Category: Tech Notes
-
Text Generation Inference (TGI) in Production: A Real Serving Example (Hugging Face Series, Part 12)
TGI turns a Hugging Face model into an OpenAI-compatible endpoint with one docker run. Here are the flags that decide whether it fits your VRAM, how to consume it, and an honest verdict now that TGI is in maintenance mode and Hugging Face points new builds at vLLM.
-
How to Run a Hugging Face Model: Inference Providers vs Endpoints vs Self-Host (Hugging Face Series, Part 11)
Three ways to serve a Hugging Face model: the serverless Inference Providers proxy, dedicated Inference Endpoints, or self-hosting TGI on your own GPUs. A build-vs-buy verdict for the infra team that owns the bill.
-
Pushing Models and Datasets to the Hugging Face Hub: Private Repos, Versioning, and Model Cards (Hugging Face Series, Part 10)
Pushing to the Hugging Face Hub is artifact promotion. Here is how to create private repos, scope tokens and org RBAC, pin consumers to a commit, write a model card for provenance, and move large files without melting your bandwidth.
-
Accelerate: Multi-GPU Training Without Rewriting Your Code (Hugging Face Series, Part 9)
How Hugging Face accelerate turns a single-GPU training script into a multi-GPU job, when to use DDP versus FSDP versus DeepSpeed, and what the choice means for interconnect, capacity, and cost on the boxes you run.
-
Fine-Tuning with Trainer and LoRA/PEFT: When You Cannot Afford a Full Fine-Tune (Hugging Face Series, Part 8)
Full fine-tuning a 7B model can need two 80GB GPUs you do not have. Here is how the Trainer, LoRA, and QLoRA change the capacity math, with runnable code and the failure modes.
-
safetensors and Model File Formats: Why the Format Is a Security Decision (Hugging Face Series, Part 7)
A PyTorch .bin checkpoint is a pickle, and loading one can run code on your host. Here is why safetensors fixes that by design, how to convert and load safely, and where to scan models before they reach your GPU hosts.
-
The Hugging Face datasets Library: Loading, Streaming, and Disk Survival (Hugging Face Series, Part 6)
The datasets library is a data-movement and storage problem before it is a data-science one. Here is how loading, streaming, and the Arrow cache actually behave, and how to keep them from filling your disk.
-
Hugging Face Tokenizers: Context Limits, Token Budgets, and Capacity (Hugging Face Series, Part 5)
Tokenizers turn text into the integers a model reads, and they decide your context limit, throughput and token bill. An infra-first guide to encoding, batching, padding waste, and what changed in Transformers v5.
-
Running Your First Model with Transformers Pipelines (Hugging Face Series, Part 4)
The transformers pipeline is the shortest path from a model repo to a running inference call. Here is what actually loads onto your box, and how device, device_map and dtype decide whether a model fits your GPU.
-
Hugging Face Access Tokens and the hf CLI: Credentials Done Right (Hugging Face Series, Part 3)
A Hugging Face access token is a credential, not a convenience. Here is how to scope, store, rotate and use tokens with the new hf CLI, written for the infrastructure engineer who already runs registries and secret stores.
-
Navigating the Hugging Face Hub: Models, Datasets, Spaces, and How to Read a Model Card (Hugging Face Series, Part 2)
The model card is provenance and license metadata you must vet before any model enters your environment. Here is how an infrastructure engineer reads the Hub fast and decides what to reject.
-
What Hugging Face Actually Is: the Hub, the Libraries, and the Map (Hugging Face Series, Part 1)
Hugging Face is a registry, a set of open-source libraries, and a company. For infrastructure engineers moving into AI, here is the whole platform mapped onto systems you already run.
Architect’s Toolkit
PJ’s Tools
VMware Cloud Foundation
- VCF Documentation
- VCF 9 Planning & Preparation Workbook
- VCF Bill of Materials (BoM)
- VMware Compatibility Guide
- VMware Interoperability Matrix
- VMware Configuration Maximums
- VMware Ports & Protocols
- VMware Hands-on Labs
- RVTools Download
Nutanix
AI & Cloud-Native Platform
- NVIDIA Build (Model Catalog)
- NVIDIA AI Enterprise Reference Architecture
- NVIDIA NIM Performance Benchmarking
- NVIDIA NGC Catalog
- NeMo Microservices Helm Chart
- Helm Charts Repository
- Hugging Face Models
Architecture & Design
About the Author
Dr Pranay Jha
Dr. Pranay Jha is a Cloud and AI Consultant with 18+ years of experience in hybrid cloud, virtualization, and enterprise infrastructure transformation. He specializes in VMware technologies, multi-cloud strategy, and Generative AI solutions. He holds a PhD in Computer Applications with research focused on Cloud and AI, has published multiple research papers, and has been a VMware vExpert since 2016 and a VMUG Community Leader.
You May Have Missed
DrJha