{ "_name_or_path": "/pure-mlo-scratch/akoepf/checkpoints/falcon-40b-megacode2-hf/", "alibi": false, "architectures": [ "FalconForCausalLM" ], "attention_dropout": 0.0, "bias": false, "bos_token_id": 65024, "eos_token_id": 65025, "hidden_dropout": 0.0, "hidden_size": 8192, "initializer_range": 0.02, "layer_norm_epsilon": 1e-05, "model_type": "falcon", "multi_query": true, "new_decoder_architecture": true, "num_attention_heads": 128, "num_hidden_layers": 60, "num_kv_heads": 8, "parallel_attn": true, "torch_dtype": "bfloat16", "transformers_version": "4.31.0", "use_cache": true, "vocab_size": 65152 }