FAQ

BeaverYard is a fine-tuning service that takes your dataset, validates it before payment, and delivers downloadable artifacts and customer guides when the run finishes.

Click Start Run, upload your dataset, then review eligibility and the final price before you pay.

We validate your dataset first to confirm it's eligible. You'll see eligibility and the final price before payment.

Launch covers lower-capacity pricing bands. Orbit covers higher-capacity pricing bands.

Every run is pay-as-you-go by default. If you want to save up to 30% on every run plus get faster scheduling and support, see BeaverYard Plus and Pro below or on the pricing page.

No. Pricing is based on token caps and a standardized training recipe. You choose between LLaMA, Mistral, and Gemma — all at the same price.

No. You see the final price before paying, and there are no post-training surprise charges.

Payments are non-refundable after training begins. If validation fails before training starts, you won't be charged.

No. PAYG is always available. Memberships are for users who want savings, faster scheduling, and faster support.

Plus saves 20% on every run; Pro saves 30%. Both include Priority Processing and priority support. Use Pro if you expect 4 or more runs each month.

Plus is $3/mo or $30/yr (2 months free). Pro is $5/mo or $50/yr (2 months free).

Members skip the standard scheduling buffer, so runs typically dispatch sooner. PAYG users can still buy Priority Processing as a one-time checkout add-on for $2 (Launch) or $3 (Orbit) per run.

BeaverYard Plus and Pro members receive priority support with a target response time of 1–3 business days.

Yes. You can subscribe at any time, switch between Plus and Pro from the dashboard, and cancel at the end of any billing period without losing access until that period ends.

You'll see the reason and can fix your dataset. If validation fails before training starts, you are not charged.

Common reasons include:

• File is not a valid .jsonl file (we only accept JSONL)
• Missing required input/output structure
• Dataset exceeds the published size or token caps
• Record count is above 200,000 or one or more lines exceed 20,000 characters

Yes. Update your dataset to meet requirements and retry.

You'll download:

• An adapter package (adapter_model.safetensors + adapter_config.json)
• Training Focus Summary (what your adapter does, what to expect, what to do next)
• Usage Guidelines (how to load and test your adapter, where it shines, where it may struggle)

No. BeaverYard delivers downloadable artifacts and customer guides. Hosting and deployment are your responsibility.

Yes, anytime during your retention window. After it expires, files are automatically deleted.

Every run includes 7-day retention. After expiry, all files are permanently deleted.

Yes. Automatic deletion happens after the retention window, and you can request early deletion through support.

Your downloadable artifacts and customer guides are no longer available and are automatically deleted based on the retention policy.

Yes. You sign in with email and password, plus optional two-factor authentication.

We use your email for account access, status updates, and run-related notifications. We don't use your email for unrelated outreach.

Only you. Support can access it only when you request help. We apply strict access controls and auditing.

United States (AWS region: us-east-1).

You can monitor status from your run dashboard after signing in.

Contact support with your run ID and a short description of the issue.