Entries matching “huggingface”

Hugging Face's launch post covers day-0 transformers, llama.cpp and vLLM support, Inference Endpoints, speculative decoding, TRL fine-tuning and agent demos for Muse Glimmer.

A quantization-aware-trained Q4_0 GGUF of Muse Glimmer 30B for llama.cpp. On held-out tokens it measures closer to BF16 than Meta's official Q4_K_M: 0.0213 vs 0.0228 KL and 95.9% vs 95.6% top-token agreement.

A coding-specialised speculative-decoding drafter for Muse Glimmer 30B, fine-tuned on software-engineering traces. It raises mean acceptance length to about 6.9 versus 4.0 for the community DSpark, for about 3.8x over no speculation.

A prebuilt macOS arm64 bundle for the Muse Glimmer voice-agent recipe in meta-oss-cookbook: Parakeet speech helper, Muse Glimmer worker and Supertonic TTS executables built from one pinned ExecuTorch checkout, plus the shared MLX Metal library.

A 4-byte GGUF header edit that raises Muse Glimmer 30B's 131,072 context limit to 262K. The author shows YaRN scaling only adds prefill cost on this architecture, and verified 262K with 8 slots on one RTX 4090.

Meta's official Muse-Glimmer-30B repo: ~29.6B dense model with a 1.8B vision encoder, 131K context, Apache 2.0, with vLLM and SGLang serve commands.