luca0621/mmpae-1p5B-e200-bs64

MMPAE 1.5B E200 BS64

Checkpoint and evaluation artifacts for mmpae-1p5B-e200-bs64.

Training

• Config: configs/Inverse_CwA_1p5B.yaml
• Parameters: about 1.496B
• Epochs: 200
• Batch size: 64
• Dataset path in training: /data/polyone_tokenized
• Tokenizer path in training: /data/polyBERT

Final Validation Metrics

{
  "actual_params": 1495526443,
  "best_checkpoint_path": "/data/runs/mmpae-1p5B-e200-bs64/Checkpoint_BEST.pt",
  "checkpoint_path": "/data/runs/mmpae-1p5B-e200-bs64/Polyone_AE_0200.pt",
  "epoch": 200,
  "epochs": 200,
  "eval": {
    "ce_loss": 0.1119885389772201,
    "contrast_loss": 2.187204460526684,
    "eos_loss": 0.014220016577973116,
    "eval_batches": 1954,
    "eval_seconds": 2795.629148006439,
    "infer_batches": 10,
    "infer_samples": 640,
    "infer_seconds": 61.439425230026245,
    "inv_r2": 0.7720341189253714,
    "inv_rmse": 0.4378301985802189,
    "mse_loss": 4.531615727267465,
    "prop_r2": 0.6728669822216033,
    "prop_rmse": 0.5227363497018814,
    "sim_score": 0.39310147781608873,
    "total_loss": 2640.4780240952055,
    "validity": 0.96875
  },
  "exp_id": "mmpae-1p5B-e200-bs64",
  "experiment_plan_metrics": {
    "Actual params from log": 1495526443,
    "Best Prop R2 \u2191": 0.6728669822216033,
    "Best Prop RMSE \u2193": 0.5227363497018814,
    "Exp ID": "mmpae-1p5B-e200-bs64",
    "Final ce_loss": 0.1119885389772201,
    "Final contrast_loss": 2.187204460526684,
    "Final epoch": 200,
    "Final mse_loss": 4.531615727267465,
    "Final total_loss": 2640.4780240952055,
    "Notes": "infer_steps=10, eval_steps=None",
    "Peak GPU memory": 92599.5908203125,
    "Run directory": "/data/runs/mmpae-1p5B-e200-bs64",
    "Runtime": "39h 06m 56s",
    "Tanimoto \u2191": 0.39310147781608873,
    "Target R2 \u2191": 0.7720341189253714,
    "Target RMSE \u2193": 0.4378301985802189,
    "Validity \u2191": 0.96875
  },
  "global_step": 204800,
  "is_best": false,
  "peak_gpu_memory_mb": 92599.5908203125,
  "phase": "Eval",
  "runtime": "39h 06m 56s",
  "runtime_seconds": 140816.74091649055,
  "score": 0.33420392034515256,
  "train": {
    "ce_loss": 0.1439999616559362,
    "contrast_loss": 0.8673558533191681,
    "eos_loss": 0.017205233216373017,
    "grad_norm": NaN,
    "mse_loss": 5.9219387280754745,
    "total_loss": 1459.6937276124954
  }
}