» Tag

transformers

16 posts

Jul 28, 20269.01
Transformers vLLM Backend Now Matches Native Inference Speed
The transformers vLLM backend now matches native inference speed via torch.fx graph analysis and automatic kernel fusion for LLMs.
huggingface llm-inference moe torch-fx
Jul 29, 20268.60
Transformer Transformer: A Unified Model for Robot Co-Design
A diffusion transformer that generates and validates complete robot embodiments from motion demos, using unified RoboTokens across robot classes.
co-design diffusion-models embodied-ai manipulation
Jul 21, 20268.61
Kimi Linear: Hybrid Attention Architecture Beats Full Attention
Kimi Linear's KDA module outperforms full attention, cutting KV cache by 75% and boosting throughput 6x at 1M context.
attention-mechanism inference-efficiency linear-attention llm
Ad
CommitBrief — AI code reviews, right in your terminal
A provider-agnostic, local-first CLI that reviews your staged changes, a historic range, or a whole GitHub pull request. Zero telemetry, no server. Free and open source.
commitbrief.com
Jul 26, 20268.10
Randomized KV-Cache Eviction Enables Verifiable Error Certificates
A new paper shows randomized KV-cache eviction, unlike deterministic top-k, enables verifiable per-step error certificates for LLM serving.
efficient-inference kv-cache llm-inference model-serving
Jul 8, 20267.90
Anthropic Discovers a Global Workspace Inside Language Models
Anthropic's new J-lens technique uncovers an internal region in Claude-like models that behaves like the brain's global workspace of consciousness.
ai anthropic cognitive-science interpretability
Jul 9, 20267.80
Inside LLM 'private thoughts': J-Space isn't consciousness, it's control
Researchers spot J-Space, an internal workspace in transformers guiding reasoning. Not consciousness, but a practical clue for AI developers and safety.
ai-safety alignment interpretability llm
Sep 24, 20267.62
Analog In-Memory Computing Attention Mechanism for Fast, Energy-Efficient LLM
Enhancing energy efficiency in attention mechanisms using analog in-memory computing.
ai energy-efficiency hardware ml
Jul 24, 20267.50
Compiler Converts Computation Graphs to Vanilla Transformer Weights
Generate transformer weights from computation graphs with a compiler, no training needed.
compilers huggingface machinelearning python
Oct 5, 20267.40
Full-fabric VHDL LLM Inference Engine: Runs Qwen3.5-class
The full-fabric VHDL LLM inference engine runs Qwen3.5-class transformer inference on FPGA.
fpga inference llm transformers
Aug 10, 20267.30
Meta Launches Muse Glimmer: Local, Agentic, Multimodal, and Open Source
Meta's new Muse Glimmer model is a multimodal AI solution designed for local use cases.
ai meta ml opensource