inference/research
verifiedhigh · 0.90updated 2026-10-04

RWKV/v5-Eagle-7B-HF config.json

repo

Fields

id
sources/mbl-cfg-rwkv5-eagle-7b
Publisher
Hugging Face
Kind
repo
Accessed
2026-10-04
notes
rwkv5. num_hidden_layers 32, hidden_size 4096. No num_attention_heads, no num_key_value_heads, no head_dim -> the WKV mixer is a linear-attention recurrence, not a softmax attention stack; KV cache and GQA are undefined. Read raw 2026-10-04.

Notes

rwkv5. num_hidden_layers 32, hidden_size 4096. No num_attention_heads, no num_key_value_heads, no head_dim -> the WKV mixer is a linear-attention recurrence, not a softmax attention stack; KV cache and GQA are undefined. Read raw 2026-10-04.

Cross-references

This record references no other records.