config.json, deepseek-ai/DeepSeek-V2
repo
Fields
id
sources/arch-hf-config-deepseek-v2Publisher
Hugging Face
Kind
repo
Accessed
2026-10-03
notes
60 layers, hidden 5120, kv_lora_rank 512, qk_rope_head_dim 64, qk_nope_head_dim 128, n_routed_experts 160, n_shared_experts 2, num_experts_per_tok 6, topk_group 3, n_group 8, max_position_embeddings 163840.
Notes
60 layers, hidden 5120, kv_lora_rank 512, qk_rope_head_dim 64, qk_nope_head_dim 128, n_routed_experts 160, n_shared_experts 2, num_experts_per_tok 6, topk_group 3, n_group 8, max_position_embeddings 163840.
Cross-references
This record references no other records.