CoolFace
Modelpublic

laion/tasktrove-dq-unitsyn-python-step20-30b-a3b

sourceHugging Faceapache-2.0updated 2mo agoView on Hugging Face
0likes20downloads
config.json42 linesDownload Raw Back to root
1{2  "architectures": [3    "Qwen3MoeForCausalLM"4  ],5  "attention_bias": false,6  "attention_dropout": 0.0,7  "bos_token_id": null,8  "decoder_sparse_step": 1,9  "dtype": "bfloat16",10  "eos_token_id": 151645,11  "head_dim": 128,12  "hidden_act": "silu",13  "hidden_size": 2048,14  "initializer_range": 0.02,15  "intermediate_size": 6144,16  "max_position_embeddings": 262144,17  "max_window_layers": 48,18  "mlp_only_layers": [],19  "model_type": "qwen3_moe",20  "moe_intermediate_size": 768,21  "norm_topk_prob": true,22  "num_attention_heads": 32,23  "num_experts_per_tok": 8,24  "num_hidden_layers": 48,25  "num_key_value_heads": 4,26  "num_local_experts": 128,27  "output_router_logits": false,28  "pad_token_id": 151643,29  "rms_norm_eps": 1e-06,30  "rope_parameters": {31    "rope_theta": 10000000,32    "rope_type": "default"33  },34  "router_aux_loss_coef": 0.001,35  "sliding_window": null,36  "tie_word_embeddings": false,37  "transformers_version": "5.8.1",38  "use_cache": true,39  "use_sliding_window": false,40  "vocab_size": 15193641}42