[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6GGPYSXYY1DQEB1FZGPHD2X":3,"skill-same:01M6GGPYSXYY1DQEB1FZGPHD2X":76},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":39,"linkOnly":40,"markdownContent":41,"markdownBody":42,"frontMatter":43,"refName":54,"commitSha":55,"skillDirectory":56,"forksCount":57,"repositoryHtmlUrl":58,"repositoryHomepageUrl":6,"ownerHtmlUrl":59,"ownerBlogUrl":6,"ownerDescription":60,"updatedAt":61,"install":62,"resources":75},"01M6GGPYSXYY1DQEB1FZGPHD2X","lambda-labs-gpu-cloud",null,"Reserved and on-demand GPU cloud instances for ML training and inference. Use when you need dedicated GPU instances with simple SSH access, persistent filesystems, or high-performance multi-node clusters for large-scale training.","zapabob\u002FSkills","zapabob","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F130830376?v=4","01M6DBMYZAGRTFQDSQYW3W764P","峯岸 亮","01M6DBMZ3GKDSJ01NZ9XTZD7JD",2,"MIT","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Farchive\u002Fskill-md-reviewed\u002Fsha256\u002Fef\u002Fef10cb06a745504fc8997e9a9d6fdb03774ea0cef36ff2994279afa65b55cda6\u002FSKILL.md",[],[19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38],"gpu","cloud","GPU Cloud","Lambda Labs","lambda","lambda labs gpu cloud","access","you","with","Multi-Node","multi","high-performance","instances","use","node","performance","Large-Scale Training","training","infrastructure","ml","2026-11-10 10:12:30",false,"---\r\nname: lambda-labs-gpu-cloud\r\ndescription: Reserved and on-demand GPU cloud instances for ML training and inference. Use when you need dedicated GPU instances with simple SSH access, persistent filesystems, or high-performance multi-node clusters for large-scale training.\r\nversion: 1.0.0\r\nauthor: Orchestra Research\r\nlicense: MIT\r\ndependencies: [lambda-cloud-client>=1.0.0]\r\nmetadata:\r\n  hermes:\r\n    tags: [Infrastructure, GPU Cloud, Training, Inference, Lambda Labs]\r\n\r\n---\r\n\r\n# Lambda Labs GPU Cloud\r\n\r\nComprehensive guide to running ML workloads on Lambda Labs GPU cloud with on-demand instances and 1-Click Clusters.\r\n\r\n## When to use Lambda Labs\r\n\r\n**Use Lambda Labs when:**\r\n- Need dedicated GPU instances with full SSH access\r\n- Running long training jobs (hours to days)\r\n- Want simple pricing with no egress fees\r\n- Need persistent storage across sessions\r\n- Require high-performance multi-node clusters (16-512 GPUs)\r\n- Want pre-installed ML stack (Lambda Stack with PyTorch, CUDA, NCCL)\r\n\r\n**Key features:**\r\n- **GPU variety**: B200, H100, GH200, A100, A10, A6000, V100\r\n- **Lambda Stack**: Pre-installed PyTorch, TensorFlow, CUDA, cuDNN, NCCL\r\n- **Persistent filesystems**: Keep data across instance restarts\r\n- **1-Click Clusters**: 16-512 GPU Slurm clusters with InfiniBand\r\n- **Simple pricing**: Pay-per-minute, no egress fees\r\n- **Global regions**: 12+ regions worldwide\r\n\r\n**Use alternatives instead:**\r\n- **Modal**: For serverless, auto-scaling workloads\r\n- **SkyPilot**: For multi-cloud orchestration and cost optimization\r\n- **RunPod**: For cheaper spot instances and serverless endpoints\r\n- **Vast.ai**: For GPU marketplace with lowest prices\r\n\r\n## Quick start\r\n\r\n### Account setup\r\n\r\n1. Create account at https:\u002F\u002Flambda.ai\r\n2. Add payment method\r\n3. Generate API key from dashboard\r\n4. Add SSH key (required before launching instances)\r\n\r\n### Launch via console\r\n\r\n1. Go to https:\u002F\u002Fcloud.lambda.ai\u002Finstances\r\n2. Click \"Launch instance\"\r\n3. Select GPU type and region\r\n4. Choose SSH key\r\n5. Optionally attach filesystem\r\n6. Launch and wait 3-15 minutes\r\n\r\n### Connect via SSH\r\n\r\n```bash\r\n# Get instance IP from console\r\nssh ubuntu@\u003CINSTANCE-IP>\r\n\r\n# Or with specific key\r\nssh -i ~\u002F.ssh\u002Flambda_key ubuntu@\u003CINSTANCE-IP>\r\n```\r\n\r\n## GPU instances\r\n\r\n### Available GPUs\r\n\r\n| GPU | VRAM | Price\u002FGPU\u002Fhr | Best For |\r\n|-----|------|--------------|----------|\r\n| B200 SXM6 | 180 GB | $4.99 | Largest models, fastest training |\r\n| H100 SXM | 80 GB | $2.99-3.29 | Large model training |\r\n| H100 PCIe | 80 GB | $2.49 | Cost-effective H100 |\r\n| GH200 | 96 GB | $1.49 | Single-GPU large models |\r\n| A100 80GB | 80 GB | $1.79 | Production training |\r\n| A100 40GB | 40 GB | $1.29 | Standard training |\r\n| A10 | 24 GB | $0.75 | Inference, fine-tuning |\r\n| A6000 | 48 GB | $0.80 | Good VRAM\u002Fprice ratio |\r\n| V100 | 16 GB | $0.55 | Budget training |\r\n\r\n### Instance configurations\r\n\r\n```\r\n8x