Models & Labs
Async GRPO with LoRA Enhances HF Jobs Efficiency
Hugging Face Bloghigh confidence
Why it matters
- →LoRA adapters reduce data transfer needs, optimizing resource usage.
- →Storage Buckets enable seamless communication between jobs without NCCL.
- →Enhances scalability and efficiency in AI training setups.
Hugging Face has enhanced its AsyncGRPOTrainer to support LoRA adapters, enabling more efficient training and inference processes. This update allows the synchronization of only the LoRA adapter, which is significantly smaller than the full model, reducing the need for extensive data transfer. By using Storage Buckets as shared filesystems, the setup facilitates communication between training and inference jobs without relying on NCCL. This development highlights the potential for scalable AI training setups using Hugging Face Jobs, optimizing resource usage and improving efficiency.
Read originalMore in Models & Labs
Models & Labsmodels
llama.cpp b10955 release addresses heap corruption
The latest release of llama.cpp, b10955, tackles a critical issue of heap corruption by disabling the ggml-cpu precompiled header and fixing CACHE_LINE_SIZE ambiguity. This update ensures consistent CACHE_LINE_SIZE values across C++ kernels and C work-buffer sizing code, preventing heap-buffer-overflow and subsequent crashes. By restoring the natural include order and removing the std::hardware_destructive_interference_size branch, the update makes the value deterministic and include-order independent. This release is a technical fix that stabilizes the runtime environment for developers using llama.cpp.
llama.cpp Releases
Models & Labsmodels
llama.cpp b10956 release enhances SYCL backend
The latest llama.cpp release, b10956, introduces significant improvements to the SYCL backend, particularly for handling large k values in TOP_K operations. By implementing a radix select method, the update allows for efficient GPU-resident processing, avoiding previous limitations that forced operations to fall back to the CPU. This change enhances performance, especially in scenarios requiring large k values, such as qwen4exp's sparse-attention indexer. The update ensures that operations are more efficient and scalable, providing a notable boost in processing speed without regressing any measured shapes.