| { | |
| "train/loss": 0.3984375, | |
| "train/contrastive": 0.310546875, | |
| "train/recons_loss": 0.68359375, | |
| "train/balance_loss": 2.015625, | |
| "train/balance_loss_contrastive": 1.0, | |
| "train/balance_loss_recons": 1.015625, | |
| "train/contrastive_std": 0.015625, | |
| "train/recons_std": 0.0888671875, | |
| "train/contrastive_min": 0.27734375, | |
| "train/contrastive_max": 0.3359375, | |
| "train/recons_min": 0.58203125, | |
| "train/recons_max": 0.84375, | |
| "train/Qwen3_0.6B_layer_2": 0.58203125, | |
| "train/Qwen3_0.6B_layer_4": 0.69140625, | |
| "train/Qwen3_1.7B_layer_2": 0.65234375, | |
| "train/Qwen3_1.7B_layer_4": 0.84375, | |
| "train/Qwen3_4B_layer_2": 0.63671875, | |
| "train/Qwen3_4B_layer_4": 0.6953125, | |
| "train/contrastives": null, | |
| "train/epoch": 1, | |
| "train/n_tokens": 1000448, | |
| "train/step": 977 | |
| } |