CharlesPing commited on
Commit
89c0403
·
verified ·
1 Parent(s): 0e7a5c9

Add new CrossEncoder model

Browse files
Files changed (3) hide show
  1. README.md +133 -92
  2. config.json +1 -1
  3. model.safetensors +2 -2
README.md CHANGED
@@ -3,19 +3,19 @@ tags:
3
  - sentence-transformers
4
  - cross-encoder
5
  - generated_from_trainer
6
- - dataset_size:18660
7
  - loss:BinaryCrossEntropyLoss
8
- base_model: cross-encoder/ms-marco-MiniLM-L6-v2
9
  datasets:
10
- - CharlesPing/climate-cross-encoder-mixed-neg-v2
11
  pipeline_tag: text-ranking
12
  library_name: sentence-transformers
13
  metrics:
14
  - map
15
- - mrr@5
16
- - ndcg@5
17
  model-index:
18
- - name: CrossEncoder based on cross-encoder/ms-marco-MiniLM-L6-v2
19
  results:
20
  - task:
21
  type: cross-encoder-reranking
@@ -25,29 +25,29 @@ model-index:
25
  type: climate-rerank-multineg
26
  metrics:
27
  - type: map
28
- value: 0.8845
29
  name: Map
30
- - type: mrr@5
31
- value: 0.9363
32
- name: Mrr@5
33
- - type: ndcg@5
34
- value: 0.8786
35
- name: Ndcg@5
36
  ---
37
 
38
- # CrossEncoder based on cross-encoder/ms-marco-MiniLM-L6-v2
39
 
40
- This is a [Cross Encoder](https://www.sbert.net/docs/cross_encoder/usage/usage.html) model finetuned from [cross-encoder/ms-marco-MiniLM-L6-v2](https://huggingface.co/cross-encoder/ms-marco-MiniLM-L6-v2) on the [climate-cross-encoder-mixed-neg-v2](https://huggingface.co/datasets/CharlesPing/climate-cross-encoder-mixed-neg-v2) dataset using the [sentence-transformers](https://www.SBERT.net) library. It computes scores for pairs of texts, which can be used for text reranking and semantic search.
41
 
42
  ## Model Details
43
 
44
  ### Model Description
45
  - **Model Type:** Cross Encoder
46
- - **Base model:** [cross-encoder/ms-marco-MiniLM-L6-v2](https://huggingface.co/cross-encoder/ms-marco-MiniLM-L6-v2) <!-- at revision ce0834f22110de6d9222af7a7a03628121708969 -->
47
  - **Maximum Sequence Length:** 512 tokens
48
  - **Number of Output Labels:** 1 label
49
  - **Training Dataset:**
50
- - [climate-cross-encoder-mixed-neg-v2](https://huggingface.co/datasets/CharlesPing/climate-cross-encoder-mixed-neg-v2)
51
  <!-- - **Language:** Unknown -->
52
  <!-- - **License:** Unknown -->
53
 
@@ -76,11 +76,11 @@ from sentence_transformers import CrossEncoder
76
  model = CrossEncoder("CharlesPing/finetuned-ce-climate-multineg-v1")
77
  # Get scores for pairs of texts
78
  pairs = [
79
- ['“Global warming and climate change, even if it is 100% caused by humans, is so slow that it cannot be observed by anyone in their lifetime.', 'The effects of climate change on human systems, mostly due to warming and shifts in precipitation, have been detected worldwide.'],
80
- ['“Global warming and climate change, even if it is 100% caused by humans, is so slow that it cannot be observed by anyone in their lifetime.', 'It is extremely likely (95–100% probability) that human influence was the dominant cause of global warming between 1951–2010.'],
81
- ['“Global warming and climate change, even if it is 100% caused by humans, is so slow that it cannot be observed by anyone in their lifetime.', 'It was eventually determined that the destruction of the ozone was caused by chlorofluorocarbons (CFCs) emitted by human products.'],
82
- ['“Global warming and climate change, even if it is 100% caused by humans, is so slow that it cannot be observed by anyone in their lifetime.', 'Ross Petty (born August 29, 1946 in Winnipeg, Manitoba) is a Canadian actor and theatre producer.'],
83
- ['“Global warming and climate change, even if it is 100% caused by humans, is so slow that it cannot be observed by anyone in their lifetime.', 'On December 23, 2013, due to low ratings, Univision started broadcasting Mentir para vivir weeknights at 7pm/6c, replacing Corazón indomable.'],
84
  ]
85
  scores = model.predict(pairs)
86
  print(scores.shape)
@@ -88,13 +88,13 @@ print(scores.shape)
88
 
