noahjax commited on
Commit
d23f1b5
·
verified ·
1 Parent(s): 15c433c

Upload fine-tuned chart reranker model

Browse files
README.md CHANGED
@@ -4,7 +4,7 @@ tags:
4
  - cross-encoder
5
  - reranker
6
  - generated_from_trainer
7
- - dataset_size:12349
8
  - loss:BinaryCrossEntropyLoss
9
  base_model: Alibaba-NLP/gte-multilingual-reranker-base
10
  pipeline_tag: text-ranking
@@ -23,10 +23,10 @@ model-index:
23
  type: validation
24
  metrics:
25
  - type: pearson
26
- value: 0.8643473065020739
27
  name: Pearson
28
  - type: spearman
29
- value: 0.8620968090164374
30
  name: Spearman
31
  ---
32
 
@@ -70,11 +70,11 @@ from sentence_transformers import CrossEncoder
70
  model = CrossEncoder("cross_encoder_model_id")
71
  # Get scores for pairs of texts
72
  pairs = [
73
- ['DJ mixers compatible with Apple Music 2025', 'Title: "Music devices - radio (United States)"\nCollections: YouGov Trackers\nDatasets: YouGovTrackerValueV2\nChart Type: survey:timeseries\nSources: YouGov'],
74
- ['current USD to PLN exchange rate', 'Title: "Conversion rate from PLN to USD"\nCollections: Foreign Exchange Rates\nDatasets: Forex\nChart Type: exchange:currency\nSources: Xignite'],
75
- ['Aktuelle Investmenttrends 2025', 'Title: "Financial activity - next 12 months (United States)"\nCollections: YouGov Trackers\nDatasets: YouGovTrackerValueV2\nChart Type: survey:timeseries\nSources: YouGov'],
76
- ["What are Amazon's accrued liabilities?", 'Title: "Amazon Expenses Accrued (Quarterly)"\nCollections: Companies\nDatasets: StandardIncomeStatement\nChart Type: timeseries:eav_v2\nCanonical forms: "Expenses Accrued"="accrued_expenses_total"\nSources: S&P Global'],
77
- ["Costco's long-term lease obligations", 'Title: "Air Lease Overview"\nCollections: Companies\nChart Type: company:finance\nCanonical forms: "Air Lease"="Air Lease Corporation", "Overview"="Stock Overview"\nSources: S&P Global'],
78
  ]
79
  scores = model.predict(pairs)
80
  print(scores.shape)
@@ -82,13 +82,13 @@ print(scores.shape)
82
 
83
  # Or rank different texts based on similarity to a single text
84
  ranks = model.rank(
85
- 'DJ mixers compatible with Apple Music 2025',
86
  [
87
- 'Title: "Music devices - radio (United States)"\nCollections: YouGov Trackers\nDatasets: YouGovTrackerValueV2\nChart Type: survey:timeseries\nSources: YouGov',
88
- 'Title: "Conversion rate from PLN to USD"\nCollections: Foreign Exchange Rates\nDatasets: Forex\nChart Type: exchange:currency\nSources: Xignite',
89
- 'Title: "Financial activity - next 12 months (United States)"\nCollections: YouGov Trackers\nDatasets: YouGovTrackerValueV2\nChart Type: survey:timeseries\nSources: YouGov',
90
- 'Title: "Amazon Expenses Accrued (Quarterly)"\nCollections: Companies\nDatasets: StandardIncomeStatement\nChart Type: timeseries:eav_v2\nCanonical forms: "Expenses Accrued"="accrued_expenses_total"\nSources: S&P Global',
91
- 'Title: "Air Lease Overview"\nCollections: Companies\nChart Type: company:finance\nCanonical forms: "Air Lease"="Air Lease Corporation", "Overview"="Stock Overview"\nSources: S&P Global',
92
  ]
93
  )
94
  # [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]
@@ -129,8 +129,8 @@ You can finetune this model on your own dataset.
129
 
130
  | Metric | Value |
131
  |:-------------|:-----------|
132
- | pearson | 0.8643 |
133
- | **spearman** | **0.8621** |
134
 
135
  <!--
136
  ## Bias, Risks and Limitations
@@ -150,19 +150,19 @@ You can finetune this model on your own dataset.
150
 
