GPU virtual machine provider - chunhualiao/public-docs GitHub Wiki

Given your workflow:

✅ VS Code + Cline + SSH remote access ✅ Train small language models (likely under 15B parameters) ✅ Need GPU support (e.g., A100, V100, 4090) ✅ Minimal friction to launch and connect

Here are top GPU cloud providers that best support your setup:

---

Best overall for your use case: GPU + SSH with minimal cloud fluff

• ✅ Supports SSH login, clean IP access.

• ✅ Compatible with VS Code Remote + cline SSH.

• ✅ Choose GPU, RAM, storage à la carte.

• ✅ Super clean web UI — select specs and launch in ~30 seconds.

• ✅ Prices:

  • RTX 4090: ~$0.48/hr
  • A100 80GB: ~$2.10/hr
  • V100: ~$0.75/hr

• ✅ Free public IP by default, no NAT hassles.

• 🌎 Data centers across US, Europe, Asia.

👉 https://tensordock.com

---

my verdict

• bad: too stripped-down containers: no vim, no package management
• cannot use vS code + Cline with ssh remote easily

Containerized but SSH-enabled when using “Secure Cloud” (not Community Cloud)

• ✅ Launch full Linux VMs with public IP, SSH.
• ✅ Works great with VS Code Remote if you choose Secure Cloud (not community pods).
• ✅ GPU options: A100, 3090, 4090, H100.
• ❗Some images are containerized by default, but you can bring your own setup.
• 🕹 UI makes it fast: pick an image, click go, wait ~1 minute.

👉 https://runpod.io

---

ML-first cloud with classic VM experience and public IP

• ✅ Full VM with SSH, easy to use with VS Code + cline.

• ✅ Powerful GPUs like A100 80GB, 4090, V100.

• ❗No spot pricing, but reliable long-running instances.

• ✅ Price example:

  • A100 80GB: $2.49/hr
  • RTX 4090: ~$1/hr

👉 https://lambdalabs.com/service/gpu-cloud

---

Extreme budget option—best A100 price on the market

• ✅ Supports SSH + public IP
• ✅ Slightly rougher UI than TensorDock, but very fast.
• 💰 $0.66/hr for A100 (on-demand), $0.19/hr (spot).
• ❗No browser Jupyter or container setup – just a clean Linux box.

👉 https://thundercompute.com

---

Provider	SSH Access	VS Code Remote	Easiest to Use	Best Price GPU
TensorDock	✅	✅	✅✅✅	🟢 Very Good (4090 ~$0.48/hr)
RunPod (Secure)	✅	✅	✅✅	🟢 (4090 ~$0.42/hr)
Lambda Labs	✅	✅	✅✅	🟠 (4090 ~$1/hr)
ThunderCompute	✅	✅	✅	🟢 Best for A100 ($0.66/hr)

---

1. Go to tensordock.com, sign up.

2. Choose instance type → Pick GPU → Pick storage → Launch.

3. Copy public IP, SSH in:

   ssh ubuntu@<public-ip>

4. Use VS Code → “Remote SSH” → Add SSH config entry or connect directly.

5. Start training your model with PyTorch/Transformers.

---

Would you like a pre-configured SSH + PyTorch + Transformers base image to save even more time? I can show you which providers support that too.