luca0621/mmpae-1p5B-e260-2gpu-bs128-lr5e5-b2000

MMPAE 1.5B E260 2GPU BS128 LR5e-5 Beta2000

Checkpoint and evaluation artifacts for mmpae-1p5B-e260-2gpu-bs128-lr5e5-b2000.

Training

• Config: configs/Inverse_CwA_1p5B.yaml
• Parameters: about 1.496B
• Epochs: 260
• Batch size: 128
• Learning rate: 5e-5
• Alpha: 100
• Beta: 2000
• 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-e260-2gpu-bs128-lr5e5-b2000/Checkpoint_BEST.pt",
  "checkpoint_path": "/data/runs/mmpae-1p5B-e260-2gpu-bs128-lr5e5-b2000/Polyone_AE_0260.pt",
  "epoch": 260,
  "epochs": 260,
  "eval": {
    "ce_loss": 0.06199822479529171,
    "contrast_loss": 1.667678924016235,
    "eos_loss": 0.004588520026621823,
    "eval_batches": 1954,
    "eval_seconds": 2812.8629097938538,
    "infer_batches": 10,
    "infer_samples": 1280,
    "infer_seconds": 61.243648052215576,
    "inv_r2": 0.8862593219244033,
    "inv_rmse": 0.31655585015420434,
    "mse_loss": 1.1142550623587724,
    "prop_r2": 0.9547484576702118,
    "prop_rmse": 0.20580836683511733,
    "sim_score": 0.5962892207249391,
    "total_loss": 3446.84535515589,
    "validity": 0.97734375
  },
  "exp_id": "mmpae-1p5B-e260-2gpu-bs128-lr5e5-b2000",
  "experiment_plan_metrics": {
    "Actual params from log": 1495526443,
    "Best Prop R2 \u2191": 0.9547484576702118,
    "Best Prop RMSE \u2193": 0.20580836683511733,
    "Exp ID": "mmpae-1p5B-e260-2gpu-bs128-lr5e5-b2000",
    "Final ce_loss": 0.06199822479529171,
    "Final contrast_loss": 1.667678924016235,
    "Final epoch": 260,
    "Final mse_loss": 1.1142550623587724,
    "Final total_loss": 3446.84535515589,
    "Notes": "infer_steps=10, eval_steps=None",
    "Peak GPU memory": 94621.39013671875,
    "Run directory": "/data/runs/mmpae-1p5B-e260-2gpu-bs128-lr5e5-b2000",
    "Runtime": "56h 56m 25s",
    "Tanimoto \u2191": 0.5962892207249391,
    "Target R2 \u2191": 0.8862593219244033,
    "Target RMSE \u2193": 0.31655585015420434,
    "Validity \u2191": 0.97734375
  },
  "global_step": 266240,
  "is_best": false,
  "peak_gpu_memory_mb": 94621.39013671875,
  "phase": "Eval",
  "runtime": "56h 56m 25s",
  "runtime_seconds": 204985.66281485558,
  "score": 0.569703471770199,
  "train": {
    "ce_loss": 0.09326180632342584,
    "contrast_loss": 0.8654914498329163,
    "eos_loss": 0.011635430310889205,
    "grad_norm": Infinity,
    "mse_loss": 2.242397526046261,
    "total_loss": 1955.3159149885178
  }
}