[Bug]: FP8 on RDNA3/gfx1100 exceeds the 600 s engine-ready timeout on first start — no AMD_Radeon_Graphics tuned configs shipped

[Bug][ROCm/gfx942]: DeepSeek-V4-Flash silent retrieval corruption for prompts ≥ ~4-5k tokens (AITER sparse indexer)

[RFC]: LoRA adapter support for DFlash speculative decoding draft models

[Bug]: upgrade vllm from 0.26.0 to 0.27.0 run deepseek v4 flash error

[Bug]: online FP8 (--quantization fp8) produces corrupted, non-EOS-terminating output on Qwen2.5-1.5B-Instruct

[RFC]: Programmatic Session-Aware KV Cache Management

[Bug]: System crash: Rust frontend aborts the process on prompt logprobs for a zero-token prompt

[Bug]: LoRA set_lora on merged QKV/gate_up raises IndexError when lora_a is a single tensor

[CI Failure]: AMD Entrypoints Integration (LLM) hangs in test_guidance_no_additional_properties on mi300

[RFC][kv_offload]: Tiering Admission Policy Design