hyperpod-version-checker

Name: hyperpod-version-checker
Author: awslabs

843stars132forks

How to install hyperpod-version-checker?

awslabs/agent-plugins/hyperpod-version-checker

$npx -y skills add awslabs/agent-plugins --skill hyperpod-version-checker

Installs into the current project.

Summary

Check and compare software component versions on SageMaker HyperPod cluster nodes - NVIDIA drivers, CUDA toolkit, cuDNN, NCCL, EFA, AWS OFI NCCL, GDRCopy, MPI, Neuron SDK (Trainium/Inferentia), Python, and PyTorch. Use when checking component versions, verifying CUDA/driver compa

Files · 1

View on GitHub

SKILL.md

1	# HyperPod Version Checker
2
3	Upload to cluster nodes via `hyperpod-ssm` skill, then execute.
4
5	## Usage
6
7	```bash
8	# Text report to console + file
9	bash hyperpod_check_versions.sh
10
11	# JSON only to stdout (text report still saved to file) — best for piping/parsing
12	bash hyperpod_check_versions.sh --json
13
14	# Custom output file
15	bash hyperpod_check_versions.sh --output /tmp/versions.txt
16
17	# No color (for logging)
18	bash hyperpod_check_versions.sh --no-color
19	```
20
21	Output file: `component_versions_<hostname>_<timestamp>.txt` (default)
22
23	## What It Checks
24
25	\| Component \| Detection Method \| Applicable When \|
26	\| ----------------- \| ----------------------------------------------- \| --------------------------------------------- \|
27	\| NVIDIA Driver \| `nvidia-smi` \| GPU instances (p3/p4/p5/g5) \|
28	\| CUDA Toolkit \| `nvcc`, `/usr/local/cuda` symlink \| GPU instances \|
29	\| cuDNN \| Header file, packages \| GPU instances doing deep learning \|
30	\| NCCL \| Library filename, header, packages \| Distributed GPU training \|
31	\| EFA \| `/opt/amazon/efa_installed_packages`, `fi_info` \| EFA-capable instances (p4d/p4de/p5/trn1/trn2) \|
32	\| AWS OFI NCCL \| `efa_installed_packages`, library search \| EFA + NCCL workloads \|
33	\| GDRCopy \| rpm/dpkg, kernel module \| GPU instances with RDMA (p4d+/p5) \|
34	\| MPI \| `mpirun`, `/opt/amazon/openmpi` \| Distributed training \|
35	\| Neuron SDK \| `neuronx-cc`, `neuron-ls`, packages \| Trainium/Inferentia (trn1/trn2/inf1/inf2) \|
36	\| Python/PyTorch \| `python3`, `torch` import \| ML workloads \|
37	\| Container runtime \| `docker`, `containerd`, `kubectl`, `nvidia-ctk` \| EKS clusters \|
38
39	## Multi-Node Comparison
40
41	Run on each node individually via the `hyperpod-ssm` skill. With `--json`, stdout is clean JSON for easy diffing.
42
43	## Compatibility Reference
44
45	The script automatically analyzes CUDA/driver compatibility. For reference:
46
47	\| Driver Series \| Supported CUDA \|
48	\| ------------- \| ----------------------------- \|
49	\| 580+ \| 13.x, 12.x, 11.x \|
50	\| 570+ \| 12.8+ (Blackwell), 12.x, 11.x \|
51	\| 545+ \| 12.3-12.7, 11.x \|
52	\| 525-535 \| 12.0-12.2, 11.x \|
53	\| 450+ \| 11.x only \|
54
55	NCCL: Use 2.18+ for CUDA 12.x, 2.12+ for CUDA 11.x. Must be consistent across all nodes.
56
57	\| EFA Installer \| AWS OFI NCCL \|
58	\| ------------- \| --------------------- \|
59	\| 1.29+ \| v1.7.3+ (recommended) \|
60	\| 1.26-1.28 \| v1.7.0-v1.7.2 \|
61	\| 1.20-1.25 \| v1.6.0+ \|

People who liked this also viewed

6 picks

Type

hyperpod-version-checker

843stars132forks

View on GitHub

How to install hyperpod-version-checker?

awslabs/agent-plugins/hyperpod-version-checker

$npx -y skills add awslabs/agent-plugins --skill hyperpod-version-checker

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# HyperPod Version Checker
2
3	Upload to cluster nodes via `hyperpod-ssm` skill, then execute.
4
5	## Usage
6
7	```bash
8	# Text report to console + file
9	bash hyperpod_check_versions.sh
10
11	# JSON only to stdout (text report still saved to file) — best for piping/parsing
12	bash hyperpod_check_versions.sh --json
13
14	# Custom output file
15	bash hyperpod_check_versions.sh --output /tmp/versions.txt
16
17	# No color (for logging)
18	bash hyperpod_check_versions.sh --no-color
19	```
20
21	Output file: `component_versions_<hostname>_<timestamp>.txt` (default)
22
23	## What It Checks
24
25	\| Component \| Detection Method \| Applicable When \|
26	\| ----------------- \| ----------------------------------------------- \| --------------------------------------------- \|
27	\| NVIDIA Driver \| `nvidia-smi` \| GPU instances (p3/p4/p5/g5) \|
28	\| CUDA Toolkit \| `nvcc`, `/usr/local/cuda` symlink \| GPU instances \|
29	\| cuDNN \| Header file, packages \| GPU instances doing deep learning \|
30	\| NCCL \| Library filename, header, packages \| Distributed GPU training \|
31	\| EFA \| `/opt/amazon/efa_installed_packages`, `fi_info` \| EFA-capable instances (p4d/p4de/p5/trn1/trn2) \|
32	\| AWS OFI NCCL \| `efa_installed_packages`, library search \| EFA + NCCL workloads \|
33	\| GDRCopy \| rpm/dpkg, kernel module \| GPU instances with RDMA (p4d+/p5) \|
34	\| MPI \| `mpirun`, `/opt/amazon/openmpi` \| Distributed training \|
35	\| Neuron SDK \| `neuronx-cc`, `neuron-ls`, packages \| Trainium/Inferentia (trn1/trn2/inf1/inf2) \|
36	\| Python/PyTorch \| `python3`, `torch` import \| ML workloads \|
37	\| Container runtime \| `docker`, `containerd`, `kubectl`, `nvidia-ctk` \| EKS clusters \|
38
39	## Multi-Node Comparison
40
41	Run on each node individually via the `hyperpod-ssm` skill. With `--json`, stdout is clean JSON for easy diffing.
42
43	## Compatibility Reference
44
45	The script automatically analyzes CUDA/driver compatibility. For reference:
46
47	\| Driver Series \| Supported CUDA \|
48	\| ------------- \| ----------------------------- \|
49	\| 580+ \| 13.x, 12.x, 11.x \|
50	\| 570+ \| 12.8+ (Blackwell), 12.x, 11.x \|
51	\| 545+ \| 12.3-12.7, 11.x \|
52	\| 525-535 \| 12.0-12.2, 11.x \|
53	\| 450+ \| 11.x only \|
54
55	NCCL: Use 2.18+ for CUDA 12.x, 2.12+ for CUDA 11.x. Must be consistent across all nodes.
56
57	\| EFA Installer \| AWS OFI NCCL \|
58	\| ------------- \| --------------------- \|
59	\| 1.29+ \| v1.7.3+ (recommended) \|
60	\| 1.26-1.28 \| v1.7.0-v1.7.2 \|
61	\| 1.20-1.25 \| v1.6.0+ \|

People who liked this also viewed

6 picks

Type