|
|
tags = [t for t in tags if t.lower() != tag.lower()]
|
|
elif op == "replace" and tag.lower() in low:
|
|
tags = [new if t.lower() == tag.lower() else t for t in tags]
|
|
else:
|
|
continue
|
|
text = join_tags(tags)
|
|
if text != item.get("caption"):
|
|
item["caption"] = text
|
|
changed += 1
|
|
return changed
|
|
|