| { | |
| "architectures": [ | |
| "NemotronForCausalLM" | |
| ], | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "bos_token_id": 2, | |
| "eos_token_id": 3, | |
| "hidden_act": "relu2", | |
| "hidden_size": 18432, | |
| "initializer_range": 0.0063, | |
| "intermediate_size": 73728, | |
| "max_position_embeddings": 4096, | |
| "model_type": "nemotron", | |
| "num_attention_heads": 96, | |
| "num_hidden_layers": 96, | |
| "num_key_value_heads": 8, | |
| "quantization_config": { | |
| "activation_scheme": "dynamic", | |
| "ignored_layers": [ | |
| "lm_head" | |
| ], | |
| "quant_method": "fp8" | |
| }, | |
| "pretraining_tp": 1, | |
| "norm_eps": 1e-05, | |
| "rope_scaling": null, | |
| "rope_percent": 0.5, | |
| "rope_theta": 10000.0, | |
| "tie_word_embeddings": false, | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.40.0.dev0", | |
| "use_cache": true, | |
| "vocab_size": 256000 | |
| } |