model_config.json

{"max_len": 1024, "d_model": 960, "num_layers": 32, "num_heads": 15, "num_kv_heads": 5, "d_ff": 4096, "learning_rate": 0.0001, "batch_size": 48, "gradient_accumulation_steps": 2, "devices": "auto", "device_count": 2, "global_batch_size": 96, "effective_batch_size": 96, "global_effective_batch_size": 192, "lr_schedule": "warmup_cosine", "lr_warmup_steps": 1000, "min_learning_rate": 2e-05, "min_learning_rate_ratio": 0.2, "pad_token_id": 0, "bos_token_id": 2, "eos_token_id": 3, "train_dataset_signature": "44bbfca21e62efa8", "validation_dataset_signature": "8fb1014e14cb23c7", "train_dataset_case": {"name": "synthetic-textbook-jp-train", "genre": "textbook", "language": "ja", "dataset_path": "KeisukeMiyamoto/SyntheticTextbook-jp", "config_name": "default", "split": "train", "text_column": "rewrite"}, "validation_dataset_case": {"name": "synthetic-textbook-jp-validation", "genre": "textbook", "language": "ja", "dataset_path": "KeisukeMiyamoto/SyntheticTextbook-jp", "config_name": "default", "split": "validation", "text_column": "rewrite"}, "validation_cache_path": "models/lambda-360m-midtrained/validation-cache-8fb1014e14cb23c7-bucket-packing-v1-len1024-samples6144.pt", "validation_sample_count": 6144, "packing_version": "bucket-packing-v1", "shuffle_buffer_size": 10000, "base_shuffle_seed": 17, "shuffle_seed": 17, "trained_steps": 20480, "training_precision": "fp8", "float8_recipe": "tensorwise", "midtraining_source_model": "models/lambda-360m", "midtraining_max_steps": 20480}