GPU: Best for distributed training (DDP, FSDP)\r\n4x GPU: Large models, multi-GPU training\r\n2x GPU: Medium workloads\r\n1x GPU: Fine-tuning, inference, development\r\n```\r\n\r\n### Launch times\r\n\r\n- Single-GPU: 3-5 minutes\r\n- Multi-GPU: 10-15 minutes\r\n\r\n## Lambda Stack\r\n\r\nAll instances come with Lambda Stack pre-installed:\r\n\r\n```bash\r\n# Included software\r\n- Ubuntu 22.04 LTS\r\n- NVIDIA drivers (latest)\r\n- CUDA 12.x\r\n- cuDNN 8.x\r\n- NCCL (for multi-GPU)\r\n- PyTorch (latest)\r\n- TensorFlow (latest)\r\n- JAX\r\n- JupyterLab\r\n```\r\n\r\n### Verify installation\r\n\r\n```bash\r\n# Check GPU\r\nnvidia-smi\r\n\r\n# Check PyTorch\r\npython -c \"import torch; print(torch.cuda.is_available())\"\r\n\r\n# Check CUDA version\r\nnvcc --version\r\n```\r\n\r\n## Python API\r\n\r\n### Installation\r\n\r\n```bash\r\npip install lambda-cloud-client\r\n```\r\n\r\n### Authentication\r\n\r\n```python\r\nimport os\r\nimport lambda_cloud_client\r\n\r\n# Configure with API key\r\nconfiguration = lambda_cloud_client.Configuration(\r\n    host=\"https:\u002F\u002Fcloud.lambdalabs.com\u002Fapi\u002Fv1\",\r\n    access_token=os.environ[\"LAMBDA_API_KEY\"]\r\n)\r\n```\r\n\r\n### List available instances\r\n\r\n```python\r\nwith lambda_cloud_client.ApiClient(configuration) as api_client:\r\n    api = lambda_cloud_client.DefaultApi(api_client)\r\n\r\n    # Get available instance types\r\n    types = api.instance_types()\r\n    for name, info in types.data.items():\r\n        print(f\"{name}: {info.instance_type.description}\")\r\n```\r\n\r\n### Launch instance\r\n\r\n```python\r\nfrom lambda_cloud_client.models import LaunchInstanceRequest\r\n\r\nrequest = LaunchInstanceRequest(\r\n    region_name=\"us-west-1\",\r\n    instance_type_name=\"gpu_1x_h100_sxm5\",\r\n    ssh_key_names=[\"my-ssh-key\"],\r\n    file_system_names=[\"my-filesystem\"],  # Optional\r\n    name=\"training-job\"\r\n)\r\n\r\nresponse = api.launch_instance(request)\r\ninstance_id = response.data.instance_ids[0]\r\nprint(f\"Launched: {instance_id}\")\r\n```\r\n\r\n### List running instances\r\n\r\n```python\r\ninstances = api.list_instances()\r\nfor instance in instances.data:\r\n    print(f\"{instance.name}: {instance.ip} ({instance.status})\")\r\n```\r\n\r\n### Terminate instance\r\n\r\n```python\r\nfrom lambda_cloud_client.models import TerminateInstanceRequest\r\n\r\nrequest = TerminateInstanceRequest(\r\n    instance_ids=[instance_id]\r\n)\r\napi.terminate_instance(request)\r\n```\r\n\r\n### SSH key management\r\n\r\n```python\r\nfrom lambda_cloud_client.models import AddSshKeyRequest\r\n\r\n# Add SSH key\r\nrequest = AddSshKeyRequest(\r\n    name=\"my-key\",\r\n    public_key=\"ssh-rsa AAAA...\"\r\n)\r\napi.add_ssh_key(request)\r\n\r\n# List keys\r\nkeys = api.list_ssh_keys()\r\n\r\n# Delete key\r\napi.delete_ssh_key(key_id)\r\n```\r\n\r\n## CLI with curl\r\n\r\n### List instance types\r\n\r\n```bash\r\ncurl -u $LAMBDA_API_KEY: \\\r\n  https:\u002F\u002Fcloud.lambdalabs.com\u002Fapi\u002Fv1\u002Finstance-types | jq\r\n```\r\n\r\n### Launch instance\r\n\r\n```bash\r\ncurl -u $LAMBDA_API_KEY: \\\r\n  -X POST https:\u002F\u002Fcloud.lambdalabs.com\u002Fapi\u002Fv1\u002Finstance-operations\u002Flaunch \\\r\n  -H \"Content-Type: application\u002Fjson\" \\\r\n  -d '{\r\n    \"region_name\": \"us-west-1\",\r\n    \"instance_type_name\": \"gpu_1x_h100_sxm5\",\r\n    \"ssh_key_names\": [\"my-key\"]\r\n  }' | jq\r\n```\r\n\r\n### Terminate instance\r\n\r\n```bash\r\ncurl -u $LAMBDA_API_KEY: \\\r\n  -X POST https:\u002F\u002Fcloud.lambdalabs.com\u002Fapi\u002Fv1\u002Finstance-operations\u002Fterminate \\\r\n  -H \"Content-Type: application\u002Fjson\" \\\r\n  -d '{\"instance_ids\": [\"\u003CINSTANCE-ID>\"]}' | jq\r\n```\r\n\r\n## Persistent storage\r\n\r\n### Filesystems\r\n\r\nFilesystems persist data across instance restarts:\r\n\r\n```bash\r\n# Mount location\r\n\u002Flambda\u002Fnfs\u002F\u003CFILESYSTEM_NAME>\r\n\r\n# Example: save checkpoints\r\npython train.py --checkpoint-dir \u002Flambda\u002Fnfs\u002Fmy-storage\u002Fcheckpoints\r\n```\r\n\r\n### Create filesystem\r\n\r\n1. Go to Storage in Lambda console\r\n2. Click \"Create filesystem\"\r\n3. Select region (must match instance region)\r\n4. Name and create\r\n\r\n### Attach to instance\r\n\r\nFilesystems must be attached at instance launch time:\r\n- Via console: Select filesystem when launching\r\n- Via API: Include `file_system_names` in launch request\r\n\r\n### Best practices\r\n\r\n```bash\r\n# Store on filesystem (persists)\r\n\u002Flambda\u002Fnfs\u002Fstorage\u002F\r\n  ├── datasets\u002F\r\n  ├── checkpoints\u002F\r\n  ├── models\u002F\r\n  └── outputs\u002F\r\n\r\n# Local SSD (faster, ephemeral)\r\n\u002Fhome\u002Fubuntu\u002F\r\n  └── working\u002F  # Temporary files\r\n```\r\n\r\n## SSH configuration\r\n\r\n### Add SSH key\r\n\r\n```bash\r\n# Generate key locally\r\nssh-keygen -t ed25519 -f ~\u002F.ssh\u002Flambda_key\r\n\r\n# Add public key to Lambda console\r\n# Or via API\r\n```\r\n\r\n### Multiple keys\r\n\r\n```bash\r\n# On instance, add more keys\r\necho 'ssh-rsa AAAA...' >> ~\u002F.ssh\u002Fauthorized_keys\r\n```\r\n\r\n### Import from GitHub\r\n\r\n```bash\r\n# On instance\r\nssh-import-id gh:username\r\n```\r\n\r\n### SSH tunneling\r\n\r\n```bash\r\n# Forward Jupyter\r\nssh -L 8888:localhost:8888 ubuntu@\u003CIP>\r\n\r\n# Forward TensorBoard\r\nssh -L 6006:localhost:6006 ubuntu@\u003CIP>\r\n\r\n# Multiple ports\r\nssh -L 8888:localhost:8888 -L 6006:localhost:6006 ubuntu@\u003CIP>\r\n```\r\n\r\n## JupyterLab\r\n\r\n### Launch from console\r\n\r\n1. Go to Instances page\r\n2. Click \"Launch\" in Cloud IDE column\r\n3. JupyterLab opens in browser\r\n\r\n### Manual access\r\n\r\n```bash\r\n# On instance\r\njupyter lab --ip=0.0.0.0 --port=8888\r\n\r\n# From local machine with tunnel\r\nssh -L 8888:localhost:8888 ubuntu@\u003CIP>\r\n# Open http:\u002F\u002Flocalhost:8888\r\n```\r\n\r\n## Training workflows\r\n\r\n### Single-GPU training\r\n\r\n```bash\r\n# SSH to instance\r\nssh ubuntu@\u003CIP>\r\n\r\n# Clone repo\r\ngit clone https:\u002F\u002Fgithub.com\u002Fuser\u002Fproject\r\ncd project\r\n\r\n# Install dependencies\r\npip install -r requirements.txt\r\n\r\n# Train\r\npython train.py --epochs 100 --checkpoint-dir \u002Flambda\u002Fnfs\u002Fstorage\u002Fcheckpoints\r\n```\r\n\r\n### Multi-GPU training (single node)\r\n\r\n```python\r\n# train_ddp.py\r\nimport torch\r\nimport torch.distributed as dist\r\nfrom torch.nn.parallel import DistributedDataParallel as DDP\r\n\r\ndef main():\r\n    dist.init_process_group(\"nccl\")\r\n    rank = dist.get_rank()\r\n    device = rank % torch.cuda.device_count()\r\n\r\n    model = MyModel().to(device)\r\n    model = DDP(model, device_ids=[device])\r\n\r\n    # Training loop...\r\n\r\nif __name__ == \"__main__\":\r\n    main()\r\n```\r\n\r\n```bash\r\n# Launch with torchrun (8 GPUs)\r\ntorchrun --nproc_per_node=8 train_ddp.py\r\n```\r\n\r\n### Checkpoint to filesystem\r\n\r\n```python\r\nimport os\r\n\r\ncheckpoint_dir = \"\u002Flambda\u002Fnfs\u002Fmy-storage\u002Fcheckpoints\"\r\nos.makedirs(checkpoint_dir, exist_ok=True)\r\n\r\n# Save checkpoint\r\ntorch.save({\r\n    'epoch': epoch,\r\n    'model_state_dict': model.state_dict(),\r\n    'optimizer_state_dict': optimizer.state_dict(),\r\n    'loss': loss,\r\n}, f\"{checkpoint_dir}\u002Fcheckpoint_{epoch}.pt\")\r\n```\r\n\r\n## 1-Click Clusters\r\n\r\n### Overview\r\n\r\nHigh-performance Slurm clusters with:\r\n- 16-512 NVIDIA H100 or B200 GPUs\r\n- NVIDIA Quantum-2 400 Gb\u002Fs InfiniBand\r\n- GPUDirect RDMA at 3200 Gb\u002Fs\r\n- Pre-installed distributed ML stack\r\n\r\n### Included software\r\n\r\n- Ubuntu 22.04 LTS + Lambda Stack\r\n- NCCL, Open MPI\r\n- PyTorch with DDP and FSDP\r\n- TensorFlow\r\n- OFED drivers\r\n\r\n### Storage\r\n\r\n- 24 TB NVMe per compute node (ephemeral)\r\n- Lambda filesystems for persistent data\r\n\r\n### Multi-node training\r\n\r\n```bash\r\n# On Slurm cluster\r\nsrun --nodes=4 --ntasks-per-node=8 --gpus-per-node=8 \\\r\n  torchrun --nnodes=4 --nproc_per_node=8 \\\r\n  --rdzv_backend=c10d --rdzv_endpoint=$MASTER_ADDR:29500 \\\r\n  train.py\r\n```\r\n\r\n## Networking\r\n\r\n### Bandwidth\r\n\r\n- Inter-instance (same region): up to 200 Gbps\r\n- Internet outbound: 20 Gbps max\r\n\r\n### Firewall\r\n\r\n- Default: Only port 22 (SSH) open\r\n- Configure additional ports in Lambda console\r\n- ICMP traffic allowed by default\r\n\r\n### Private IPs\r\n\r\n```bash\r\n# Find private IP\r\nip addr show | grep 'inet '\r\n```\r\n\r\n## Common workflows\r\n\r\n### Workflow 1: Fine-tuning LLM\r\n\r\n```bash\r\n# 1. Launch 8x H100 instance with filesystem\r\n\r\n# 2. SSH and setup\r\nssh ubuntu@\u003CIP>\r\npip install transformers accelerate peft\r\n\r\n# 3. Download model to filesystem\r\npython -c \"\r\nfrom transformers import AutoModelForCausalLM\r\nmodel = AutoModelForCausalLM.from_pretrained('meta-llama\u002FLlama-2-7b-hf')\r\nmodel.save_pretrained('\u002Flambda\u002Fnfs\u002Fstorage\u002Fmodels\u002Fllama-2-7b')\r\n\"\r\n\r\n# 4. Fine-tune with checkpoints on filesystem\r\naccelerate launch --num_processes 8 train.py \\\r\n  --model_path \u002Flambda\u002Fnfs\u002Fstorage\u002Fmodels\u002Fllama-2-7b \\\r\n  --output_dir \u002Flambda\u002Fnfs\u002Fstorage\u002Foutputs \\\r\n  --checkpoint_dir \u002Flambda\u002Fnfs\u002Fstorage\u002Fcheckpoints\r\n```\r\n\r\n### Workflow 2: Batch inference\r\n\r\n```bash\r\n# 1. Launch A10 instance (cost-effective for inference)\r\n\r\n# 2. Run inference\r\npython inference.py \\\r\n  --model \u002Flambda\u002Fnfs\u002Fstorage\u002Fmodels\u002Ffine-tuned \\\r\n  --input \u002Flambda\u002Fnfs\u002Fstorage\u002Fdata\u002Finputs.jsonl \\\r\n  --output \u002Flambda\u002Fnfs\u002Fstorage\u002Fdata\u002Foutputs.jsonl\r\n```\r\n\r\n## Cost optimization\r\n\r\n### Choose right GPU\r\n\r\n| Task | Recommended GPU |\r\n|------|-----------------|\r\n| LLM fine-tuning (7B) | A100 40GB |\r\n| LLM fine-tuning (70B) | 8x H100 |\r\n| Inference | A10, A6000 |\r\n| Development | V100, A10 |\r\n| Maximum performance | B200 |\r\n\r\n### Reduce costs\r\n\r\n1. **Use filesystems**: Avoid re-downloading data\r\n2. **Checkpoint frequently**: Resume interrupted training\r\n3. **Right-size**: Don't over-provision GPUs\r\n4. **Terminate idle**: No auto-stop, manually terminate\r\n\r\n### Monitor usage\r\n\r\n- Dashboard shows real-time GPU utilization\r\n- API for programmatic monitoring\r\n\r\n## Common issues\r\n\r\n| Issue | Solution |\r\n|-------|----------|\r\n| Instance won't launch | Check region availability, try different GPU |\r\n| SSH connection refused | Wait for instance to initialize (3-15 min) |\r\n| Data lost after terminate | Use persistent filesystems |\r\n| Slow data transfer | Use filesystem in same region |\r\n| GPU not detected | Reboot instance, check drivers |\r\n\r\n## References\r\n\r\n- **[Advanced Usage](references\u002Fadvanced-usage.md)** - Multi-node training, API automation\r\n- **[Troubleshooting](references\u002Ftroubleshooting.md)** - Common issues and solutions\r\n\r\n## Resources\r\n\r\n- **Documentation**: https:\u002F\u002Fdocs.lambda.ai\r\n- **Console**: https:\u002F\u002Fcloud.lambda.ai\r\n- **Pricing**: https:\u002F\u002Flambda.ai\u002Finstances\r\n- **Support**: https:\u002F\u002Fsupport.lambdalabs.com\r\n- **Blog**: https:\u002F\u002Flambda.ai\u002Fblog\r\n","# Lambda Labs GPU Cloud\r\n\r\nComprehensive guide to running ML workloads on Lambda Labs GPU cloud with on-demand instances and 1-Click Clusters.\r\n\r\n## When to use Lambda Labs\r\n\r\n**Use Lambda Labs when:**\r\n- Need dedicated GPU instances with full SSH access\r\n- Running long training jobs (hours to days)\r\n- Want simple pricing with no egress fees\r\n- Need persistent storage across sessions\r\n- Require high-performance multi-node clusters (16-512 GPUs)\r\n- Want pre-installed ML stack (Lambda Stack with PyTorch, CUDA, NCCL)\r\n\r\n**Key features:**\r\n- **GPU variety**: B200, H100, GH200, A100, A10, A6000, V100\r\n- **Lambda Stack**: Pre-installed PyTorch, TensorFlow, CUDA, cuDNN, NCCL\r\n- **Persistent filesystems**: Keep data across instance restarts\r\n- **1-Click Clusters**: 16-512 GPU Slurm clusters with InfiniBand\r\n- **Simple pricing**: Pay-per-minute, no egress fees\r\n- **Global regions**: 12+ regions worldwide\r\n\r\n**Use alternatives instead:**\r\n- **Modal**: For serverless, auto-scaling workloads\r\n- **SkyPilot**: For multi-cloud orchestration and cost optimization\r\n- **RunPod**: For cheaper spot instances and serverless endpoints\r\n- **Vast.ai**: For GPU marketplace with lowest prices\r\n\r\n## Quick start\r\n\r\n### Account setup\r\n\r\n1. Create account at https:\u002F\u002Flambda.ai\r\n2. Add payment method\r\n3. Generate API key from dashboard\r\n4. Add SSH key (required before launching instances)\r\n\r\n### Launch via console\r\n\r\n1. Go to https:\u002F\u002Fcloud.lambda.ai\u002Finstances\r\n2. Click \"Launch instance\"\r\n3. Select GPU type and region\r\n4. Choose SSH key\r\n5. Optionally attach filesystem\r\n6. Launch and wait 3-15 minutes\r\n\r\n### Connect via SSH\r\n\r\n```bash\r\n# Get instance IP from console\r\nssh ubuntu@\u003CINSTANCE-IP>\r\n\r\n# Or with specific key\r\nssh -i ~\u002F.ssh\u002Flambda_key ubuntu@\u003CINSTANCE-IP>\r\n```\r\n\r\n## GPU instances\r\n\r\n### Available GPUs\r\n\r\n| GPU | VRAM | Price\u002FGPU\u002Fhr | Best For |\r\n|-----|------|--------------|----------|\r\n| B200 SXM6 | 180 GB | $4.99 | Largest models, fastest training |\r\n| H100 SXM | 80 GB | $2.99-3.29 | Large model training |\r\n| H100 PCIe | 80 GB | $2.49 | Cost-effective H100 |\r\n| GH200 | 96 GB | $1.49 | Single-GPU large models |\r\n| A100 80GB | 80 GB | $1.79 | Production