For anyone learning AI engineering, the order should be:
Transformers
RAG
Agents
Because agents are not magic. They sit on top of a model. The model is usually a transformer. RAG then gives the model external knowledge. Agents then use the model plus tools, memory, and workflows
In LLM training The key benefit of LoRA is efficiency!
LoRA means Low-Rank Adaptation. It is basically a way to fine-tune a large pretrained model without updating all of the model’s original weights.
🧬Cisco ACI uses a sharded distributed database architecture. Each “shard” is a partition of the overall configuration and operational state. Each APIC takes responsibility for certain shards. Think of it as spreading out the brain of the ACI fabric across multiple controllers.
Recently I had an engineer reach out to me, they wanted to enable SSH key-based authentication for logging into a large number of network devices as part of their work security requirement.
Basically, they wanted a way to automate the entire ssh key distribution process.
#devnet
@dmfigol@PulumiCorp It’s really great to use your favourite GPL to declare cloud infra rather that DSL. Pulumi is great, I used it sometime ago and not sure why I left it tbh? Time to head back!
I hope you all enjoy this easy to follow Nexus vPC Pair + Palo Alto HA Pair in 7 easy steps!
All configs are cli - the level of support and feedback on this has been absolutely amazing by the community on LinkedIn.
✌️
@danieldibswe If it was just for testing purposes- why not run it locally using an inference, such as huggingface/transformers or even Ollama?
Regardless, this is a bit too steep even on Nvidia H100 AZ
But then again google clouds TPU pod with multiple TPU v4 can cost at least 10k per hour
One thing I love about network automation- it forces you to learn a particular technology-if automating deployment/configuration SD devices such as ACI,ISE,SDA,Palo Alto,fortigate, F5 etc or traditional route/switch devices IOS/XE/NXOS
You will need to understand their workflow