NVIDIA NIM Connecteur
Integrate NVIDIA NIM (NVIDIA Inference Microservices) with OnPremiseAgent for optimized GPU-accelerated model serving. NIM provides pre-optimized containers for popular models with TensorRT-LLM acceleration, delivering maximum performance on NVIDIA hardware.
API Key
Authentification
API Key
Catégorie
Runtimes IA
Compatibilité
NVIDIA A100, NVIDIA H100, NVIDIA L40S, CUDA 12.x
Niveau
Bientôt disponible
Des capacités conçues sur mesure
Tout ce dont vous avez besoin pour intégrer NVIDIA NIM à vos workflows d'agents on-premise.
TensorRT-LLM
Hardware-optimized inference with TensorRT-LLM for maximum throughput on NVIDIA GPUs.
Pre-Built Containers
Deploy pre-optimized NIM containers for Llama, Mistral, and other popular models.
Multi-GPU Scaling
Automatic tensor parallelism across multiple GPUs for serving large models.
OpenAI Compatible
Industry-standard OpenAI-compatible API for seamless integration.
01
Pull NIM Container
Pull the NVIDIA NIM container for your chosen model from NVIDIA NGC catalog.
01
Avantages clés
Pourquoi les entreprises choisissent ce connecteur
- Maximum inference performance on NVIDIA hardware
- Pre-optimized containers — no manual optimization needed
- Multi-GPU tensor parallelism for large models
- Enterprise support from NVIDIA
01
Maximum Performance
Deploy models with TensorRT-LLM optimization for the lowest possible latency on NVIDIA hardware.
02
Large Model Serving
Serve 70B+ parameter models across multiple GPUs with automatic tensor parallelism.
03
Enterprise Standardization
Standardize on NVIDIA NIM for all AI inference with enterprise support and SLA guarantees.
Questions fréquentes
Do I need an NVIDIA AI Enterprise license?
NIM containers are available through NVIDIA NGC. Some models require an NVIDIA AI Enterprise subscription for production use.
Which models are available as NIM containers?
NIM supports Llama 3, Mistral, Mixtral, and many other popular models with pre-optimized TensorRT-LLM configurations.
Fonctionne parfaitement avec
Associez NVIDIA NIM à ces connecteurs pour une stack d'intégration complète.
Prêt à connecter NVIDIA NIM?
Déployez sur votre propre infrastructure avec une souveraineté totale des données. Lancez-vous en quelques minutes.