| { | |
| "model_name": "Qwen3-8M-GPT2", | |
| "model_type": "Qwen3ForCausalLM", | |
| "tokenizer": "gpt2", | |
| "dtype": "bfloat16", | |
| "vocab_size": 50257, | |
| "hidden_size": 128, | |
| "num_layers": 12, | |
| "num_attention_heads": 4, | |
| "num_key_value_heads": 2, | |
| "head_dim": 32, | |
| "intermediate_size": 384, | |
| "max_position_embeddings": 2048, | |
| "rope_theta": 10000, | |
| "parameters": 8796160, | |
| "tie_word_embeddings": true, | |
| "attention_type": "full_attention", | |
| "positional_encoding": "rope", | |
| "normalization": "rmsnorm", | |
| "activation": "swiglu", | |
| "xsa_enabled": true, | |
| "xsa_paper": "arxiv 2603.09078" | |
| } |