| { |
| "architectures": [ |
| "DeciLMForCausalLM" |
| ], |
| "attention_bias": false, |
| "attention_dropout": 0.0, |
| "auto_map": { |
| "AutoConfig": "configuration_decilm.DeciLMConfig", |
| "AutoModelForCausalLM": "modeling_decilm.DeciLMForCausalLM" |
| }, |
| "block_configs": [ |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 16, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 2.625, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 32, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 2.625, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 32, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 2.625, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 64, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 2.625, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 64, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 2.625, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 32, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 2.625, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 32, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 2.625, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 2.625, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 64, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 2.625, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 32, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 2.625, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 32, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 2.625, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 2.625, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 2.625, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 2.625, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 2.625, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 2.625, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": true, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": true, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": true, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": true, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": true, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": true, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": true, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": true, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": null, |
| "no_op": false, |
| "replace_with_linear": true |
| }, |
| "ffn": { |
| "ffn_mult": 1.3125, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| }, |
| { |
| "attention": { |
| "n_heads_in_group": 8, |
| "no_op": false, |
| "replace_with_linear": false |
| }, |
| "ffn": { |
| "ffn_mult": 5.25, |
| "no_op": false, |
| "replace_with_linear": false |
| } |
| } |
| ], |
| "bos_token_id": 128000, |
| "eos_token_id": 128009, |
| "hidden_act": "silu", |
| "hidden_size": 8192, |
| "initializer_range": 0.02, |
| "intermediate_size": null, |
| "max_position_embeddings": 131072, |
| "mlp_bias": false, |
| "model_type": "nemotron-nas", |
| "num_attention_heads": 64, |
| "num_hidden_layers": 80, |
| "num_key_value_heads": null, |
| "pretraining_tp": 1, |
| "rms_norm_eps": 1e-05, |
| "rope_scaling": { |
| "factor": 8.0, |
| "high_freq_factor": 4.0, |
| "low_freq_factor": 1.0, |
| "original_max_position_embeddings": 8192, |
| "rope_type": "llama3" |
| }, |
| "rope_theta": 500000.0, |
| "tie_word_embeddings": false, |
| "torch_dtype": "bfloat16", |
| "transformers_version": "4.51.3", |
| "use_cache": true, |
| "vocab_size": 128256 |
| } |
|
|