ChristianZ97/SATP-aesop-policy

Aesop-RL Model Checkpoint

This repository contains trained model weights for the Aesop-RL project.

Model Details

• Model Type: Policy Network for Aesop Configuration
• Training Stage: warm_up
• Framework: PyTorch
• Lean Version: 4

Architecture Validation

• warm_up.pt: Valid (shared_mlp, tactic_heads, lemma_heads, config_heads, base_encoder)
• warm_up_best.pt: Valid (shared_mlp, tactic_heads, lemma_heads, config_heads, base_encoder)
• warm_up_final.pt: Valid (shared_mlp, tactic_heads, lemma_heads, config_heads, base_encoder)

Validation Results

• warm_up.pt: 25.00% (61/244)
• warm_up_best.pt: 25.82% (63/244)
• warm_up_final.pt: 25.82% (63/244)

Files

• warm_up.pt: PyTorch checkpoint (1056.4 MB)
• warm_up_best.pt: PyTorch checkpoint (1056.4 MB)
• warm_up_final.pt: PyTorch checkpoint (1056.4 MB)

Usage

import torch
from aesop.models.policy import AesopPolicy

# Load checkpoint
checkpoint = torch.load("warm_up.pt", map_location="cpu")
model = AesopPolicy()
model.load_state_dict(checkpoint['model_state_dict'])

Training Info

{
  "stage": "warm_up",
  "files_description": "- `warm_up.pt`: PyTorch checkpoint (1056.4 MB)\n- `warm_up_best.pt`: PyTorch checkpoint (1056.4 MB)\n- `warm_up_final.pt`: PyTorch checkpoint (1056.4 MB)",
  "architecture_summary": "- **warm_up.pt**: Valid (shared_mlp, tactic_heads, lemma_heads, config_heads, base_encoder)\n- **warm_up_best.pt**: Valid (shared_mlp, tactic_heads, lemma_heads, config_heads, base_encoder)\n- **warm_up_final.pt**: Valid (shared_mlp, tactic_heads, lemma_heads, config_heads, base_encoder)",
  "validation_summary": "- **warm_up.pt**: 25.00% (61/244)\n- **warm_up_best.pt**: 25.82% (63/244)\n- **warm_up_final.pt**: 25.82% (63/244)",
  "checkpoints": [
    {
      "stage": "warm_up",
      "keys": [
        "epoch",
        "model_state_dict",
        "trainer_state_dict",
        "train_loss",
        "val_acc",
        "best_val_acc"
      ],
      "epoch": 1,
      "best_val_acc": 0.25,
      "model_architecture": {
        "is_valid": true,
        "components": [
          {
            "name": "shared_mlp",
            "found_keys": 2,
            "total_expected": 3
          },
          {
            "name": "tactic_heads",
            "found_keys": 2,
            "total_expected": 2
          },
          {
            "name": "lemma_heads",
            "found_keys": 2,
            "total_expected": 2
          },
          {
            "name": "config_heads",
            "found_keys": 2,
            "total_expected": 2
          },
          {
            "name": "base_encoder",
            "found_keys": 280,
            "total_expected": "variable"
          }
        ],
        "warnings": [],
        "errors": []
      },
      "file_size_mb": 1056.3982362747192,
      "filename": "warm_up.pt"
    },
    {
      "stage": "warm_up",
      "keys": [
        "epoch",
        "model_state_dict",
        "trainer_state_dict",
        "train_loss",
        "val_acc",
        "best_val_acc"
      ],
      "epoch": 8,
      "best_val_acc": 0.2581967213114754,
      "model_architecture": {
        "is_valid": true,
        "components": [
          {
            "name": "shared_mlp",
            "found_keys": 2,
            "total_expected": 3
          },
          {
            "name": "tactic_heads",
            "found_keys": 2,
            "total_expected": 2
          },
          {
            "name": "lemma_heads",
            "found_keys": 2,
            "total_expected": 2
          },
          {
            "name": "config_heads",
            "found_keys": 2,
            "total_expected": 2
          },
          {
            "name": "base_encoder",
            "found_keys": 280,
            "total_expected": "variable"
          }
        ],
        "warnings": [],
        "errors": []
      },
      "file_size_mb": 1056.3955039978027,
      "filename": "warm_up_best.pt"
    },
    {
      "stage": "warm_up",
      "keys": [
        "model_state_dict",
        "trainer_state_dict",
        "final_train_loss",
        "final_val_acc",
        "best_val_acc"
      ],
      "best_val_acc": 0.2581967213114754,
      "model_architecture": {
        "is_valid": true,
        "components": [
          {
            "name": "shared_mlp",
            "found_keys": 2,
            "total_expected": 3
          },
          {
            "name": "tactic_heads",
            "found_keys": 2,
            "total_expected": 2
          },
          {
            "name": "lemma_heads",
            "found_keys": 2,
            "total_expected": 2
          },
          {
            "name": "config_heads",
            "found_keys": 2,
            "total_expected": 2
          },
          {
            "name": "base_encoder",
            "found_keys": 280,
            "total_expected": "variable"
          }
        ],
        "warnings": [],
        "errors": []
      },
      "file_size_mb": 1056.396414756775,
      "filename": "warm_up_final.pt"
    }
  ]
}

Citation

If you use this model, please cite the Aesop-RL project.