| { | |
| "_name_or_path": "ds-8x6.7b-top-2-universal-evol-instruct-2e-5_bs_64_epoch_4_e_4", | |
| "add_weight_norm": false, | |
| "architectures": [ | |
| "LlamaMoEUpscalingForCausalLM" | |
| ], | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "bos_token_id": 32013, | |
| "calculator_type": "UniversalCalculator", | |
| "capacity_factor": 1.25, | |
| "drop_tokens": true, | |
| "dropped_padding": "zero", | |
| "eos_token_id": 32014, | |
| "ffn_folder_path": "ds-8x6.7b-top-2-universal-evol-instruct-2e-5_bs_64_epoch_4_e_4_ffn", | |
| "gate_add_noise": true, | |
| "gate_balance_loss_weight": 0.01, | |
| "gate_network": "mlp", | |
| "gate_noise_epsilon": 0.01, | |
| "gate_type": "TopKUniversalBalancedNoisyGate", | |
| "gate_use_balance": true, | |
| "gate_use_softmax": true, | |
| "hidden_act": "silu", | |
| "hidden_size": 4096, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 11008, | |
| "max_position_embeddings": 16384, | |
| "model_type": "llama_moe_upscaling", | |
| "multiply_gate_scores": true, | |
| "num_attention_heads": 32, | |
| "num_experts": 8, | |
| "num_hidden_layers": 32, | |
| "num_key_value_heads": 32, | |
| "num_selects": 2, | |
| "pad_token_id": 32014, | |
| "pretraining_tp": 1, | |
| "rms_norm_eps": 1e-06, | |
| "rope_scaling": { | |
| "factor": 4.0, | |
| "type": "linear" | |
| }, | |
| "rope_theta": 10000.0, | |
| "score_scale_factor": 1.0, | |
| "shared_expert_weight": 0.75, | |
| "size_experts": null, | |
| "tie_word_embeddings": false, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.40.1", | |
| "use_cache": true, | |
| "vocab_size": 32256 | |
| } | |