Commit
•
61ab149
1
Parent(s):
47cf1a0
Update README.md
Browse files
README.md
CHANGED
|
@@ -72,7 +72,7 @@ SPLADE models are a fine balance between retrieval effectiveness (quality) and r
|
|
| 72 |
|
| 73 |
**TL;DR of Our attempt & results**
|
| 74 |
1. FLOPS tuning: Seperate **Seq lens and Severely restrictive token budget** doc(128) & query(24) NOT 256 unlike Official SPLADE++. Inspired from **SparseEmbed** (instead of 2 models for query & doc).
|
| 75 |
-
3. Init Weights: **bert-base-uncased**.
|
| 76 |
4. Achieves a modest yet competitive effectiveness **MRR@10 37.22** in ID data (& OOD) and a retrieval latency of - **47.27ms**. (multi-threaded) all On **Consumer grade-GPUs** with **only 5 negatives per query**.
|
| 77 |
4. For Industry setting: Effectiveness on custom domains needs more than just **Trading FLOPS for tiny gains** and The Premise "SPLADE++ are not well suited to mono-cpu retrieval" does not hold.
|
| 78 |
|
|
|
|
| 72 |
|
| 73 |
**TL;DR of Our attempt & results**
|
| 74 |
1. FLOPS tuning: Seperate **Seq lens and Severely restrictive token budget** doc(128) & query(24) NOT 256 unlike Official SPLADE++. Inspired from **SparseEmbed** (instead of 2 models for query & doc).
|
| 75 |
+
3. Init Weights: Vanilla **bert-base-uncased**. No corpus awarness unlike Official splade++ / ColBERT
|
| 76 |
4. Achieves a modest yet competitive effectiveness **MRR@10 37.22** in ID data (& OOD) and a retrieval latency of - **47.27ms**. (multi-threaded) all On **Consumer grade-GPUs** with **only 5 negatives per query**.
|
| 77 |
4. For Industry setting: Effectiveness on custom domains needs more than just **Trading FLOPS for tiny gains** and The Premise "SPLADE++ are not well suited to mono-cpu retrieval" does not hold.
|
| 78 |
|