training |\r\n| A100 40GB | 40 GB | $1.29 | Standard training |\r\n| A10 | 24 GB | $0.75 | Inference, fine-tuning |\r\n| A6000 | 48 GB | $0.80 | Good VRAM\u002Fprice ratio |\r\n| V100 | 16 GB | $0.55 | Budget training |\r\n\r\n### Instance configurations\r\n\r\n```\r\n8x GPU: Best for distributed training (DDP, FSDP)\r\n4x GPU: Large models, multi-GPU training\r\n2x GPU: Medium workloads\r\n1x GPU: Fine-tuning, inference, development\r\n```\r\n\r\n### Launch times\r\n\r\n- Single-GPU: 3-5 minutes\r\n- Multi-GPU: 10-15 minutes\r\n\r\n## Lambda Stack\r\n\r\nAll instances come with Lambda Stack pre-installed:\r\n\r\n```bash\r\n# Included software\r\n- Ubuntu 22.04 LTS\r\n- NVIDIA drivers (latest)\r\n- CUDA 12.x\r\n- cuDNN 8.x\r\n- NCCL (for multi-GPU)\r\n- PyTorch (latest)\r\n- TensorFlow (latest)\r\n- JAX\r\n- JupyterLab\r\n```\r\n\r\n### Verify installation\r\n\r\n```bash\r\n# Check GPU\r\nnvidia-smi\r\n\r\n# Check PyTorch\r\npython -c \"import torch; print(torch.cuda.is_available())\"\r\n\r\n# Check CUDA version\r\nnvcc --version\r\n```\r\n\r\n## Python API\r\n\r\n### Installation\r\n\r\n```bash\r\npip install lambda-cloud-client\r\n```\r\n\r\n### Authentication\r\n\r\n```python\r\nimport os\r\nimport lambda_cloud_client\r\n\r\n# Configure with API key\r\nconfiguration = lambda_cloud_client.Configuration(\r\n    host=\"https:\u002F\u002Fcloud.lambdalabs.com\u002Fapi\u002Fv1\",\r\n    access_token=os.environ[\"LAMBDA_API_KEY\"]\r\n)\r\n```\r\n\r\n### List available instances\r\n\r\n```python\r\nwith lambda_cloud_client.ApiClient(configuration) as api_client:\r\n    api = lambda_cloud_client.DefaultApi(api_client)\r\n\r\n    # Get available instance types\r\n    types = api.instance_types()\r\n    for name, info in types.data.items():\r\n        print(f\"{name}: {info.instance_type.description}\")\r\n```\r\n\r\n### Launch instance\r\n\r\n```python\r\nfrom lambda_cloud_client.models import LaunchInstanceRequest\r\n\r\nrequest = LaunchInstanceRequest(\r\n    region_name=\"us-west-1\",\r\n    instance_type_name=\"gpu_1x_h100_sxm5\",\r\n    ssh_key_names=[\"my-ssh-key\"],\r\n    file_system_names=[\"my-filesystem\"],  # Optional\r\n    name=\"training-job\"\r\n)\r\n\r\nresponse = api.launch_instance(request)\r\ninstance_id = response.data.instance_ids[0]\r\nprint(f\"Launched: {instance_id}\")\r\n```\r\n\r\n### List running instances\r\n\r\n```python\r\ninstances = api.list_instances()\r\nfor instance in instances.data:\r\n    print(f\"{instance.name}: {instance.ip} ({instance.status})\")\r\n```\r\n\r\n### Terminate instance\r\n\r\n```python\r\nfrom lambda_cloud_client.models import TerminateInstanceRequest\r\n\r\nrequest = TerminateInstanceRequest(\r\n    instance_ids=[instance_id]\r\n)\r\napi.terminate_instance(request)\r\n```\r\n\r\n### SSH key management\r\n\r\n```python\r\nfrom lambda_cloud_client.models import AddSshKeyRequest\r\n\r\n# Add SSH key\r\nrequest = AddSshKeyRequest(\r\n    name=\"my-key\",\r\n    public_key=\"ssh-rsa AAAA...\"\r\n)\r\napi.add_ssh_key(request)\r\n\r\n# List keys\r\nkeys = api.list_ssh_keys()\r\n\r\n# Delete key\r\napi.delete_ssh_key(key_id)\r\n```\r\n\r\n## CLI with curl\r\n\r\n### List instance types\r\n\r\n```bash\r\ncurl -u $LAMBDA_API_KEY: \\\r\n  https:\u002F\u002Fcloud.lambdalabs.com\u002Fapi\u002Fv1\u002Finstance-types | jq\r\n```\r\n\r\n### Launch instance\r\n\r\n```bash\r\ncurl -u $LAMBDA_API_KEY: \\\r\n  -X POST https:\u002F\u002Fcloud.lambdalabs.com\u002Fapi\u002Fv1\u002Finstance-operations\u002Flaunch \\\r\n  -H \"Content-Type: application\u002Fjson\" \\\r\n  -d '{\r\n    \"region_name\": \"us-west-1\",\r\n    \"instance_type_name\": \"gpu_1x_h100_sxm5\",\r\n    \"ssh_key_names\": [\"my-key\"]\r\n  }' | jq\r\n```\r\n\r\n### Terminate instance\r\n\r\n```bash\r\ncurl -u $LAMBDA_API_KEY: \\\r\n  -X POST https:\u002F\u002Fcloud.lambdalabs.com\u002Fapi\u002Fv1\u002Finstance-operations\u002Fterminate \\\r\n  -H \"Content-Type: application\u002Fjson\" \\\r\n  -d '{\"instance_ids\": [\"\u003CINSTANCE-ID>\"]}' | jq\r\n```\r\n\r\n## Persistent storage\r\n\r\n### Filesystems\r\n\r\nFilesystems persist data across instance restarts:\r\n\r\n```bash\r\n# Mount location\r\n\u002Flambda\u002Fnfs\u002F\u003CFILESYSTEM_NAME>\r\n\r\n# Example: save checkpoints\r\npython train.py --checkpoint-dir \u002Flambda\u002Fnfs\u002Fmy-storage\u002Fcheckpoints\r\n```\r\n\r\n### Create filesystem\r\n\r\n1. Go to Storage in Lambda console\r\n2. Click \"Create filesystem\"\r\n3. Select region (must match instance region)\r\n4. Name and create\r\n\r\n### Attach to instance\r\n\r\nFilesystems