| { | |
| "architectures": [ | |
| "MinimalLLMForCausalLM" | |
| ], | |
| "attention_bias": false, | |
| "auto_map": { | |
| "AutoConfig": "configuration_minimalllm.MinimalLLMConfig", | |
| "AutoModelForCausalLM": "modeling_minimalllm.MinimalLLMForCausalLM" | |
| }, | |
| "d_ff": 1536, | |
| "d_k": 48, | |
| "d_model": 384, | |
| "dropout": 0.1, | |
| "dtype": "float32", | |
| "max_seq_len": 512, | |
| "model_type": "minimal_llm", | |
| "n_heads": 8, | |
| "n_kv_groups": 2, | |
| "n_kv_heads": 4, | |
| "n_layers": 6, | |
| "rms_norm_eps": 1e-06, | |
| "sliding_window": 4096, | |
| "transformers_version": "5.13.1", | |
| "vocab_size": 63982, | |
| "tokenizer_class": "TokenizersBackend", | |
| "pad_token_id": 63982, | |
| "eos_token_id": 63982, | |
| "bos_token_id": 63982, | |
| "unk_token_id": 1 | |
| } |