Nemo mbridge multi node slurm vs Physicsnemo shard tensor
Nemo mbridge multi node slurm is an open-source agent skill for developer tools. Physicsnemo shard tensor is an open-source agent skill for developer tools.
- Nemo mbridge multi node slurm is best for convert a single-node training script into a multi-node sbatch job.
- Physicsnemo shard tensor is best for add domain parallelism to an existing physics training script.
- What it is
- Nemo mbridge multi node slurm is an open-source agent skill for developer tools.Physicsnemo shard tensor is an open-source agent skill for developer tools.
- Best for
- Convert a single-node training script into a multi-node sbatch jobAdd domain parallelism to an existing physics training script
- Kind
- SkillSkill
- Pricing
- Open sourceOpen source
- Platforms
- Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf, GooseClaude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf, Goose
- GitHub stars
- 4K4K
- Last commit
- Nov 2026Nov 2026
- License
- Apache-2.0Apache-2.0
- Open source
- Good for
- Convert a single-node training script into a multi-node sbatch job; Debug an NCCL timeout that appears on nodes after the first; Size expert parallelism for a large mixture-of-experts modelAdd domain parallelism to an existing physics training script; Scatter a spatial input tensor across GPUs for inference; Write a shard patch so a new layer works on ShardTensor inputs
What Nemo mbridge multi node slurm does
Direct srun multi-node launch
Explains the preferred launch shape where Slurm spawns eight tasks per node and Bridge derives distributed environment variables.
Legacy torchrun launch path
Covers converting scripts that rely on torch.distributed.run, including one task per node and the TRAIN_CMD escaping rules.
NCCL timeout triage
Lists first-log checks and likely causes for NCCL timeouts at barriers, with fixes such as longer timeouts.
MoE memory sizing
Shows how expert parallelism splits expert memory across GPUs and gives rules of thumb for roundtrip, inference and training.
Interactive allocation workflow
Walks through salloc, container shell setup and common partition and flag errors for ad-hoc testing.
Shared cache configuration
Describes placing uv, Hugging Face and NeMo caches on shared storage so all nodes see prepared data.
What Physicsnemo shard tensor does
Domain parallel script setup
Covers building a DeviceMesh with ddp and domain axes and scattering inputs and targets over the domain group.
DDP and FSDP2 wrapper choice
Explains which data-parallel wrapper fits plain versus sharded parameters and how to sync weights across the domain axis.
Shard patch authoring
Describes when to write a patch for a new layer or op and how to register it at import time.
Multi-GPU correctness testing
Shows how to compare distributed outputs and gradients against a single-GPU run using torchrun and pytest markers.
torch.compile guidance
Lists regional compilation rules for sharded attention and dynamic shape settings.
Debugging pitfalls
Collects known failure modes such as misleading operator errors, in-place grad flags and unwaited collectives.
Nemo mbridge multi node slurm vs Physicsnemo shard tensor: questions
Is Nemo mbridge multi node slurm or Physicsnemo shard tensor cheaper?
Both are open source. Compare current plans on each official site.
What is the difference between Nemo mbridge multi node slurm and Physicsnemo shard tensor?
Nemo mbridge multi node slurm is an open-source agent skill for developer tools. Physicsnemo shard tensor is an open-source agent skill for developer tools.
