Posts by Inesh Chakrabarti

4-bit KV Caching in LMCache: Offloading Quantized KV Beyond HBM for Context-Heavy Agents on AMD MI355X

As agents carry ever-longer context from turn to turn, the KV cache becomes the resource that runs out first. Two techniques ease that pressure from different angles: KV quantization makes each cached token smaller, while a hierarchical cache like LMCache spills cold KV to CPU DRAM. The two have mostly been developed independently.

Read more ...


Quark Support for HuggingFace Diffusers and SVDQuant

Diffusion models are heavy on memory and compute: a single text-to-image call runs a large transformer or UNet dozens of times. Quantization — storing weights (and sometimes activations) in low precision — is one of the most effective ways to cut both the memory footprint and the latency of these models.

Read more ...


Productionizing TurboQuant on AMD GPUs for KV-Cache-Bound LLM Inference

*The first three authors (Chakrabarti, Limpus, Rana) contributed equally to this work.

Read more ...