UniversalComputingResearch/Limen0.2B
963
1{2 "architectures": [3 "GPTForCausalLM"4 ],5 "auto_map": {6 "AutoConfig": "config.GPTConfig",7 "AutoModelForCausalLM": "model.GPTForCausalLM"8 },9 "block_size": 1024,10 "dtype": "float32",11 "head_dim": 128,12 "hidden_size": 768,13 "intermediate_size": 1920,14 "labels_are_shifted": true,15 "max_position_embeddings": 1024,16 "model_type": "gpt",17 "num_attention_heads": 6,18 "num_hidden_layers": 35,19 "num_key_value_heads": 2,20 "rms_norm_eps": 1e-06,21 "rope_theta": 100000.0,22 "tie_word_embeddings": true,23 "transformers_version": "5.14.1",24 "vocab_size": 16384,25 "xsa_projection": true26}27 