Tracked topic
Qwen3 is an AI model family developed by Alibaba, released as a set of large language models for natural-language tasks.
Experimental results for pruning and distillation from Qwen3 8B using Model Optimizer show that Qwen3 Depth Pruned 6B model is 30% faster than the Qwen3 4B model, and it also performs better on the MMLU (Massive Multitask Language Understanding) benchmark. Depth pruning was applied to reduce the model from 36 to 24 layers, resulting in a 6B model, using one NVIDIA H100 80 GB HBM3. The Pruned model is distilled from Qwen3-8B using the OptimalScale/ClimbMix data processed from nvidia/ClimbMix pretraining dataset. The experiment uses 25% of the data, which is approximately 90B tokens. Distillatio
Pruning and Distilling LLMs Using NVIDIA TensorRT Model Optimizer | NVIDIA Technical Blog…GR00T 1.7 introduces a vision-language-action model pretrained on approximately 32,000 hours of real human demonstration and 8,000 hours of simulation, incorporating a Cosmos-Reason2-2B backbone (Qwen3…
…Over-provisioning a static table wastes memory on rows that will never be touched, while under-provisioning causes expensive copies that can degrade model performance and quality. DynamicEmb replaces static tables with…
…These changes build on the performance considerations outlined in our first post , which focused on the serving architecture underneath agentic inference: the frontend, router, and KV cache management. This follow-up focuses…
…Riva Parakeet ASR for real-time speech recognition, Chatterbox TTS for expressive voice synthesis Audio2Face for facial animation and performance delivery The latest workflow allows NPCs to perform dynamically in-character during…
…How does Nemotron 3.5 Lightning perform on the accuracy-speed Pareto frontier? Nemotron 3.5 Lightning delivers leading accuracy at the highest output speed in its class, winning the accuracy-versus…
…Performance, resiliency, maturity Now in its sixth generation, NVLink provides the delivered performance and resiliency needed to maximize factory goodput, and over ten years of scale-up technology investment and proven deployments…
…VLMs are trained on massive amounts of vision-text data and show notable zero-shot performance on many vision tasks. The VLA recipe then adapts these pretrained representations for action generation. However…
…Dense transformers (Llama, GPT, Qwen): All parameters participate in every forward pass. ~0% exact zeros → ~1.25× ANS, ~1.27× ZSTD. MoE models (Mixtral, DeepSeek, OLMoE): Only a subset of experts activate…
…software-based time-slicing/MPS (enabling dynamic sharing but lacking hardware isolation) and hardware-based NVIDIA Multi-Instance GPU (MIG) partitioning (delivering strict fault isolation and deterministic performance). Benchmarking a production voice…
…define the question, set the budget, decide which mutations are allowed, and review the ledger, while the AI agent performs the repetitive work of trying bounded candidate strategies and recording the results…
To show you the most relevant results, we’ve omitted some entries very similar to those already shown. Repeat the search with the omitted results included.