Conversation
Signed-off-by: Ohad Mosafi <omosafi@nvidia.com>
|
/nvskills-ci |
 |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
| args.output_dir.mkdir(parents=True, exist_ok=False) | ||
| (args.output_dir / "response.json").write_text(json.dumps(result, indent=2) + "\n", encoding="utf-8") | ||
| for database in dict.fromkeys(args.databases): | ||
| alignment = result["alignments"][database]["a3m"]["alignment"] | ||
| path = args.output_dir / f"{database}.a3m" | ||
| path.write_text(alignment, encoding="utf-8") |
There was a problem hiding this comment.
response.json succeeds but a later A3M write fails, the client exits with an error while leaving incomplete files behind. The next run cannot reuse the directory because --output-dir must not already exist. Consider publishing the directory only after all writes succeed, or removing it on failure, and testing a failure during writing. The distributed copy has the same behavior.
| a3m = formats.get("a3m") if isinstance(formats, dict) else None | ||
| alignment = a3m.get("alignment") if isinstance(a3m, dict) else None | ||
| if not isinstance(alignment, str) or not alignment.lstrip().startswith(">"): | ||
| raise SearchError(f"Hosted MSA returned no A3M alignment for {database}.") | ||
| if not any(line.strip() and not line.startswith((">", "#")) for line in alignment.splitlines()): | ||
| raise SearchError(f"Hosted MSA returned an empty A3M alignment for {database}.") |
There was a problem hiding this comment.
> and has any non-header line; it does not check the returned format or whether the records form usable A3M. A malformed response can therefore be saved as a successful .a3m file and fail later in a structure-prediction tool. Check the format and basic record structure before reporting success. The distributed copy has the same check.
Signed-off-by: nvskills-svc-account <svc-nvskills-signing@nvidia.com>
| ### License/Terms of Use: <br> | ||
| Apache-2.0 AND CC-BY-4.0 <br> | ||
| ## Use Case: <br> | ||
| Developers and engineers use this skill to generate multiple sequence alignments for protein sequences via GPU-accelerated MMSeqs2, supporting hosted NVIDIA API or local Docker NIM deployment for homolog search, paired MSA for complexes, and structural template retrieval. <br> |
There was a problem hiding this comment.
| Developers and engineers use this skill to generate multiple sequence alignments for protein sequences via GPU-accelerated MMSeqs2, supporting hosted NVIDIA API or local Docker NIM deployment for homolog search, paired MSA for complexes, and structural template retrieval. <br> | |
| Developers and engineers use this skill to generate multiple sequence alignments for protein sequences via GPU-accelerated MMSeqs2. Hosted NVIDIA API and local Docker NIM support homolog search and paired MSA for complexes; structural template retrieval requires a local Docker NIM. <br> |
2 participants
No description provided.