Train 400x faster Static Embedding Models with Sentence Transformers
ā¢
119
Tom Aarsen
tomaarsen
AI & ML interests
NLP: text embeddings, information retrieval, named entity recognition, few-shot text classification
Recent Activity
liked
a model
about 6 hours ago
HuggingFaceTB/SmolVLM-256M-Instruct
posted
an
update
about 7 hours ago
I just released Sentence Transformers v3.4.0, featuring a memory leak fix, compatibility between the powerful Cached... losses and the Matryoshka loss modifier, and a bunch of fixes & small features.
šŖ Matryoshka & Cached loss compatibility
It is now possible to combine the powerful Cached... losses (which use in-batch negatives & a caching mechanism to allow for endless batch size & negatives) with the Matryoshka loss modifier which modifies a base loss such that it is trained not only on the maximum dimensionality (e.g. 1024 dimensions), but also on many lower dimensions (e.g. 768, 512, 256, 128, 64, 32).
After training, these models' embeddings can be truncated for faster retrieval, etc.
šļø Resolve memory leak when Model and Trainer are reinitialized
Due to a circular dependency between Trainer -> Model -> ModelCardData -> Trainer, deleting both the trainer & model still didn't free up the memory.
This led to a memory leak in scripts where you repeatedly do so.
ā New Features
Many new small features, e.g. multi-GPU support for 'mine_hard_negatives', a 'margin' parameter to TripletEvaluator, and Matthews Correlation Coefficient in the BinaryClassificationEvaluator.
š Bug Fixes
Also a bunch of fixes, for example that subsequent batches were not sorted when using the "no_duplicates" batch sampler. See the release notes for more details.
Full release notes: https://github.com/UKPLab/sentence-transformers/releases/tag/v3.4.0
Big thanks to all community members who assisted in this release. 10 folks with their first contribution this time around!
Articles
Organizations
tomaarsen's activity
Should I be concerned about "UserWarning: TypedStorage is deprecated. " when using the model with python 3.11?
1
#97 opened about 7 hours ago
by
poohlio
Convert git-lfs md, py, json files to normal git files
#8 opened 1 day ago
by
tomaarsen
patch inference on CPU & Windows + Update README snippets
#2 opened 1 day ago
by
tomaarsen
Complete Sentence Transformers integration + patch inference on CPU & Windows
1
#4 opened 1 day ago
by
tomaarsen
Update path to ""
1
#4 opened 2 days ago
by
tomaarsen
Add openvino converted tokenizers
2
#96 opened 2 days ago
by
rhecker
Work towards Sentence Transformers compatibility
1
#3 opened 2 days ago
by
tomaarsen
Add Sentence Transformers integration
2
#7 opened 3 days ago
by
tomaarsen
Fix Sentence Transformers integration; currently uses mean pooling accidentally
#3 opened 3 days ago
by
tomaarsen
Add library_name and pipeline_tag tags
1
#5 opened 4 days ago
by
nielsr
Add pipeline tag, library name and link to paper
1
#2 opened 4 days ago
by
nielsr
Upload ONNX weights
#1 opened 6 days ago
by
Xenova
Upload ONNX weights
#2 opened 7 days ago
by
Xenova
Finetuning snowflake-arctic-embed-l-v2.0 for a specific domain
3
#4 opened about 1 month ago
by
wilfoderek
Quants?
2
#1 opened 7 days ago
by
ctranslate2-4you
Suggestion: Different base model
#1 opened 7 days ago
by
tomaarsen
Performance
1
#1 opened 7 days ago
by
tomaarsen
Restart space to add new models
1
#150 opened 8 days ago
by
zhonghhh
Update images with clearer ones
1
#423 opened 8 days ago
by
tomaarsen
Add more optimized images
#420 opened 10 days ago
by
tomaarsen