Correct the model card (pooling is at the LABELS suffix's last token); add code/classifier from project commit 4ced7554f
3d7c238 verified | """Own market-psychology classifier: Qwen3.5-9B with LoRA and one classification head per field. | |
| Spec: docs/market-psychology-own-classifier-training.md. Adapted from the read-only reference | |
| ~/yt_live_analysis/training (heads arm), with the fixes the spec lists. The pure-Python modules | |
| (taxonomy, calibrate, metrics, formats, data, ledger, robustness) run anywhere; model, train, predict and | |
| check_env need torch and run on the pod. | |
| """ | |