89
  # Or rank different texts based on similarity to a single text
90
  ranks = model.rank(
91
- '“Global warming and climate change, even if it is 100% caused by humans, is so slow that it cannot be observed by anyone in their lifetime.',
92
  [
93
- 'The effects of climate change on human systems, mostly due to warming and shifts in precipitation, have been detected worldwide.',
94
- 'It is extremely likely (95–100% probability) that human influence was the dominant cause of global warming between 1951–2010.',
95
- 'It was eventually determined that the destruction of the ozone was caused by chlorofluorocarbons (CFCs) emitted by human products.',
96
- 'Ross Petty (born August 29, 1946 in Winnipeg, Manitoba) is a Canadian actor and theatre producer.',
97
- 'On December 23, 2013, due to low ratings, Univision started broadcasting Mentir para vivir weeknights at 7pm/6c, replacing Corazón indomable.',
98
  ]
99
  )
100
  # [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]
@@ -134,16 +134,16 @@ You can finetune this model on your own dataset.
134
  * Evaluated with [<code>CrossEncoderRerankingEvaluator</code>](https://sbert.net/docs/package_reference/cross_encoder/evaluation.html#sentence_transformers.cross_encoder.evaluation.CrossEncoderRerankingEvaluator) with these parameters:
135
  ```json
136
  {
137
- "at_k": 5,
138
  "always_rerank_positives": false
139
  }
140
  ```
141
 
142
  | Metric | Value |
143
  |:-----------|:---------------------|
144
- | map | 0.8845 (-0.1155) |
145
- | mrr@5 | 0.9363 (-0.0637) |
146
- | **ndcg@5** | **0.8786 (-0.1214)** |
147
 
148
  <!--
149
  ## Bias, Risks and Limitations
@@ -161,22 +161,22 @@ You can finetune this model on your own dataset.
161
 
