aps6992 commited on
Commit
4952134
1 Parent(s): b6edb72

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +25 -7
README.md CHANGED
@@ -41,21 +41,38 @@ adapter_name = model.load_adapter("allenai/specter2", source="hf", set_active=Tr
41
 
42
  <!-- Provide a quick summary of what the model is/does. -->
43
 
44
- SPECTER 2.0 is the successor to [SPECTER](allenai/specter) and is capable of generating task specific embeddings for scientific tasks when paired with [adapters](https://huggingface.co/models?search=allenai/specter-2).
45
- This is the proximity adapter and should be used for all general embedding purposes.
 
 
 
 
 
 
 
 
 
 
 
 
 
46
  Given the combination of title and abstract of a scientific paper or a short texual query, the model can be used to generate effective embeddings to be used in downstream applications.
47
 
 
 
 
 
48
  # Model Details
49
 
50
  ## Model Description
51
 
52
  SPECTER 2.0 has been trained on over 6M triplets of scientific paper citations, which are available [here](https://huggingface.co/datasets/allenai/scirepeval/viewer/cite_prediction_new/evaluation).
53
- Post that it is trained on all the [SciRepEval](https://huggingface.co/datasets/allenai/scirepeval) training tasks, with task format specific adapters.
54
 
55
  Task Formats trained on:
56
  - Classification
57
  - Regression
58
- - Proximity
59
  - Adhoc Search
60
 
61
  This is a retrieval specific adapter. For tasks where given a paper query, other relevant papers have to be retrieved from a corpus, use this adapter to generate the embeddings.
@@ -87,12 +104,13 @@ It builds on the work done in [SciRepEval: A Multi-Format Benchmark for Scientif
87
 
88
  |Model|Name and HF link|Description|
89
  |--|--|--|
90
- |Retrieval*|[allenai/specter2_proximity](https://huggingface.co/allenai/specter2)|Encode papers as queries and candidates eg. Link Prediction, Nearest Neighbor Search|
91
- |Adhoc Query|[allenai/specter2_adhoc_query](https://huggingface.co/allenai/specter2_adhoc_query)|Encode short raw text queries for search tasks. (Candidate papers can be encoded with proximity)|
92
  |Classification|[allenai/specter2_classification](https://huggingface.co/allenai/specter2_classification)|Encode papers to feed into linear classifiers as features|
93
  |Regression|[allenai/specter2_regression](https://huggingface.co/allenai/specter2_regression)|Encode papers to feed into linear regressors as features|
94
 
95
- *Retrieval model should suffice for downstream task types not mentioned above
 
96
 
97
  ```python
98
  from transformers import AutoTokenizer, AutoModel
 
41
 
42
  <!-- Provide a quick summary of what the model is/does. -->
43
 
44
+ **Aug 2023 Update:**
45
+ 1. The SPECTER 2.0 Base and proximity adapter models have been renamed in Hugging Face based upon usage patterns as follows:
46
+
47
+ |Old Name|New Name|
48
+ |--|--|
49
+ |allenai/specter2|[allenai/specter2_base](https://huggingface.co/allenai/specter2_base)|
50
+ |allenai/specter2_proximity|[allenai/specter2](https://huggingface.co/allenai/specter2)|
51
+
52
+ 2. We have a parallel version (termed [aug2023refresh](https://huggingface.co/allenai/specter2_aug2023refresh)) where the base transformer encoder version is pre-trained on a collection of newer papers (published after 2018).
53
+ However, for benchmarking purposes, please continue using the current version.
54
+
55
+
56
+
57
+ SPECTER 2.0 is the successor to [SPECTER](https://huggingface.co/allenai/specter) and is capable of generating task specific embeddings for scientific tasks when paired with [adapters](https://huggingface.co/models?search=allenai/specter-2_).
58
+ This is the base model to be used along with the adapters.
59
  Given the combination of title and abstract of a scientific paper or a short texual query, the model can be used to generate effective embeddings to be used in downstream applications.
60
 
61
+ **Note:For general embedding purposes, please use [allenai/specter2](https://huggingface.co/allenai/specter2).**
62
+
63
+ **To get the best performance on a downstream task type please load the associated adapter with the base model as in the example below.**
64
+
65
  # Model Details
66
 
67
  ## Model Description
68
 
69
  SPECTER 2.0 has been trained on over 6M triplets of scientific paper citations, which are available [here](https://huggingface.co/datasets/allenai/scirepeval/viewer/cite_prediction_new/evaluation).
70
+ Post that it is trained with additionally attached task format specific adapter modules on all the [SciRepEval](https://huggingface.co/datasets/allenai/scirepeval) training tasks.
71
 
72
  Task Formats trained on:
73
  - Classification
74
  - Regression
75
+ - Proximity (Retrieval)
76
  - Adhoc Search
77
 
78
  This is a retrieval specific adapter. For tasks where given a paper query, other relevant papers have to be retrieved from a corpus, use this adapter to generate the embeddings.
 
104
 
105
  |Model|Name and HF link|Description|
106
  |--|--|--|
107
+ |Proximity*|[allenai/specter2](https://huggingface.co/allenai/specter2)|Encode papers as queries and candidates eg. Link Prediction, Nearest Neighbor Search|
108
+ |Adhoc Query|[allenai/specter2_adhoc_query](https://huggingface.co/allenai/specter2_adhoc_query)|Encode short raw text queries for search tasks. (Candidate papers can be encoded with the proximity adapter)|
109
  |Classification|[allenai/specter2_classification](https://huggingface.co/allenai/specter2_classification)|Encode papers to feed into linear classifiers as features|
110
  |Regression|[allenai/specter2_regression](https://huggingface.co/allenai/specter2_regression)|Encode papers to feed into linear regressors as features|
111
 
112
+ *Proximity model should suffice for downstream task types not mentioned above
113
+
114
 
115
  ```python
116
  from transformers import AutoTokenizer, AutoModel