Tom Aarsen
commited on
Commit
·
bcbd909
1
Parent(s):
56ccabc
Update README model architecture
Browse files
README.md
CHANGED
@@ -377,8 +377,8 @@ This is a [Asymmetric Inference-free SPLADE Sparse Encoder](https://www.sbert.ne
|
|
377 |
```
|
378 |
SparseEncoder(
|
379 |
(0): Router(
|
380 |
-
(
|
381 |
-
(document_0_MLMTransformer): MLMTransformer({'max_seq_length': 512, 'do_lower_case': False
|
382 |
(document_1_SpladePooling): SpladePooling({'pooling_strategy': 'max', 'activation_function': 'relu', 'word_embedding_dimension': 30522})
|
383 |
)
|
384 |
)
|
@@ -591,7 +591,7 @@ You can finetune this model on your own dataset.
|
|
591 |
- `fp16`: True
|
592 |
- `batch_sampler`: no_duplicates
|
593 |
- `router_mapping`: {'query': 'query', 'answer': 'document'}
|
594 |
-
- `learning_rate_mapping`: {'
|
595 |
|
596 |
#### All Hyperparameters
|
597 |
<details><summary>Click to expand</summary>
|
@@ -710,7 +710,7 @@ You can finetune this model on your own dataset.
|
|
710 |
- `batch_sampler`: no_duplicates
|
711 |
- `multi_dataset_batch_sampler`: proportional
|
712 |
- `router_mapping`: {'query': 'query', 'answer': 'document'}
|
713 |
-
- `learning_rate_mapping`: {'
|
714 |
|
715 |
</details>
|
716 |
|
|
|
377 |
```
|
378 |
SparseEncoder(
|
379 |
(0): Router(
|
380 |
+
(query_0_SparseStaticEmbedding): SparseStaticEmbedding({'frozen': False}, dim:30522, tokenizer: DistilBertTokenizerFast)
|
381 |
+
(document_0_MLMTransformer): MLMTransformer({'max_seq_length': 512, 'do_lower_case': False, 'architecture': 'DistilBertForMaskedLM'})
|
382 |
(document_1_SpladePooling): SpladePooling({'pooling_strategy': 'max', 'activation_function': 'relu', 'word_embedding_dimension': 30522})
|
383 |
)
|
384 |
)
|
|
|
591 |
- `fp16`: True
|
592 |
- `batch_sampler`: no_duplicates
|
593 |
- `router_mapping`: {'query': 'query', 'answer': 'document'}
|
594 |
+
- `learning_rate_mapping`: {'SparseStaticEmbedding\\.weight': 0.001}
|
595 |
|
596 |
#### All Hyperparameters
|
597 |
<details><summary>Click to expand</summary>
|
|
|
710 |
- `batch_sampler`: no_duplicates
|
711 |
- `multi_dataset_batch_sampler`: proportional
|
712 |
- `router_mapping`: {'query': 'query', 'answer': 'document'}
|
713 |
+
- `learning_rate_mapping`: {'SparseStaticEmbedding\\.weight': 0.001}
|
714 |
|
715 |
</details>
|
716 |
|