162
  ### Training Dataset
163
 
164
- #### climate-cross-encoder-mixed-neg-v2
165
 
166
- * Dataset: [climate-cross-encoder-mixed-neg-v2](https://huggingface.co/datasets/CharlesPing/climate-cross-encoder-mixed-neg-v2) at [c30b8df](https://huggingface.co/datasets/CharlesPing/climate-cross-encoder-mixed-neg-v2/tree/c30b8dffb3440f0bfdd43534562c480fe24a20b4)
167
- * Size: 18,660 training samples
168
  * Columns: <code>query</code>, <code>doc</code>, and <code>label</code>
169
  * Approximate statistics based on the first 1000 samples:
170
- | | query | doc | label |
171
- |:--------|:-------------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------|:--------------------------------------------------------------|
172
- | type | string | string | float |
173
- | details | <ul><li>min: 34 characters</li><li>mean: 124.48 characters</li><li>max: 267 characters</li></ul> | <ul><li>min: 3 characters</li><li>mean: 140.91 characters</li><li>max: 654 characters</li></ul> | <ul><li>min: 0.0</li><li>mean: 0.2</li><li>max: 1.0</li></ul> |
174
  * Samples:
175
- | query | doc | label |
176
- |:-----------------------------------------------------------------------------------------------------------|:--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------|
177
- | <code>Every year air pollution protections are delayed, another 34,000 people will die prematurely.</code> | <code>The Chernobyl Forum predicts that the eventual death toll could reach 4,000 among those exposed to the highest levels of radiation (200,000 emergency workers, 116,000 evacuees and 270,000 residents of the most contaminated areas); this figure is a total causal death toll prediction, combining the deaths of approximately 50 emergency workers who died soon after the accident from acute radiation syndrome, 15 children who have died of thyroid cancer and a future predicted total of 3,935 deaths from radiation-induced cancer and leukaemia.</code> | <code>1.0</code> |
178
- | <code>Every year air pollution protections are delayed, another 34,000 people will die prematurely.</code> | <code>Inguinal, femoral and abdominal hernias were present in 18.5 million people and resulted in 59,800 deaths in 2015.</code> | <code>0.0</code> |
179
- | <code>Every year air pollution protections are delayed, another 34,000 people will die prematurely.</code> | <code>It treated victims of tuberculosis at a time when 60, 000 people a year were dying from the disease in England and Wales and the annual mortality rate in Newcastle alone was nearly 600.</code> | <code>0.0</code> |
180
  * Loss: [<code>BinaryCrossEntropyLoss</code>](https://sbert.net/docs/package_reference/cross_encoder/losses.html#binarycrossentropyloss) with these parameters:
181
  ```json
182
  {
@@ -187,22 +187,22 @@ You can finetune this model on your own dataset.
187
 
188
  ### Evaluation Dataset
189
 
190
- #### climate-cross-encoder-mixed-neg-v2
191
 
192
- * Dataset: [climate-cross-encoder-mixed-neg-v2](https://huggingface.co/datasets/CharlesPing/climate-cross-encoder-mixed-neg-v2) at [c30b8df](https://huggingface.co/datasets/CharlesPing/climate-cross-encoder-mixed-neg-v2/tree/c30b8dffb3440f0bfdd43534562c480fe24a20b4)
193
- * Size: 1,950 evaluation samples
194
  * Columns: <code>query</code>, <code>doc</code>, and <code>label</code>
195
  * Approximate statistics based on the first 1000 samples:
196
- | | query | doc | label |
197
- |:--------|:-------------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------|:--------------------------------------------------------------|
198
- | type | string | string | float |
199
- | details | <ul><li>min: 26 characters</li><li>mean: 115.07 characters</li><li>max: 240 characters</li></ul> | <ul><li>min: 9 characters</li><li>mean: 135.98 characters</li><li>max: 731 characters</li></ul> | <ul><li>min: 0.0</li><li>mean: 0.2</li><li>max: 1.0</li></ul> |
200
  * Samples:
201
- | query | doc | label |
202
- |:---------------------------------------------------------------------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------|
203
- | <code>“Global warming and climate change, even if it is 100% caused by humans, is so slow that it cannot be observed by anyone in their lifetime.</code> | <code>The effects of climate change on human systems, mostly due to warming and shifts in precipitation, have been detected worldwide.</code> | <code>1.0</code> |
204
- | <code>“Global warming and climate change, even if it is 100% caused by humans, is so slow that it cannot be observed by anyone in their lifetime.</code> | <code>It is extremely likely (95–100% probability) that human influence was the dominant cause of global warming between 1951–2010.</code> | <code>0.0</code> |
205
- | <code>“Global warming and climate change, even if it is 100% caused by humans, is so slow that it cannot be observed by anyone in their lifetime.</code> | <code>It was eventually determined that the destruction of the ozone was caused by chlorofluorocarbons (CFCs) emitted by human products.</code> | <code>0.0</code> |
206
  * Loss: [<code>BinaryCrossEntropyLoss</code>](https://sbert.net/docs/package_reference/cross_encoder/losses.html#binarycrossentropyloss) with these parameters:
207
  ```json
208
  {
@@ -343,43 +343,84 @@ You can finetune this model on your own dataset.
343
  </details>
344
 
