Token Classification
Transformers
ONNX
Safetensors
English
Japanese
Chinese
bert
anime
filename-parsing
Eval Results (legacy)
Instructions to use ModerRAS/AniFileBERT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use ModerRAS/AniFileBERT with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("token-classification", model="ModerRAS/AniFileBERT")# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForTokenClassification tokenizer = AutoTokenizer.from_pretrained("ModerRAS/AniFileBERT") model = AutoModelForTokenClassification.from_pretrained("ModerRAS/AniFileBERT", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Tighten LM Studio title-boundary prompt
Browse files
tools/rust_dmhy_template_apply/src/main.rs
CHANGED
|
@@ -1338,10 +1338,11 @@ fn title_boundary_review_prompt(recipe: &Value, decision: &Value) -> String {
|
|
| 1338 |
let payload = json!({
|
| 1339 |
"task": "Independently classify whether two adjacent TITLE slots in an anime filename template should be one logical title span or kept separate.",
|
| 1340 |
"rules": [
|
| 1341 |
-
"Choose merge only when the two slots are contiguous
|
| 1342 |
-
"
|
| 1343 |
-
"Choose
|
| 1344 |
-
"
|
|
|
|
| 1345 |
],
|
| 1346 |
"allowed_decisions": {
|
| 1347 |
"merge": "same title continuation; use one logical title span",
|
|
|
|
| 1338 |
let payload = json!({
|
| 1339 |
"task": "Independently classify whether two adjacent TITLE slots in an anime filename template should be one logical title span or kept separate.",
|
| 1340 |
"rules": [
|
| 1341 |
+
"Choose merge only when concrete examples or left/right top literals show the two slots are contiguous words of the same title/subtitle phrase.",
|
| 1342 |
+
"Do not choose merge merely because the slots are adjacent or separated only by SEP/O; adjacency alone is not evidence.",
|
| 1343 |
+
"Choose keep when slots are duplicate titles, path/leaf repeats, multilingual aliases, category tags plus title, release group plus title, or distinct title candidates.",
|
| 1344 |
+
"Choose skip when examples/top literals are mixed, insufficient, contradictory, or you cannot cite a concrete left/right title-word pairing.",
|
| 1345 |
+
"Use literal_similarity only as a clue, not as the answer. Do not infer from any heuristic decision."
|
| 1346 |
],
|
| 1347 |
"allowed_decisions": {
|
| 1348 |
"merge": "same title continuation; use one logical title span",
|