151
  #### Unnamed Dataset
152
 
153
- * Size: 12,349 training samples
154
  * Columns: <code>sentence_0</code>, <code>sentence_1</code>, and <code>label</code>
155
  * Approximate statistics based on the first 1000 samples:
156
- | | sentence_0 | sentence_1 | label |
157
- |:--------|:-----------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------|:---------------------------------------------------------------|
158
- | type | string | string | float |
159
- | details | <ul><li>min: 5 characters</li><li>mean: 46.81 characters</li><li>max: 123 characters</li></ul> | <ul><li>min: 77 characters</li><li>mean: 182.4 characters</li><li>max: 495 characters</li></ul> | <ul><li>min: 0.0</li><li>mean: 0.48</li><li>max: 1.0</li></ul> |
160
  * Samples:
161
- | sentence_0 | sentence_1 | label |
162
- |:--------------------------------------------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:------------------|
163
- | <code>DJ mixers compatible with Apple Music 2025</code> | <code>Title: "Music devices - radio (United States)"<br>Collections: YouGov Trackers<br>Datasets: YouGovTrackerValueV2<br>Chart Type: survey:timeseries<br>Sources: YouGov</code> | <code>0.25</code> |
164
- | <code>current USD to PLN exchange rate</code> | <code>Title: "Conversion rate from PLN to USD"<br>Collections: Foreign Exchange Rates<br>Datasets: Forex<br>Chart Type: exchange:currency<br>Sources: Xignite</code> | <code>0.75</code> |
165
- | <code>Aktuelle Investmenttrends 2025</code> | <code>Title: "Financial activity - next 12 months (United States)"<br>Collections: YouGov Trackers<br>Datasets: YouGovTrackerValueV2<br>Chart Type: survey:timeseries<br>Sources: YouGov</code> | <code>0.75</code> |
166
  * Loss: [<code>BinaryCrossEntropyLoss</code>](https://sbert.net/docs/package_reference/cross_encoder/losses.html#binarycrossentropyloss) with these parameters:
167
  ```json
168
  {
@@ -308,29 +308,40 @@ You can finetune this model on your own dataset.
308
  ### Training Logs
309
  | Epoch | Step | Training Loss | validation_spearman |
310
  |:------:|:----:|:-------------:|:-------------------:|
311
- | 0.2591 | 100 | - | 0.7835 |
312
- | 0.5181 | 200 | - | 0.8161 |
313
- | 0.7772 | 300 | - | 0.8369 |
314
- | 1.0 | 386 | - | 0.8392 |
315
- | 1.0363 | 400 | - | 0.8442 |
316
- | 1.2953 | 500 | 0.47 | 0.8475 |
317
- | 1.5544 | 600 | - | 0.8533 |
318
- | 1.8135 | 700 | - | 0.8544 |
319
- | 2.0 | 772 | - | 0.8579 |
320
- | 2.0725 | 800 | - | 0.8585 |
321
- | 2.3316 | 900 | - | 0.8548 |
322
- | 2.5907 | 1000 | 0.3926 | 0.8577 |
323
- | 2.8497 | 1100 | - | 0.8569 |
324
- | 3.0 | 1158 | - | 0.8607 |
325
- | 3.1088 | 1200 | - | 0.8573 |
326
- | 3.3679 | 1300 | - | 0.8614 |
327
- | 3.6269 | 1400 | - | 0.8594 |
328
- | 3.8860 | 1500 | 0.3602 | 0.8591 |
329
- | 4.0 | 1544 | - | 0.8596 |
330
- | 4.1451 | 1600 | - | 0.8611 |
331
- | 4.4041 | 1700 | - | 0.8619 |
332
- | 4.6632 | 1800 | - | 0.8618 |
333
- | 4.9223 | 1900 | - | 0.8621 |
 
 
 
 
 
 
 
 
 
 
 
334
 
335
 
336
  ### Framework Versions
 
4
  - cross-encoder
5
  - reranker
6
  - generated_from_trainer
7
+ - dataset_size:20347
8
  - loss:BinaryCrossEntropyLoss
9
  base_model: Alibaba-NLP/gte-multilingual-reranker-base
10
  pipeline_tag: text-ranking
 
23
  type: validation
24
  metrics:
25
  - type: pearson
26
+ value: 0.8381245620713855
27
  name: Pearson
28
  - type: spearman
29
+ value: 0.8388188648567115
30
  name: Spearman
31
  ---
32
 
 
70
  model = CrossEncoder("cross_encoder_model_id")
71
  # Get scores for pairs of texts
72
  pairs = [
73
+ ['Thanks, now you have everything pick the most important item or 2 or three if you find it really appropriate from each group. Just simplify this list a bit, to make sure I have my micro nutrients, vitamins, whatever checked off.', 'Title: "Natural Grocers by Vitamin Cottage Overview"\nCollections: Companies\nDatasets: InstrumentClosePrice1Day\nChart Type: timeseries:eav_v3\nCanonical forms: "Natural Grocers by Vitamin Cottage"="closing_price"'],
74
+ ['How do people feel about Nicola Sturgeon?', 'Title: "Nicola Sturgeon fame & popularity tracker (United Kingdom)"\nCollections: YouGov Trackers\nDatasets: YouGovTrackerValueV2\nChart Type: survey:timeseries\nSources: YouGov'],
75
+ ['Create a skit about hino. It should be a horror theme and humor in the end. Without the need of driving a truck. it can be about hino genuine spareparts or technician services', 'Title: "Hino Motors Overview"\nCollections: Companies\nChart Type: company:finance\nCanonical forms: "Hino Motors"="Hino Motors, Ltd.", "Overview"="Stock Overview"\nSources: S&P Global'],
76
