Downloads · 30 days
0
ironprose-omni
jpoehnelt/ironprose-omni
ironprose-omni is a machine learning model from jpoehnelt. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Unified diagnostic classifier with 17 binary heads sharing a single google/bertuncasedL-4H-256A-4 encoder.
Access
Public
Updated May 1, 2026
Repo size
180 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pt44.8 MB · 49%
At a glance
- Access
- Public
- Created
- Mar 1, 2026
- Updated
- May 1, 2026
- SHA
- 94f5aac5
Model card
From the Hugging Face model README
Omni Multi-Task Model
Unified diagnostic classifier with 17 binary heads sharing a single
google/bert_uncased_L-4_H-256_A-4 encoder.
Tasks
| Index | Task | Description |
|---|---|---|
| 0 | show_tell | Binary classifier |
| 1 | on_the_nose | Binary classifier |
| 2 | info_dump | Binary classifier |
| 3 | purple_prose | Binary classifier |
| 4 | white_room | Binary classifier |
| 5 | trope_cliche | Binary classifier |
| 6 | as_you_know_bob | Binary classifier |
| 7 | tautological_tag | Binary classifier |
| 8 | floating_body_part | Binary classifier |
| 9 | spatial_grounding | Binary classifier |
| 10 | over_explained_emotion | Binary classifier |
| 11 | redundant_dialogue_tag | Binary classifier |
| 12 | psychic_distance | Binary classifier |
| 13 | bloat | Binary classifier |
| 14 | passive_protagonist | Binary classifier |
| 15 | emotional_whiplash | Binary classifier |
| 16 | anachronism | Binary classifier |
Metrics
- Val Accuracy: 0.9223
- Val F1 (macro): 0.9172
- Training samples: 19449
Output Shape
logits: (batch_size, 17, 2) — softmax over dim=2 gives per-task probabilities.
Usage
import onnxruntime as ort
import numpy as np
session = ort.InferenceSession("model.onnx")
logits = session.run(None, {"input_ids": ids, "attention_mask": mask})[0]
# logits shape: (1, 17, 2)
probs = np.exp(logits) / np.exp(logits).sum(axis=-1, keepdims=True)
# probs[:, task_idx, 1] = probability of "flagged" for each task
- Created
- Mar 1, 2026
- Updated
- May 1, 2026