{ "architectures": [ "HARP1BCUDAModel" ], "model_type": "harp", "vocab_size": 32000, "hidden_size": 768, "num_hidden_layers": 12, "num_attention_heads": 12, "num_key_value_heads": 3, "intermediate_size": 2048, "head_dim": 64, "max_position_embeddings": 512, "rms_norm_eps": 1e-06, "rope_theta": 10000.0, "torch_dtype": "bfloat16", "transformers_version": "4.45.0", "organization": "FrontTerrain", "model_name": "HARP-1B", "creator": "Rishi Shandilya (@dar2quasar)", "ai_lead": "Arka Brahma", "step": 52000, "loss": 0.2773, "tokens_trained": 1400832000, "target_tokens": 50000000000 }