Metadata-Version: 2.4
Name: linear-probe
Version: 0.1.1
Summary: Add your description here
Requires-Python: >=3.12
Description-Content-Type: text/markdown
Requires-Dist: datasets>=5.0.1
Requires-Dist: pysimdjson>=7.0.2
Requires-Dist: torch>=2.13.0

# linear-probe
```bash
ruff check --select I --fix . && ruff format .
```

Expected input is `embeddings` and `labels` columns. `labels` are raw values are encoding automatically.

| task | labels | example |
|---|---|---|
| `binary` | scalar | `1`, `"spam"` |
| `multiclass` | scalar | `2`, `"sports"` |
| `multilabel` | list of labels present | `["politics", "eu"]`, `[0, 2]` |

```py
# binary
{"embeddings": [0.12, -0.53, 0.88, ...], "labels": "spam"}

# multiclass
{"embeddings": [0.12, -0.53, 0.88, ...], "labels": "sports"}

# multilabel
{"embeddings": [0.12, -0.53, 0.88, ...], "labels": ["politics", "eu"]}
```
