Update model card with v2 training details (600 iter, 2.3K examples, val loss 1.057)

LEK-GPT-OSS-20B v2: 600 iter LoRA on 2,299 examples (up from 160)

Rename LEM -> LEK (proof-of-concept designation)

Tags: lek->lem for trained models, standardise LEM tags

Standardise tags: lek->lem for trained models

LEM-GPT-OSS-20B: LoRA fine-tuned with LEK-1 (iter 100 checkpoint, val_loss 1.052)

initial commit