345
  ### Training Logs
346
- | Epoch | Step | Training Loss | Validation Loss | climate-rerank-multineg_ndcg@5 |
347
  |:----------:|:--------:|:-------------:|:---------------:|:------------------------------:|
348
- | 0.0857 | 100 | 0.7355 | - | - |
349
- | 0.1714 | 200 | 0.4147 | - | - |
350
- | 0.2571 | 300 | 0.3579 | - | - |
351
- | 0.3428 | 400 | 0.286 | - | - |
352
- | 0.4284 | 500 | 0.2954 | 0.2659 | 0.8646 (-0.1354) |
353
- | 0.5141 | 600 | 0.2885 | - | - |
354
- | 0.5998 | 700 | 0.2904 | - | - |
355
- | 0.6855 | 800 | 0.2533 | - | - |
356
- | 0.7712 | 900 | 0.2656 | - | - |
357
- | 0.8569 | 1000 | 0.229 | 0.2649 | 0.8769 (-0.1231) |
358
- | 0.9426 | 1100 | 0.2619 | - | - |
359
- | 1.0283 | 1200 | 0.2348 | - | - |
360
- | 1.1140 | 1300 | 0.219 | - | - |
361
- | 1.1997 | 1400 | 0.2139 | - | - |
362
- | 1.2853 | 1500 | 0.2339 | 0.2661 | 0.8850 (-0.1150) |
363
- | 1.3710 | 1600 | 0.2369 | - | - |
364
- | 1.4567 | 1700 | 0.21 | - | - |
365
- | 1.5424 | 1800 | 0.1978 | - | - |
366
- | 1.6281 | 1900 | 0.2005 | - | - |
367
- | **1.7138** | **2000** | **0.2147** | **0.2623** | **0.8789 (-0.1211)** |
368
- | 1.7995 | 2100 | 0.2237 | - | - |
369
- | 1.8852 | 2200 | 0.2155 | - | - |
370
- | 1.9709 | 2300 | 0.1941 | - | - |
371
- | 2.0566 | 2400 | 0.1932 | - | - |
372
- | 2.1422 | 2500 | 0.1799 | 0.2995 | 0.8735 (-0.1265) |
373
- | 2.2279 | 2600 | 0.1482 | - | - |
374
- | 2.3136 | 2700 | 0.1752 | - | - |
375
- | 2.3993 | 2800 | 0.1811 | - | - |
376
- | 2.4850 | 2900 | 0.1531 | - | - |
377
- | 2.5707 | 3000 | 0.1745 | 0.3011 | 0.8773 (-0.1227) |
378
- | 2.6564 | 3100 | 0.1837 | - | - |
379
- | 2.7421 | 3200 | 0.1841 | - | - |
380
- | 2.8278 | 3300 | 0.1816 | - | - |
381
- | 2.9135 | 3400 | 0.1655 | - | - |
382
- | 2.9991 | 3500 | 0.1948 | 0.2974 | 0.8786 (-0.1214) |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
383
 
384
  * The bold row denotes the saved checkpoint.
385
 
 
3
  - sentence-transformers
4
  - cross-encoder
5
  - generated_from_trainer
6
+ - dataset_size:41052
7
  - loss:BinaryCrossEntropyLoss
8
+ base_model: cross-encoder/ms-marco-MiniLM-L12-v2
9
  datasets:
10
+ - CharlesPing/climate-cross-encoder-mixed-neg-v3
11
  pipeline_tag: text-ranking
12
  library_name: sentence-transformers
13
  metrics:
14
  - map
15
+ - mrr@1
16
+ - ndcg@1
17
  model-index:
18
+ - name: CrossEncoder based on cross-encoder/ms-marco-MiniLM-L12-v2
19
  results:
20
  - task:
21
  type: cross-encoder-reranking
 
25
  type: climate-rerank-multineg
26
  metrics:
27
  - type: map
28
+ value: 0.6809
29
  name: Map
30
+ - type: mrr@1
31
+ value: 0.6748
32
+ name: Mrr@1
33
+ - type: ndcg@1
34
+ value: 0.6748
35
+ name: Ndcg@1
36
  ---
37
 
38
+ # CrossEncoder based on cross-encoder/ms-marco-MiniLM-L12-v2
39
 
40
+ This is a [Cross Encoder](https://www.sbert.net/docs/cross_encoder/usage/usage.html) model finetuned from [cross-encoder/ms-marco-MiniLM-L12-v2](https://huggingface.co/cross-encoder/ms-marco-MiniLM-L12-v2) on the [climate-cross-encoder-mixed-neg-v3](https://huggingface.co/datasets/CharlesPing/climate-cross-encoder-mixed-neg-v3) dataset using the [sentence-transformers](https://www.SBERT.net) library. It computes scores for pairs of texts, which can be used for text reranking and semantic search.
41
 
