krzysztof/landing_page
1
0
Fork 0
mirror of https://github.com/qdrant/landing_page.git synced 2026-10-10 21:38:30 +02:00
Code Issues Packages Projects Releases Wiki Activity

Fix SPLADE architecture order in parts 1 & 2

Browse Source 
Log saturation (ReLU + log1p) is applied per-token before max pooling
across positions, not after. Fix text descriptions and pipeline diagram.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
Evgeniya Sukhodolskaya
2026-06-04 11:32:49 +02:00
co-authored by Claude Sonnet 4.6
parent 4a727882dc
commit f4d989b1cb
3 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
qdrant-landing/content/articles/sparse-embeddings-ecommerce-part-2.md
+1 -1
View File
@@ -154,7 +154,7 @@ No S3 uploads, no checkpoint management code, no lost training runs.
Sentence Transformers v5 introduced `SparseEncoder`, making SPLADE training straightforward. The model has two components:
1. **MLMTransformer**: A transformer with a masked language model head that outputs logits over the full vocabulary
2. **SpladePooling**: Max-pools the token-level logits and applies ReLU + log saturation
2. **SpladePooling**: Applies ReLU + log saturation to the token-level logits and max-pools across positions
```python
from sentence_transformers import SparseEncoder