1{
2 "architectures": ["LlamaForCausalLM"],
3 "model_type": "llama",
4 "hidden_size": 512,
5 "intermediate_size": 2048,
6 "num_hidden_layers": 8,
7 "num_attention_heads": 8,
8 "num_key_value_heads": 8,
9 "head_dim": 64,
10 "hidden_act": "silu",
11 "max_position_embeddings": 2048,
12 "vocab_size": 32768,
13 "attention_bias": false,
14 "attention_dropout": 0.0,
15 "mlp_bias": false,
16 "rms_norm_eps": 1e-06,
17 "rope_theta": 10000.0,
18 "rope_scaling": null,
19 "tie_word_embeddings": false,
20 "initializer_range": 0.02,
21 "bos_token_id": 1,
22 "eos_token_id": 2,
23 "torch_dtype": "float32",
24 "use_cache": true,
25 "transformers_version": "4.51.3"
26}
Il modello può generare contenuti scorretti o fuorvianti.
È responsabilità dell’utente:
1@misc{dac60m,
2 title={DAC60M: A Compact LLaMA-style Language Model},
3 author={M.INC. Research and Mattimax},
4 year={2025},
5 url={https://huggingface.co/Mattimax}
6}