42
  ## Model Details
43
 
44
  ### Model Description
45
  - **Model Type:** Cross Encoder
46
+ - **Base model:** [cross-encoder/ms-marco-MiniLM-L12-v2](https://huggingface.co/cross-encoder/ms-marco-MiniLM-L12-v2) <!-- at revision a34da8fab3ad458d48778dea3276ce729857efaf -->
47
  - **Maximum Sequence Length:** 512 tokens
48
  - **Number of Output Labels:** 1 label
49
  - **Training Dataset:**
50
+ - [climate-cross-encoder-mixed-neg-v3](https://huggingface.co/datasets/CharlesPing/climate-cross-encoder-mixed-neg-v3)
51
  <!-- - **Language:** Unknown -->
52
  <!-- - **License:** Unknown -->
53
 
 
76
  model = CrossEncoder("CharlesPing/finetuned-ce-climate-multineg-v1")
77
  # Get scores for pairs of texts
78
  pairs = [
79
+ ['Scientific analysis of past climates\xa0shows that greenhouse gasses, principally CO2,\xa0have controlled most ancient\xa0climate changes.', 'Greenhouse gases, in particular carbon dioxide and methane, played a significant role during the Eocene in controlling the surface temperature.'],
80
+ ['Scientific analysis of past climates\xa0shows that greenhouse gasses, principally CO2,\xa0have controlled most ancient\xa0climate changes.', 'Climatic geomorphology is of limited use to study recent (Quaternary, Holocene) large climate changes since there are seldom discernible in the geomorphological record.'],
81
+ ['Scientific analysis of past climates\xa0shows that greenhouse gasses, principally CO2,\xa0have controlled most ancient\xa0climate changes.', 'There is also a close correlation between CO2 and temperature, where CO2 has a strong control over global temperatures in Earth history.'],
82
+ ['Scientific analysis of past climates\xa0shows that greenhouse gasses, principally CO2,\xa0have controlled most ancient\xa0climate changes.', 'While scientists knew of past climate change such as the ice ages, the concept of climate as unchanging was useful in the development of a general theory of what determines climate.'],
83
+ ['Scientific analysis of past climates\xa0shows that greenhouse gasses, principally CO2,\xa0have controlled most ancient\xa0climate changes.', 'Some long term modifications along the history of the planet have been significant, such as the incorporation of oxygen to the atmosphere.'],
84
  ]
85
  scores = model.predict(pairs)
86
  print(scores.shape)
 
88
 
89
  # Or rank different texts based on similarity to a single text
90
  ranks = model.rank(
91
+ 'Scientific analysis of past climates\xa0shows that greenhouse gasses, principally CO2,\xa0have controlled most ancient\xa0climate changes.',
92
  [
93
+ 'Greenhouse gases, in particular carbon dioxide and methane, played a significant role during the Eocene in controlling the surface temperature.',
94
+ 'Climatic geomorphology is of limited use to study recent (Quaternary, Holocene) large climate changes since there are seldom discernible in the geomorphological record.',
95
+ 'There is also a close correlation between CO2 and temperature, where CO2 has a strong control over global temperatures in Earth history.',
96
+ 'While scientists knew of past climate change such as the ice ages, the concept of climate as unchanging was useful in the development of a general theory of what determines climate.',
97
+ 'Some long term modifications along the history of the planet have been significant, such as the incorporation of oxygen to the atmosphere.',
98
  ]
99
  )
100
  # [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]
 
134
  * Evaluated with [<code>CrossEncoderRerankingEvaluator</code>](https://sbert.net/docs/package_reference/cross_encoder/evaluation.html#sentence_transformers.cross_encoder.evaluation.CrossEncoderRerankingEvaluator) with these parameters:
135
  ```json
136
  {
137
+ "at_k": 1,
138
  "always_rerank_positives": false
139
  }
140
  ```
141
 
142
  | Metric | Value |
143
  |:-----------|:---------------------|
144
+ | map | 0.6809 (-0.3191) |
145
+ | mrr@1 | 0.6748 (-0.3252) |
146
+ | **ndcg@1** | **0.6748 (-0.3252)** |
147
 
