Feature Extraction
OpenCLIP
Safetensors
English
clap2
audio
clap
contrastive
audio-text-retrieval
zero-shot-audio-classification
Instructions to use laion/clap2-xs-audioxl-s1B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- OpenCLIP
How to use laion/clap2-xs-audioxl-s1B with OpenCLIP:
import open_clip model, preprocess_train, preprocess_val = open_clip.create_model_and_transforms('hf-hub:laion/clap2-xs-audioxl-s1B') tokenizer = open_clip.get_tokenizer('hf-hub:laion/clap2-xs-audioxl-s1B') - Notebooks
- Google Colab
- Kaggle
Commit History
7 days ago
-
Transformers weights: MLP parameters renamed to Wi/Wo, config field hidden_activation
Verified signatureSame weights as before under the names the Transformers CLAP2 implementation uses (its MLP is ModernBERT's). The OpenCLIP files are unchanged.4ad2cb1
9 days ago
-
Fix writing
Verified signature76ef521 -
Remove the Transformers usage example (CLAP2 support in Transformers is not released yet)
Verified signature0cccce1 -
Add arXiv paper link and citation
Verified signaturecaf8cbd
13 days ago
-
Model card: install open_clip_torch>=4.2.1 (released)
Verified signatureadc5107 -
Card: remove two limitation notes
Verified signature5ce3d4f -
Model card: simpler OpenCLIP quick start (PyPI >=4.2.1.dev0, load_audio, sentence prompts, probabilities)
Verified signature095608f -
Model card: drop the 30 s note in the intro and the notes on the numbers
Verified signature1f2fb05 -
Model card: link CLIP_benchmark, drop the PR-pending note
Verified signatureaa3c746 -
Model card: Apache 2.0 licence, LAION-Audio-XL numbers from the final corpus (366,006,046 pairs), shorter training-data section
Verified signature41f6cfe
14 days ago
-
Add Transformers weights, config and processor; README: Transformers usage, prompts
Verified signature315b9b7
24 days ago
-
Bundle the r50k_base tokenizer files (OpenCLIP push_to_hf_hub layout)
Verified signaturef04445e -
Use OpenCLIP's default r50k_base tokenizer instead of a bundled copy
Verified signature7fc7ba8 -
Set audio_seq_len to 1504 (30 s, the training maximum)
Verified signature3e610ab -
Update repository names
Verified signatureaa5e4eb -
Add model
Verified signature48695dd -
Add model card
Verified signature768f664 -
initial commit
Verified signature3d0c6f4