fix(rope): YaRN attention_factor is (0.1*ln(factor)+1)*attn_factor, not the bare multiplier 444b2a1 verified joerowell commited on 1 day ago
README: require vLLM #47311 for tool calling; glm47 parser fallback on older builds 07133fb verified joerowell commited on 6 days ago
Fix sliding-window RoPE and expert loading for trust_remote_code a156999 verified joerowell commited on 10 days ago