148
  <!--
149
  ## Bias, Risks and Limitations
 
161
 
162
  ### Training Dataset
163
 
164
+ #### climate-cross-encoder-mixed-neg-v3
165
 
166
+ * Dataset: [climate-cross-encoder-mixed-neg-v3](https://huggingface.co/datasets/CharlesPing/climate-cross-encoder-mixed-neg-v3) at [cd49b57](https://huggingface.co/datasets/CharlesPing/climate-cross-encoder-mixed-neg-v3/tree/cd49b575d132b8a377614b398f1a30378ce0c563)
167
+ * Size: 41,052 training samples
168
  * Columns: <code>query</code>, <code>doc</code>, and <code>label</code>
169
  * Approximate statistics based on the first 1000 samples:
170
+ | | query | doc | label |
171
+ |:--------|:-------------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------|:---------------------------------------------------------------|
172
+ | type | string | string | float |
173
+ | details | <ul><li>min: 49 characters</li><li>mean: 140.03 characters</li><li>max: 306 characters</li></ul> | <ul><li>min: 4 characters</li><li>mean: 136.03 characters</li><li>max: 731 characters</li></ul> | <ul><li>min: 0.0</li><li>mean: 0.09</li><li>max: 1.0</li></ul> |
174
  * Samples:
175
+ | query | doc | label |
176
+ |:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------|
177
+ | <code>“A leading Canadian authority on polar bears, Mitch Taylor, said: ‘We’re seeing an increase in bears that’s really unprecedented, and in places where we’re seeing a decrease in the population</code> | <code>Warnings about the future of the polar bear are often contrasted with the fact that worldwide population estimates have increased over the past 50 years and are relatively stable today.</code> | <code>1.0</code> |
178
+ | <code>“A leading Canadian authority on polar bears, Mitch Taylor, said: ‘We’re seeing an increase in bears that’s really unprecedented, and in places where we’re seeing a decrease in the population</code> | <code>Species distribution models of recent years indicate that the deer tick, known as "I. scapularis," is pushing its distribution to higher latitudes of the Northeastern United States and Canada, as well as pushing and maintaining populations in the South Central and Northern Midwest regions of the United States.</code> | <code>0.0</code> |
179
+ | <code>“A leading Canadian authority on polar bears, Mitch Taylor, said: ‘We’re seeing an increase in bears that’s really unprecedented, and in places where we’re seeing a decrease in the population</code> | <code>Bear and deer are among the animals present.</code> | <code>0.0</code> |
180
  * Loss: [<code>BinaryCrossEntropyLoss</code>](https://sbert.net/docs/package_reference/cross_encoder/losses.html#binarycrossentropyloss) with these parameters:
181
  ```json
182
  {
 
187
 
188
  ### Evaluation Dataset
189
 
190
+ #### climate-cross-encoder-mixed-neg-v3
191
 
192
+ * Dataset: [climate-cross-encoder-mixed-neg-v3](https://huggingface.co/datasets/CharlesPing/climate-cross-encoder-mixed-neg-v3) at [cd49b57](https://huggingface.co/datasets/CharlesPing/climate-cross-encoder-mixed-neg-v3/tree/cd49b575d132b8a377614b398f1a30378ce0c563)
193
+ * Size: 4,290 evaluation samples
194
  * Columns: <code>query</code>, <code>doc</code>, and <code>label</code>
195
  * Approximate statistics based on the first 1000 samples:
196
+ | | query | doc | label |
197
+ |:--------|:-------------------------------------------------------------------------------------------------|:-------------------------------------------------------------------------------------------------|:---------------------------------------------------------------|
198
+ | type | string | string | float |
199
+ | details | <ul><li>min: 39 characters</li><li>mean: 116.67 characters</li><li>max: 240 characters</li></ul> | <ul><li>min: 18 characters</li><li>mean: 132.92 characters</li><li>max: 731 characters</li></ul> | <ul><li>min: 0.0</li><li>mean: 0.09</li><li>max: 1.0</li></ul> |
200
  * Samples:
201
+ | query | doc | label |
202
+ |:-----------------------------------------------------------------------------------------------------------------------------------------------|:--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------|
203
+ | <code>Scientific analysis of past climates shows that greenhouse gasses, principally CO2, have controlled most ancient climate changes.</code> | <code>Greenhouse gases, in particular carbon dioxide and methane, played a significant role during the Eocene in controlling the surface temperature.</code> | <code>1.0</code> |
204
+ | <code>Scientific analysis of past climates shows that greenhouse gasses, principally CO2, have controlled most ancient climate changes.</code> | <code>Climatic geomorphology is of limited use to study recent (Quaternary, Holocene) large climate changes since there are seldom discernible in the geomorphological record.</code> | <code>0.0</code> |
205
+ | <code>Scientific analysis of past climates shows that greenhouse gasses, principally CO2, have controlled most ancient climate changes.</code> | <code>There is also a close correlation between CO2 and temperature, where CO2 has a strong control over global temperatures in Earth history.</code> | <code>0.0</code> |
206
  * Loss: [<code>BinaryCrossEntropyLoss</code>](https://sbert.net/docs/package_reference/cross_encoder/losses.html#binarycrossentropyloss) with these parameters:
207
  ```json
208
  {
 
