public-safe Runpod GPU server snapshot after initial setup
TL;DR
This is a public-safe, one-time server snapshot collected from inside the current Runpod GPU server after initial setup. It records automatically verifiable hardware, storage, OS, CUDA, Python, PyTorch, and key ML package information. Pricing, billing metadata, region, template name, SSH details, public IP addresses, credentials, project repositories, checkpoints, datasets, and experiment results are intentionally excluded.
Hardware Summary
Item
Value
Source
GPU count
1
auto-detected
GPU model
NVIDIA L40S
auto-detected
VRAM per GPU
46068 MiB
auto-detected
NVIDIA Driver
580.159.03
auto-detected
CUDA Version from nvidia-smi
13.0
auto-detected
CPU model
AMD EPYC 9554 64-Core Processor
auto-detected
vCPU count
256
auto-detected
System RAM
1.0Ti
auto-detected
/workspace storage
378T total, 125T available
auto-detected
GPU Details
GPU
Model
VRAM
Driver
CUDA from nvidia-smi
Power Limit
PCI Bus ID
0
NVIDIA L40S
46068 MiB
580.159.03
13.0
262.00 W
00000000:C3:00.0
CPU Details
Item
Value
Architecture
x86_64
CPU model
AMD EPYC 9554 64-Core Processor
vCPU count
256
Sockets
2
Cores per socket
64
Threads per core
2
CPU max MHz
3764.2019 MHz
Memory
Item
Value
Total RAM
1.0Ti
Used RAM
40Gi
Available RAM
966Gi
Storage
Mount
Size
Used
Available
Use%
Filesystem
/
100G
1.2G
99G
2%
overlay
/workspace
378T
253T
125T
67%
fuse
/dev/shm
58G
0
58G
0%
tmpfs
Operating System
Item
Value
OS
Ubuntu 24.04.3 LTS
OS ID
ubuntu
OS version
24.04
Kernel release
6.17.0-29-generic
Machine architecture
x86_64
Python Environment
Item
Value
Python
CPython 3.12.3
Python executable
/usr/local/bin/python
pip
25.2
CUDA / PyTorch Runtime
Item
Value
PyTorch
2.8.0+cu128
CUDA available in PyTorch
True
PyTorch CUDA version
12.8
cuDNN
91002
PyTorch CUDA device count
1
NVCC
release 12.8, build V12.8.93
PyTorch CUDA Devices
Device
Name
Total Memory
Compute Capability
0
NVIDIA L40S
44.39 GiB
8.9
Key Package Versions
Package
Version
torch
2.8.0+cu128
torchvision
0.23.0+cu128
torchaudio
2.8.0+cu128
numpy
2.1.2
scipy
not installed
pandas
not installed
matplotlib
not installed
opencv / cv2
not installed
diffusers
not installed
transformers
not installed
accelerate
not installed
tokenizers
not installed
einops
not installed
flash_attn
not installed
huggingface_hub
not installed
safetensors
not installed
xformers
not installed
Publication Notes
This manifest includes only information automatically detected from inside the server/container.
This manifest is intended to be safe for public blog publication.
Pricing, billing type, region, data center, template name, public IP address, SSH port, credentials, API tokens, project repositories, checkpoints, datasets, and experiment results are intentionally excluded.
Project-specific reproducibility metadata should be documented separately in each project report.
