Update README.md
Browse files
README.md
CHANGED
@@ -147,6 +147,17 @@ This tiny SPLADE model is `6x` smaller than Naver's official `splade-v3-distilbe
|
|
147 |
- `Distillation Dataset:` https://huggingface.co/datasets/yosefw/msmarco-train-distil-v2
|
148 |
- `Code:` https://github.com/rasyosef/splade-tiny-msmarco
|
149 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
150 |
## Usage
|
151 |
|
152 |
### Direct Usage (Sentence Transformers)
|
|
|
147 |
- `Distillation Dataset:` https://huggingface.co/datasets/yosefw/msmarco-train-distil-v2
|
148 |
- `Code:` https://github.com/rasyosef/splade-tiny-msmarco
|
149 |
|
150 |
+
## Performance
|
151 |
+
|
152 |
+
The splade models were evaluated on 55 thousand queries and 8 million documents from the [MSMARCO](https://huggingface.co/datasets/microsoft/ms_marco) dataset.
|
153 |
+
|
154 |
+
||Size (# Params)|MRR@10 (MS MARCO dev)|
|
155 |
+
|:---|:----|:-------------------|
|
156 |
+
|`BM25`|-|18.6|-|-|
|
157 |
+
|`rasyosef/splade-tiny`|4.4M|30.8|
|
158 |
+
|`rasyosef/splade-mini`|11.2M|32.8|
|
159 |
+
|`naver/splade-v3-distilbert`|67.0M|38.7|
|
160 |
+
|
161 |
## Usage
|
162 |
|
163 |
### Direct Usage (Sentence Transformers)
|