343
  </details>
344
 
345
  ### Training Logs
346
+ | Epoch | Step | Training Loss | Validation Loss | climate-rerank-multineg_ndcg@1 |
347
  |:----------:|:--------:|:-------------:|:---------------:|:------------------------------:|
348
+ | 0.0390 | 100 | 0.5097 | - | - |
349
+ | 0.0779 | 200 | 0.3662 | - | - |
350
+ | 0.1169 | 300 | 0.3034 | - | - |
351
+ | 0.1559 | 400 | 0.2655 | - | - |
352
+ | 0.1949 | 500 | 0.2651 | 0.2262 | 0.6585 (-0.3415) |
353
+ | 0.2338 | 600 | 0.2161 | - | - |
354
+ | 0.2728 | 700 | 0.227 | - | - |
355
+ | 0.3118 | 800 | 0.235 | - | - |
356
+ | 0.3507 | 900 | 0.2243 | - | - |
357
+ | 0.3897 | 1000 | 0.2081 | 0.2174 | 0.6992 (-0.3008) |
358
+ | 0.4287 | 1100 | 0.1961 | - | - |
359
+ | 0.4677 | 1200 | 0.207 | - | - |
360
+ | 0.5066 | 1300 | 0.2375 | - | - |
361
+ | 0.5456 | 1400 | 0.2117 | - | - |
362
+ | 0.5846 | 1500 | 0.2058 | 0.2253 | 0.6748 (-0.3252) |
363
+ | 0.6235 | 1600 | 0.2163 | - | - |
364
+ | 0.6625 | 1700 | 0.2235 | - | - |
365
+ | 0.7015 | 1800 | 0.2193 | - | - |
366
+ | 0.7405 | 1900 | 0.1924 | - | - |
367
+ | 0.7794 | 2000 | 0.2084 | 0.2095 | 0.6748 (-0.3252) |
368
+ | 0.8184 | 2100 | 0.2113 | - | - |
369
+ | 0.8574 | 2200 | 0.2276 | - | - |
370
+ | 0.8963 | 2300 | 0.2071 | - | - |
371
+ | 0.9353 | 2400 | 0.2374 | - | - |
372
+ | 0.9743 | 2500 | 0.2173 | 0.2172 | 0.6667 (-0.3333) |
373
+ | 1.0133 | 2600 | 0.2011 | - | - |
374
+ | 1.0522 | 2700 | 0.1634 | - | - |
375
+ | 1.0912 | 2800 | 0.1807 | - | - |
376
+ | 1.1302 | 2900 | 0.1878 | - | - |
377
+ | 1.1691 | 3000 | 0.2037 | 0.2147 | 0.6911 (-0.3089) |
378
+ | 1.2081 | 3100 | 0.1904 | - | - |
379
+ | 1.2471 | 3200 | 0.1911 | - | - |
380
+ | 1.2860 | 3300 | 0.1828 | - | - |
381
+ | 1.3250 | 3400 | 0.1686 | - | - |
382
+ | 1.3640 | 3500 | 0.1892 | 0.2179 | 0.6992 (-0.3008) |
383
+ | 1.4030 | 3600 | 0.188 | - | - |
384
+ | 1.4419 | 3700 | 0.1691 | - | - |
385
+ | 1.4809 | 3800 | 0.1946 | - | - |
386
+ | 1.5199 | 3900 | 0.1938 | - | - |
387
+ | **1.5588** | **4000** | **0.211** | **0.2088** | **0.6992 (-0.3008)** |
388
+ | 1.5978 | 4100 | 0.1826 | - | - |
389
+ | 1.6368 | 4200 | 0.1608 | - | - |
390
+ | 1.6758 | 4300 | 0.1782 | - | - |
391
+ | 1.7147 | 4400 | 0.1803 | - | - |
392
+ | 1.7537 | 4500 | 0.1804 | 0.2160 | 0.6911 (-0.3089) |
