Usage¶
Getting started¶
ai-suggest — Generate a config from plain English¶
Describe your fine-tuning task and get a ready-to-run YAML config:
xlmtec ai-suggest "fine-tune GPT-2 for sentiment analysis" --provider claude
xlmtec ai-suggest "qlora on llama for code generation" --provider gemini
xlmtec ai-suggest "instruction tune for customer support" --provider codex --save config.yaml
Options:
| Option | Default | Description |
|---|---|---|
--provider | claude | AI provider: claude, gemini, codex |
--api-key | env var | API key (falls back to env variable) |
--save | — | Save generated YAML to a file |
See AI Integrations for full provider setup.
hub — Browse HuggingFace models¶
Search¶
xlmtec hub search "bert"
xlmtec hub search "llama" --task text-generation --limit 10
xlmtec hub search "gpt" --sort likes
Options:
| Option | Default | Description |
|---|---|---|
--task | — | Filter by pipeline tag |
--limit | 10 | Number of results (max 100) |
--sort | downloads | Sort by: downloads, likes, lastModified |
Info¶
Trending¶
See Model Hub for more.
config validate — Check a config before training¶
xlmtec config validate config.yaml
xlmtec config validate config.yaml --strict # fail on warnings too
Reports all validation errors at once. Exits 0 on success, 1 on any error.
train — Fine-tune a model¶
Dry run (no model loading)¶
Validates the config and prints the full training plan without loading any model.