{
  "config": {
    "med_dataset": "mschonhardt/MedBerta",
    "canon_dataset": "mschonhardt/MedBerta",
    "dataset_split": "train",
    "text_column": null,
    "doc_id_column": null,
    "order_column": null,
    "model_name": "xlm-roberta-large",
    "max_seq_length": 512,
    "mlm_probability": 0.15,
    "attn_implementation": "sdpa",
    "min_chunk_tokens": 64,
    "val_doc_fraction": 0.01,
    "test_doc_fraction": 0.0,
    "max_eval_docs_per_source": 400,
    "stratify_column": "category",
    "split_seed": 13,
    "train_corpus": "canon",
    "dedup_train": false,
    "diagnose_tokenizer": true,
    "tokenizer_sample_size": 20000,
    "normalize_nfc": false,
    "diagnose_only": false,
    "learning_rate": 0.0001,
    "weight_decay": 0.01,
    "adam_beta1": 0.9,
    "adam_beta2": 0.98,
    "adam_epsilon": 1e-06,
    "max_grad_norm": 1.0,
    "warmup_ratio": 0.06,
    "lr_scheduler_type": "linear",
    "num_train_epochs": 10.0,
    "max_steps": -1,
    "optim": "adamw_torch_fused",
    "early_stopping_patience": 5,
    "early_stopping_threshold": 0.0001,
    "per_device_train_batch_size": 64,
    "per_device_eval_batch_size": 64,
    "gradient_accumulation_steps": 16,
    "gradient_checkpointing": false,
    "auto_find_batch_size": true,
    "torch_compile": false,
    "dataloader_num_workers": 8,
    "eval_accumulation_steps": 50,
    "eval_steps": 500,
    "logging_steps": 50,
    "save_steps": 500,
    "save_total_limit": 3,
    "output_dir": "./MediBerta",
    "cache_dir": ".cache_packed",
    "seed": 42,
    "report_to_wandb": false,
    "preprocess_num_proc": 16,
    "resume": true,
    "dry_run": false
  },
  "environment": {
    "timestamp": "2026-06-21T18:28:11",
    "python": "3.11.2",
    "platform": "Linux-6.1.0-49-amd64-x86_64-with-glibc2.36",
    "torch": "2.12.1+cu130",
    "transformers": "5.12.1",
    "datasets": "5.0.0",
    "cuda_available": true,
    "cuda": "13.0",
    "gpu_name": "NVIDIA RTX PRO 6000 Blackwell Server Edition",
    "gpu_total_memory_gb": 95.0,
    "gpu_capability": "12.0",
    "bf16_supported": true,
    "torch_arch_list": [
      "sm_75",
      "sm_80",
      "sm_86",
      "sm_90",
      "sm_100",
      "sm_120"
    ],
    "gpu_arch_supported_by_torch": true
  }
}