393
+ | 1.7927 | 4600 | 0.1823 | - | - |
394
+ | 1.8316 | 4700 | 0.1844 | - | - |
395
+ | 1.8706 | 4800 | 0.1727 | - | - |
396
+ | 1.9096 | 4900 | 0.1937 | - | - |
397
+ | 1.9486 | 5000 | 0.1662 | 0.2219 | 0.6829 (-0.3171) |
398
+ | 1.9875 | 5100 | 0.1653 | - | - |
399
+ | 2.0265 | 5200 | 0.1658 | - | - |
400
+ | 2.0655 | 5300 | 0.1316 | - | - |
401
+ | 2.1044 | 5400 | 0.1379 | - | - |
402
+ | 2.1434 | 5500 | 0.152 | 0.2513 | 0.6504 (-0.3496) |
403
+ | 2.1824 | 5600 | 0.1848 | - | - |
404
+ | 2.2214 | 5700 | 0.1507 | - | - |
405
+ | 2.2603 | 5800 | 0.1495 | - | - |
406
+ | 2.2993 | 5900 | 0.1469 | - | - |
407
+ | 2.3383 | 6000 | 0.1596 | 0.2407 | 0.6585 (-0.3415) |
408
+ | 2.3772 | 6100 | 0.1518 | - | - |
409
+ | 2.4162 | 6200 | 0.1351 | - | - |
410
+ | 2.4552 | 6300 | 0.1706 | - | - |
411
+ | 2.4942 | 6400 | 0.1538 | - | - |
412
+ | 2.5331 | 6500 | 0.1329 | 0.2505 | 0.6911 (-0.3089) |
413
+ | 2.5721 | 6600 | 0.147 | - | - |
414
+ | 2.6111 | 6700 | 0.1289 | - | - |
415
+ | 2.6500 | 6800 | 0.1698 | - | - |
416
+ | 2.6890 | 6900 | 0.1456 | - | - |
417
+ | 2.7280 | 7000 | 0.141 | 0.2618 | 0.6748 (-0.3252) |
418
+ | 2.7670 | 7100 | 0.1413 | - | - |
419
+ | 2.8059 | 7200 | 0.1474 | - | - |
420
+ | 2.8449 | 7300 | 0.1381 | - | - |
421
+ | 2.8839 | 7400 | 0.1252 | - | - |
422
+ | 2.9228 | 7500 | 0.1384 | 0.2608 | 0.6748 (-0.3252) |
423
+ | 2.9618 | 7600 | 0.1826 | - | - |
424
 
425
  * The bold row denotes the saved checkpoint.
426
 
config.json CHANGED
@@ -20,7 +20,7 @@
20
  "max_position_embeddings": 512,
21
  "model_type": "bert",
22
  "num_attention_heads": 12,
23
- "num_hidden_layers": 6,
24
  "pad_token_id": 0,
25
  "position_embedding_type": "absolute",
26
  "sentence_transformers": {
 
20
  "max_position_embeddings": 512,
21
  "model_type": "bert",
22
  "num_attention_heads": 12,
23
+ "num_hidden_layers": 12,
24
  "pad_token_id": 0,
25
  "position_embedding_type": "absolute",
26
  "sentence_transformers": {
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:dda4d03c74754e827f5e0eb1f3ed87f30c452e261945de76d784ff0d530f9752
3
- size 90866412
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2fcb49a81642309eb64539d1e4bf527bea3c3d4f93f63a50a2b0f7cd2b58ccac
3
+ size 133464836