Skip to content

Model Hub

The xlmtec hub commands let you search, browse, and inspect HuggingFace models without leaving the terminal.

No API key required — uses the public HuggingFace API.

Find models by keyword with optional filters:

# Basic search
xlmtec hub search "bert"

# Filter by task
xlmtec hub search "bert" --task text-classification

# Sort by likes, limit results
xlmtec hub search "gpt" --sort likes --limit 20

# Find multilingual models
xlmtec hub search "multilingual" --task translation

Options:

Option Default Description
--task, -t — Pipeline tag filter (see common tasks below)
--limit, -n 10 Number of results (1–100)
--sort, -s downloads Sort by: downloads, likes, lastModified

Common task filters:

  • text-classification
  • text-generation
  • token-classification
  • question-answering
  • summarization
  • translation
  • fill-mask
  • text2text-generation

Info

Show full details for a specific model:

xlmtec hub info google/bert-base-uncased
xlmtec hub info mistralai/Mistral-7B-v0.1
xlmtec hub info meta-llama/Llama-2-7b-hf

Output includes:

  • Model ID, author, task, library
  • Download count and likes
  • Model size (MB)
  • Tags and languages
  • Last modified date
  • Direct link to HuggingFace page

Show the most downloaded models right now:

xlmtec hub trending
xlmtec hub trending --limit 20

Typical workflow

# 1. Find candidate models
xlmtec hub search "bert" --task text-classification --limit 5

# 2. Inspect the best candidate
xlmtec hub info google/bert-base-uncased

# 3. Generate a training config for it
xlmtec ai-suggest "fine-tune google/bert-base-uncased for sentiment analysis" --provider claude

# 4. Validate and train
xlmtec config validate config.yaml
xlmtec train --config config.yaml --dry-run
xlmtec train --config config.yaml