SpecOffload: Unlocking Latent GPU Capacity for LLM Inference on Resource-Constrained Devices

Re-assign community