must be attached at instance launch time:\r\n- Via console: Select filesystem when launching\r\n- Via API: Include `file_system_names` in launch request\r\n\r\n### Best practices\r\n\r\n```bash\r\n# Store on filesystem (persists)\r\n\u002Flambda\u002Fnfs\u002Fstorage\u002F\r\n  ├── datasets\u002F\r\n  ├── checkpoints\u002F\r\n  ├── models\u002F\r\n  └── outputs\u002F\r\n\r\n# Local SSD (faster, ephemeral)\r\n\u002Fhome\u002Fubuntu\u002F\r\n  └── working\u002F  # Temporary files\r\n```\r\n\r\n## SSH configuration\r\n\r\n### Add SSH key\r\n\r\n```bash\r\n# Generate key locally\r\nssh-keygen -t ed25519 -f ~\u002F.ssh\u002Flambda_key\r\n\r\n# Add public key to Lambda console\r\n# Or via API\r\n```\r\n\r\n### Multiple keys\r\n\r\n```bash\r\n# On instance, add more keys\r\necho 'ssh-rsa AAAA...' >> ~\u002F.ssh\u002Fauthorized_keys\r\n```\r\n\r\n### Import from GitHub\r\n\r\n```bash\r\n# On instance\r\nssh-import-id gh:username\r\n```\r\n\r\n### SSH tunneling\r\n\r\n```bash\r\n# Forward Jupyter\r\nssh -L 8888:localhost:8888 ubuntu@\u003CIP>\r\n\r\n# Forward TensorBoard\r\nssh -L 6006:localhost:6006 ubuntu@\u003CIP>\r\n\r\n# Multiple ports\r\nssh -L 8888:localhost:8888 -L 6006:localhost:6006 ubuntu@\u003CIP>\r\n```\r\n\r\n## JupyterLab\r\n\r\n### Launch from console\r\n\r\n1. Go to Instances page\r\n2. Click \"Launch\" in Cloud IDE column\r\n3. JupyterLab opens in browser\r\n\r\n### Manual access\r\n\r\n```bash\r\n# On instance\r\njupyter lab --ip=0.0.0.0 --port=8888\r\n\r\n# From local machine with tunnel\r\nssh -L 8888:localhost:8888 ubuntu@\u003CIP>\r\n# Open http:\u002F\u002Flocalhost:8888\r\n```\r\n\r\n## Training workflows\r\n\r\n### Single-GPU training\r\n\r\n```bash\r\n# SSH to instance\r\nssh ubuntu@\u003CIP>\r\n\r\n# Clone repo\r\ngit clone https:\u002F\u002Fgithub.com\u002Fuser\u002Fproject\r\ncd project\r\n\r\n# Install dependencies\r\npip install -r requirements.txt\r\n\r\n# Train\r\npython train.py --epochs 100 --checkpoint-dir \u002Flambda\u002Fnfs\u002Fstorage\u002Fcheckpoints\r\n```\r\n\r\n### Multi-GPU training (single node)\r\n\r\n```python\r\n# train_ddp.py\r\nimport torch\r\nimport torch.distributed as dist\r\nfrom torch.nn.parallel import DistributedDataParallel as DDP\r\n\r\ndef main():\r\n    dist.init_process_group(\"nccl\")\r\n    rank = dist.get_rank()\r\n    device = rank % torch.cuda.device_count()\r\n\r\n    model = MyModel().to(device)\r\n    model = DDP(model, device_ids=[device])\r\n\r\n    # Training loop...\r\n\r\nif __name__ == \"__main__\":\r\n    main()\r\n```\r\n\r\n```bash\r\n# Launch with torchrun (8 GPUs)\r\ntorchrun --nproc_per_node=8 train_ddp.py\r\n```\r\n\r\n### Checkpoint to filesystem\r\n\r\n```python\r\nimport os\r\n\r\ncheckpoint_dir = \"\u002Flambda\u002Fnfs\u002Fmy-storage\u002Fcheckpoints\"\r\nos.makedirs(checkpoint_dir, exist_ok=True)\r\n\r\n# Save checkpoint\r\ntorch.save({\r\n    'epoch': epoch,\r\n    'model_state_dict': model.state_dict(),\r\n    'optimizer_state_dict': optimizer.state_dict(),\r\n    'loss': loss,\r\n}, f\"{checkpoint_dir}\u002Fcheckpoint_{epoch}.pt\")\r\n```\r\n\r\n## 1-Click Clusters\r\n\r\n### Overview\r\n\r\nHigh-performance Slurm clusters with:\r\n- 16-512 NVIDIA H100 or B200 GPUs\r\n- NVIDIA Quantum-2 400 Gb\u002Fs InfiniBand\r\n- GPUDirect RDMA at 3200 Gb\u002Fs\r\n- Pre-installed distributed ML stack\r\n\r\n### Included software\r\n\r\n- Ubuntu 22.04 LTS + Lambda Stack\r\n- NCCL, Open MPI\r\n- PyTorch with DDP and FSDP\r\n- TensorFlow\r\n- OFED drivers\r\n\r\n### Storage\r\n\r\n- 24 TB NVMe per compute node (ephemeral)\r\n- Lambda filesystems for persistent data\r\n\r\n### Multi-node training\r\n\r\n```bash\r\n# On Slurm cluster\r\nsrun --nodes=4 --ntasks-per-node=8 --gpus-per-node=8 \\\r\n  torchrun --nnodes=4 --nproc_per_node=8 \\\r\n  --rdzv_backend=c10d --rdzv_endpoint=$MASTER_ADDR:29500 \\\r\n  train.py\r\n```\r\n\r\n## Networking\r\n\r\n### Bandwidth\r\n\r\n- Inter-instance (same region): up to 200 Gbps\r\n- Internet outbound: 20 Gbps max\r\n\r\n### Firewall\r\n\r\n- Default: Only port 22 (SSH) open\r\n- Configure additional ports in Lambda console\r\n- ICMP traffic allowed by default\r\n\r\n### Private IPs\r\n\r\n```bash\r\n# Find private IP\r\nip addr show | grep 'inet '\r\n```\r\n\r\n## Common workflows\r\n\r\n### Workflow 1: Fine-tuning LLM\r\n\r\n```bash\r\n# 1. Launch 8x H100 instance with filesystem\r\n\r\n# 2. SSH and setup\r\nssh ubuntu@\u003CIP>\r\npip install transformers accelerate peft\r\n\r\n# 3. Download model to filesystem\r\npython -c \"\r\nfrom transformers import AutoModelForCausalLM\r\nmodel = AutoModelForCausalLM.from_pretrained('meta-llama\u002FLlama-2-7b-hf')\r\nmodel.save_pretrained('\u002Flambda\u002Fnfs\u002Fstorage\u002Fmodels\u002Fllama-2-7b')\r\n\"\r\n\r\n# 4. Fine-tune with checkpoints on filesystem\r\naccelerate launch --num_processes 8 train.py \\\r\n  --model_path \u002Flambda\u002Fnfs\u002Fstorage\u002Fmodels\u002Fllama-2-7b \\\r\n  --output_dir \u002Flambda\u002Fnfs\u002Fstorage\u002Foutputs \\\r\n  --checkpoint_dir \u002Flambda\u002Fnfs\u002Fstorage\u002Fcheckpoints\r\n```\r\n\r\n### Workflow 2: Batch inference\r\n\r\n```bash\r\n# 1. Launch A10 instance (cost-effective for inference)\r\n\r\n# 2. Run inference\r\npython inference.py \\\r\n  --model \u002Flambda\u002Fnfs\u002Fstorage\u002Fmodels\u002Ffine-tuned \\\r\n  --input \u002Flambda\u002Fnfs\u002Fstorage\u002Fdata\u002Finputs.jsonl \\\r\n  --output \u002Flambda\u002Fnfs\u002Fstorage\u002Fdata\u002Foutputs.jsonl\r\n```\r\n\r\n## Cost optimization\r\n\r\n### Choose right GPU\r\n\r\n| Task | Recommended GPU |\r\n|------|-----------------|\r\n| LLM fine-tuning (7B) | A100 40GB |\r\n| LLM fine-tuning (70B) | 8x H100 |\r\n| Inference | A10, A6000 |\r\n| Development | V100, A10 |\r\n| Maximum performance | B200 |\r\n\r\n### Reduce costs\r\n\r\n1. **Use filesystems**: Avoid re-downloading data\r\n2. **Checkpoint frequently**: Resume interrupted training\r\n3. **Right-size**: Don't over-provision GPUs\r\n4. **Terminate idle**: No auto-stop, manually terminate\r\n\r\n### Monitor usage\r\n\r\n- Dashboard shows real-time GPU utilization\r\n- API for programmatic monitoring\r\n\r\n## Common issues\r\n\r\n| Issue | Solution |\r\n|-------|----------|\r\n| Instance won't launch | Check region availability, try different GPU |\r\n| SSH connection refused | Wait for instance to initialize (3-15 min) |\r\n| Data lost after terminate | Use persistent filesystems |\r\n| Slow data transfer | Use filesystem in same region |\r\n| GPU not detected | Reboot instance, check drivers |\r\n\r\n## References\r\n\r\n- **[Advanced Usage](references\u002Fadvanced-usage.md)** - Multi-node training, API automation\r\n- **[Troubleshooting](references\u002Ftroubleshooting.md)** - Common issues and solutions\r\n\r\n## Resources\r\n\r\n- **Documentation**: https:\u002F\u002Fdocs.lambda.ai\r\n- **Console**: https:\u002F\u002Fcloud.lambda.ai\r\n- **Pricing**: https:\u002F\u002Flambda.ai\u002Finstances\r\n- **Support**: https:\u002F\u002Fsupport.lambdalabs.com\r\n- **Blog**: https:\u002F\u002Flambda.ai\u002Fblog\r\n",{"name":5,"author":44,"license":15,"version":45,"metadata":46,"description":7,"dependencies":52},"Orchestra Research","1.0.0",{"hermes":47},{"tags":48},[49,21,50,51,22],"Infrastructure","Training","Inference",[53],"lambda-cloud-client>=1.0.0","main","0d1434f89f4bc6b07e9d272accd8c1013768820f","archive\u002Fskill-md-reviewed\u002Fsha256\u002Fef\u002Fef10cb06a745504fc8997e9a9d6fdb03774ea0cef36ff2994279afa65b55cda6",0,"https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills","https:\u002F\u002Fgithub.com\u002Fzapabob","The Open University of JAPAN　2019-now\r\n\r\nTokyo University of Science　2007-2010","2026-10-26 16:29:01",{"source":8,"skillName":5,"commands":63,"prompt":73,"zipAvailable":74},[64,67,70],{"tool":65,"command":66},"npx","npx skills add zapabob\u002FSkills --skill lambda-labs-gpu-cloud",{"tool":68,"command":69},"bunx","bunx skills add zapabob\u002FSkills --skill lambda-labs-gpu-cloud",{"tool":71,"command":72},"pnpm","pnpm dlx skills add zapabob\u002FSkills --skill lambda-labs-gpu-cloud","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Farchive\u002Fskill-md-reviewed\u002Fsha256\u002Fef\u002Fef10cb06a745504fc8997e9a9d6fdb03774ea0cef36ff2994279afa65b55cda6\u002FSKILL.md",true,[],[77,115,150,176,203,219],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":96,"publishedAt":39},"01M6GJTE1GESGGF9K028Y7753E","xurl","X\u002FTwitter via xurl CLI: post, search, DM, media, v2 API.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fxurl\u002Fvariants\u002Fhermes-agent\u002FSKILL.md",[83,88,91],{"categoryNo":84,"categoryName":85,"taxonomyType":86,"taxonomyTypeDesc":87},"01JBSEEDCAT000000000000010","后端与接口","SCENARIO","使用场景",{"categoryNo":89,"categoryName":90,"taxonomyType":86,"taxonomyTypeDesc":87},"01JBSEEDCAT000000000000023","Skill与智能体开发",{"categoryNo":92,"categoryName":93,"taxonomyType":94,"taxonomyTypeDesc":95},"01JBSEEDCAT000000000000030","开发工程师","PROFESSION","适用职业",[79,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,54,112,113,114],"openclaw","macos","cli","github","linux","windows","official","search","https","social-media","api","Media","hermes","hermes agent","twitter","social","via","post",{"skillNo":116,"name":117,"nameZh":6,"description":118,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":119,"categories":120,"keywords":132,"publishedAt":39},"01M6GJT8S1PXQ8WPXT1HSNBWH2","writing-plans","Write implementation plans: bite-sized tasks, paths, code.