Skip to content

Latest commit

 

History

History
34 lines (24 loc) · 2.56 KB

deepseek-v4.md

File metadata and controls

34 lines (24 loc) · 2.56 KB

DeepSeek V4

DeepSeek-V4 is the next-generation Mixture-of-Experts language model from DeepSeek-AI. It extends the V3 design with Hyper-Connections (mHC) for multi-stream residual mixing, Compressed Sparse Attention (CSA) with a learned token-importance indexer (DSA), hash-routed MoE layers for the first few decoder blocks, and a refined Multi-Token Prediction (MTP) head with separate e_proj / h_proj projections.

DeepSeek V4 models are supported via the Bridge system with auto-detected configuration and weight mapping.

Model Architecture Features

  • Hybrid Attention (DSv4HybridSelfAttention): Per-layer mix of dense MLA and Compressed Sparse Attention selected by compress_ratios
  • Compressed Sparse Attention (CSA) with DSA Indexer: Top-k token selection over windowed keys; index_n_heads, index_head_dim, index_topk control the indexer
  • Hyper-Connections (mHC): 4-stream residual mixing per layer (hc_mult = 4) with sinkhorn-iterated attention; per-MTP-layer hc_head_* learns output contraction
  • Hash-Routed MoE: First few decoder layers use a deterministic vocab → expert mapping (tid2eid) instead of softmax routing
  • Multi-Token Prediction (MTP): One MTP layer with separate e_proj and h_proj projections (post-MCore #4518)
  • YaRN RoPE: rotary_scaling_factor=16, original_max_position_embeddings=65536; mscale=mscale_all_dim=1.0 for V4
  • Sigmoid Gating with Expert Bias: noaux_tc load balancing, sqrtsoftplus scoring, expert bias enabled
  • o_groups Output Projection: o_lora_rank low-rank output projection split into o_groups parallel groups

Examples, Parallelism, and Limitations

For checkpoint conversion and inference scripts, recommended parallelism settings, and current known limitations, see the DeepSeek V4 examples README.

Hugging Face Model Cards & References

Hugging Face Model Cards

Additional Resources

Related Docs