39 lines
1008 B
JSON
39 lines
1008 B
JSON
|
{
|
||
|
|
"_name_or_path": "Kendamarron/Qwen2.5-4x0.5B-cpt",
|
||
|
|
"architectures": [
|
||
|
|
"Qwen2MoeForCausalLM"
|
||
|
|
],
|
||
|
|
"attention_dropout": 0.0,
|
||
|
|
"decoder_sparse_step": 1,
|
||
|
|
"eos_token_id": 151645,
|
||
|
|
"hidden_act": "silu",
|
||
|
|
"hidden_size": 896,
|
||
|
|
"initializer_range": 0.02,
|
||
|
|
"intermediate_size": 4864,
|
||
|
|
"max_position_embeddings": 32768,
|
||
|
|
"max_window_layers": 24,
|
||
|
|
"mlp_only_layers": [],
|
||
|
|
"model_type": "qwen2_moe",
|
||
|
|
"moe_intermediate_size": 4864,
|
||
|
|
"norm_topk_prob": true,
|
||
|
|
"num_attention_heads": 14,
|
||
|
|
"num_experts": 4,
|
||
|
|
"num_experts_per_tok": 2,
|
||
|
|
"num_hidden_layers": 24,
|
||
|
|
"num_key_value_heads": 2,
|
||
|
|
"output_router_logits": false,
|
||
|
|
"rms_norm_eps": 1e-06,
|
||
|
|
"rope_scaling": null,
|
||
|
|
"rope_theta": 1000000.0,
|
||
|
|
"router_aux_loss_coef": 0.001,
|
||
|
|
"shared_expert_intermediate_size": 4864,
|
||
|
|
"sliding_window": null,
|
||
|
|
"tie_word_embeddings": true,
|
||
|
|
"torch_dtype": "bfloat16",
|
||
|
|
"transformers_version": "4.47.1",
|
||
|
|
"use_cache": false,
|
||
|
|
"use_mrope": false,
|
||
|
|
"use_sliding_window": false,
|
||
|
|
"vocab_size": 151936
|
||
|
|
}
|