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fwriting-plans\u002Fvariants\u002Fhermes-agent\u002FSKILL.md",[121,124,127,128,129],{"categoryNo":122,"categoryName":123,"taxonomyType":86,"taxonomyTypeDesc":87},"01JBSEEDCAT000000000000015","视觉设计",{"categoryNo":125,"categoryName":126,"taxonomyType":86,"taxonomyTypeDesc":87},"01JBSEEDCAT000000000000018","写作与文案",{"categoryNo":89,"categoryName":90,"taxonomyType":86,"taxonomyTypeDesc":87},{"categoryNo":92,"categoryName":93,"taxonomyType":94,"taxonomyTypeDesc":95},{"categoryNo":130,"categoryName":131,"taxonomyType":94,"taxonomyTypeDesc":95},"01JBSEEDCAT000000000000036","内容创作者",[133,134,135,136,98,137,101,138,102,139,140,141,142,143,144,145,146,147,148,149],"writing","writing plans","plans","design","review","workflow","documentation","tasks","development","code","planning","code-review","subagent","write","superpowers","paths","test-driven-development",{"skillNo":151,"name":152,"nameZh":6,"description":153,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":154,"categories":155,"keywords":159,"publishedAt":39},"01M6GJST0H5X7WHBJ3BZK0DVKH","whisper","OpenAI's general-purpose speech recognition model. Supports 99 languages, transcription, translation to English, and language identification. Six model sizes from tiny (39M params) to large (1550M params). Use for speech-to-text, podcast transcription, or multilingual audio processing. Best for robust, multilingual ASR.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fwhisper\u002Fvariants\u002Fhermes-agent-optional\u002FSKILL.md",[156],{"categoryNo":157,"categoryName":158,"taxonomyType":86,"taxonomyTypeDesc":87},"01JBSEEDCAT000000000000016","图像与多媒体",[152,98,101,160,161,162,163,164,165,166,167,168,32,169,170,171,172,173,174,175],"openai","speech","multimodal","Transformers","ASR","Audio","text","model","speech recognition","general","multilingual","translation","english","identification","processing","Audio Processing",{"skillNo":177,"name":178,"nameZh":6,"description":179,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":180,"categories":181,"keywords":185,"publishedAt":39},"01M6GJSMRQ8EETZTYMG72P70EA","weights-and-biases","W&B: log ML experiments, sweeps, model registry, dashboards.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fweights-and-biases\u002Fvariants\u002Fhermes-agent\u002FSKILL.md",[182],{"categoryNo":183,"categoryName":184,"taxonomyType":86,"taxonomyTypeDesc":87},"01JBSEEDCAT000000000000013","分析与可视化",[186,98,187,188,102,189,190,191,192,193,194,167,195,196,197,198,199,200,201,202],"weights and biases","tensorflow","experiments","pytorch","visualization","registry","real-time","tuning","dashboards","experiment","log","time","Experiment Tracking","Model Registry","Hyperparameter Tuning","WandB","Real-Time Visualization",{"skillNo":204,"name":205,"nameZh":6,"description":206,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":207,"categories":208,"keywords":210,"publishedAt":39},"01M6GJS19KDPRY112JTR326S06","watchers","Poll RSS, JSON APIs, and GitHub with watermark dedup.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fwatchers\u002Fvariants\u002Fhermes-agent-optional\u002FSKILL.md",[209],{"categoryNo":89,"categoryName":90,"taxonomyType":86,"taxonomyTypeDesc":87},[205,98,211,100,101,212,213,214,215,216,217,109,110,218,27],"terminal","automation","http","monitoring","cron","json","rss","devops",{"skillNo":220,"name":221,"nameZh":6,"description":222,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":223,"categories":224,"keywords":227,"publishedAt":39},"01M6GJRHJD2A6Q830SWF9W48BC","serving-llms-vllm","vLLM: high-throughput LLM serving, OpenAI API, quantization.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fvllm\u002Fvariants\u002Fhermes-agent\u002FSKILL.md",[225,226],{"categoryNo":84,"categoryName":85,"taxonomyType":86,"taxonomyTypeDesc":87},{"categoryNo":92,"categoryName":93,"taxonomyType":94,"taxonomyTypeDesc":95},[228,229,230,101,231,160,232,233,234,163,235,107,236,237,238,239,240,241,242,243],"llms","vLLM","serving llms vllm","research","openai api","tensor","Inference Serving","Production","llm serving","High Throughput","PagedAttention","Continuous Batching","Quantization","Tensor Parallelism","high-throughput","throughput"]