Mode A: Card Mandate Setup
Auto-FundConnect your credit/debit card with a ₹2 verification authorization. Clears per-job compute costs upon workload completion without manual pre-funding.
Mode B: Prepaid Tier Packs
8 fine-tuning compute tiers with instant balance creditingSelect your fine-tuning compute tier. Net credit is deposited directly into your balance ledger minus a fixed ₹30 platform service fee. Paddle processes global payments and compliance securely.
- ✓ Up to 24GB VRAM allocation
- ✓ Single-adapter LoRA training
- ✓ Standard community queue
- ✓ Up to 48GB VRAM allocation
- ✓ Multi-dataset validation runs
- ✓ Accelerated disk I/O
- ✓ Dual A100 / L40S allocation
- ✓ Fast checkpointing & snapshots
- ✓ High memory throughput
- ✓ 80GB SXM4/SXM5 VRAM
- ✓ Multi-epoch convergence runs
- ✓ Priority auto-failover
- ✓ 4x 80GB VRAM clusters
- ✓ Full FP16/BF16 tensor parallelism
- ✓ Priority queue placement
- ✓ 8x GPU Dedicated Node
- ✓ InfiniBand / NVLink interconnect
- ✓ Dedicated support channel
- ✓ 8x H100 80GB SXM5
- ✓ Uncapped NVLink inter-GPU bandwidth
- ✓ Instant warm starts
- ✓ Dedicated cluster reservation
- ✓ Zero-preemption guarantee
- ✓ Custom hardware runtime env
Financial Ledger & Audit Trail
Row-level cryptographic settlement history for compute deductions and tier top-ups.
Help, Guidance & Feedback Hub
Have questions about a training run, model checkpointing, account setup, or billing? Connect directly with the Vivacious Cloud engineering team for fast, personalized assistance.
Technical Support
Guidance on CLI deployment, script errors, loss anomalies, VRAM sizing, and mid-job checkpoint migration.
General & Billing Help
Assistance with prepaid balance recharges, mandate authorizations, invoice requests, account questions, and feature feedback.
Telegram Live Support
Connect directly with our engineers and active developer community on Telegram for quick chats, advice, and live status.
💡 Platform Guardrails & Policy Reminders
- 7-Day Weight Retention: Your completed models remain safely available for a full 7 days. Downloading your weights does not delete them.
- 10% Price Protection: If upstream wholesale spot market rates fluctuate more than 10% from the estimated rate mid-run, our system automatically pauses the job into a safe zone to guard your balance.
- Human Verification Guarantee: Automated security monitors strictly flag anomalies for review; no account is ever suspended by an automated bot.
Workspace Setup Guide
Follow the steps below to configure your local client and dispatch training sessions.
Important: Jobs may automatically move between data centers if the platform needs to switch GPU providers to maintain uptime or reduce cost. Your training will resume from the last saved checkpoint using standard fine-tuning methods. Models or setups requiring a specific provider's custom environment may not be compatible with this automatic failover — please ensure your model works with standard fine-tuning libraries before deploying.