Can I fine-tune LLMs with FrameTrain?

Yes. FrameTrain supports fine-tuning of Large Language Models (LLMs) via LoRA and QLoRA. You can import models like Llama, Mistral, or Phi directly from HuggingFace and adapt them to your own datasets.

Which GPUs are supported?

FrameTrain runs on all NVIDIA GPUs with CUDA support (from the GTX 10 series onward) as well as on Apple Silicon (M1/M2/M3/M4) via Metal Performance Shaders. Pure CPU-based training is also possible.

What's the difference compared to Google Colab or AWS SageMaker?

FrameTrain runs entirely locally on your machine. There are no ongoing cloud costs, your data is never uploaded, and you're not dependent on an internet connection. Ideal for sensitive or proprietary datasets.

Do I need programming knowledge?

No. FrameTrain is designed as a graphical application. You configure training parameters through an intuitive interface without writing any code. Advanced users can also use configuration files.

Does FrameTrain support LoRA training?

Yes, LoRA (Low-Rank Adaptation) is fully integrated. You can configure LoRA rank, alpha, and target modules directly in the app, which also enables fine-tuning on GPUs with limited VRAM.

Why does FrameTrain only cost €4.99/month?

FrameTrain is in Early Access. The price is intentionally low so that as many developers and researchers as possible can get access early. The price increases to €9.99 after the first 100 users – sign up now to lock in the lower price permanently.

Does FrameTrain run on Apple M1/M2/M3/M4?

Yes. FrameTrain natively supports Apple Silicon via Metal Performance Shaders (MPS). Models are accelerated on the M-chip's GPU without requiring CUDA.

Which model formats are supported?

FrameTrain supports PyTorch models (.pt, .bin), SafeTensors, and the GGUF format. Models can be imported directly from the HuggingFace Hub or loaded locally.

Can I use my own datasets?

Yes. You can load your own datasets in CSV, JSON, or JSONL format. FrameTrain also supports datasets directly from the HuggingFace Datasets Hub.

Does FrameTrain support mixed precision and gradient checkpointing?

Yes. FP16 and BF16 mixed-precision training are fully supported, as is gradient checkpointing for memory-efficient training on GPUs with limited VRAM.

Home GuidesLocal vs Cloud

ComparisonGuide

Local ML Training
vs Cloud – The Honest Comparison

Google Colab, AWS SageMaker, Paperspace – or your own GPU? We compare cost, privacy, flexibility, and performance for typical ML workloads.

Cost Comparison: Cloud vs Local

Cloud GPUs are billed by the hour. For intensive ML projects, costs add up quickly. Local training, on the other hand, has a one-time hardware cost – no further costs after that.

Provider / Solution	GPU	Price/Hour	100h Training
Google Colab Pro+	A100 40GB	~$1.30/h	~$130
AWS SageMaker	ml.g4dn.xlarge (T4)	~$0.95/h	~$95
Paperspace Gradient	A100	~$2.50/h	~$250
Lambda Cloud	RTX A6000	~$1.20/h	~$120
FrameTrain (local)	RTX 4070 (your own HW)	$0.00/h	$0 ¹

¹ After the one-time hardware investment. Electricity cost approx. $0.03–0.08 per hour (RTX 4070: ~200W).

Feature Comparison

Feature	Cloud	FrameTrain (Local)
Ongoing costs
GDPR compliance
Data leaves your device
Offline operation possible
Full GPU control
No vendor lock-in
Fast training start
Scaling to 100+ GPUs
No hardware investment

Privacy & GDPR – Why Local Is Better

With cloud services, your training data is uploaded to someone else's servers – in the US, Europe, or elsewhere. For many use cases that's problematic or simply not allowed:

Customer data

Under GDPR, customer data may not be transferred to US servers without explicit consent.

Medical data

Patient data is subject to special protection requirements that rule out cloud training.

Company data

Proprietary company data and trade secrets should never leave external servers.

Research data

Unpublished research results can become unintentionally exposed through a cloud upload.

When Does Cloud Make Sense?

Local training isn't always the best choice. Cloud makes sense for:

1
Training very large models (70B+ parameters) that need more VRAM than you have available
2
Horizontal scaling to hundreds of GPUs for fast experiments
3
No own hardware available and the training workload is one-off/low
4
Team workflows where multiple people need simultaneous access to the same resources

Start Training Locally with FrameTrain

Starting at €4.99/month – cancel anytime, no hidden costs, full control over your data.

Compare for Free

Read the LoRA Guide

Local ML Trainingvs Cloud – The Honest Comparison