+ ['no i mean talk about the trends in school', 'Title: "Should private schools be banned? (United Kingdom)"\nCollections: YouGov Trackers\nDatasets: YouGovTrackerValueV2\nChart Type: survey:timeseries\nSources: YouGov'],
77
+ ['Exchange rate Moroccan dirham to euro 29 October 2025', 'Title: "Conversion rate from EUR to MAD"\nCollections: Foreign Exchange Rates\nDatasets: Forex\nChart Type: exchange:currency\nSources: Xignite'],
78
  ]
79
  scores = model.predict(pairs)
80
  print(scores.shape)
 
82
 
83
  # Or rank different texts based on similarity to a single text
84
  ranks = model.rank(
85
+ 'Thanks, now you have everything pick the most important item or 2 or three if you find it really appropriate from each group. Just simplify this list a bit, to make sure I have my micro nutrients, vitamins, whatever checked off.',
86
  [
87
+ 'Title: "Natural Grocers by Vitamin Cottage Overview"\nCollections: Companies\nDatasets: InstrumentClosePrice1Day\nChart Type: timeseries:eav_v3\nCanonical forms: "Natural Grocers by Vitamin Cottage"="closing_price"',
88
+ 'Title: "Nicola Sturgeon fame & popularity tracker (United Kingdom)"\nCollections: YouGov Trackers\nDatasets: YouGovTrackerValueV2\nChart Type: survey:timeseries\nSources: YouGov',
89
+ 'Title: "Hino Motors Overview"\nCollections: Companies\nChart Type: company:finance\nCanonical forms: "Hino Motors"="Hino Motors, Ltd.", "Overview"="Stock Overview"\nSources: S&P Global',
90
+ 'Title: "Should private schools be banned? (United Kingdom)"\nCollections: YouGov Trackers\nDatasets: YouGovTrackerValueV2\nChart Type: survey:timeseries\nSources: YouGov',
91
+ 'Title: "Conversion rate from EUR to MAD"\nCollections: Foreign Exchange Rates\nDatasets: Forex\nChart Type: exchange:currency\nSources: Xignite',
92
  ]
93
  )
94
  # [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]
 
129
 
130
  | Metric | Value |
131
  |:-------------|:-----------|
132
+ | pearson | 0.8381 |
133
+ | **spearman** | **0.8388** |
134
 
135
  <!--
136
  ## Bias, Risks and Limitations
 
150
 
151
  #### Unnamed Dataset
152
 
153
+ * Size: 20,347 training samples
154
  * Columns: <code>sentence_0</code>, <code>sentence_1</code>, and <code>label</code>
155
  * Approximate statistics based on the first 1000 samples:
156
+ | | sentence_0 | sentence_1 | label |
157
+ |:--------|:-----------------------------------------------------------------------------------------------|:-------------------------------------------------------------------------------------------------|:---------------------------------------------------------------|
158
+ | type | string | string | float |
159
+ | details | <ul><li>min: 1 characters</li><li>mean: 84.39 characters</li><li>max: 943 characters</li></ul> | <ul><li>min: 74 characters</li><li>mean: 180.44 characters</li><li>max: 396 characters</li></ul> | <ul><li>min: 0.0</li><li>mean: 0.45</li><li>max: 1.0</li></ul> |
160
  * Samples:
161
+ | sentence_0 | sentence_1 | label |
162
+ |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------|
163
+ | <code>Thanks, now you have everything pick the most important item or 2 or three if you find it really appropriate from each group. Just simplify this list a bit, to make sure I have my micro nutrients, vitamins, whatever checked off.</code> | <code>Title: "Natural Grocers by Vitamin Cottage Overview"<br>Collections: Companies<br>Datasets: InstrumentClosePrice1Day<br>Chart Type: timeseries:eav_v3<br>Canonical forms: "Natural Grocers by Vitamin Cottage"="closing_price"</code> | <code>0.0</code> |
164
+ | <code>How do people feel about Nicola Sturgeon?</code> | <code>Title: "Nicola Sturgeon fame & popularity tracker (United Kingdom)"<br>Collections: YouGov Trackers<br>Datasets: YouGovTrackerValueV2<br>Chart Type: survey:timeseries<br>Sources: YouGov</code> | <code>1.0</code> |
165
+ | <code>Create a skit about hino. It should be a horror theme and humor in the end. Without the need of driving a truck. it can be about hino genuine spareparts or technician services</code> | <code>Title: "Hino Motors Overview"<br>Collections: Companies<br>Chart Type: company:finance<br>Canonical forms: "Hino Motors"="Hino Motors, Ltd.", "Overview"="Stock Overview"<br>Sources: S&P Global</code> | <code>0.5</code> |
166
  * Loss: [<code>BinaryCrossEntropyLoss</code>](https://sbert.net/docs/package_reference/cross_encoder/losses.html#binarycrossentropyloss) with these parameters:
167
  ```json
168
  {
 
308
  ### Training Logs
309
  | Epoch | Step | Training Loss | validation_spearman |
310
  |:------:|:----:|:-------------:|:-------------------:|
311
+ | 0.1572 | 100 | - | 0.7137 |
312
+ | 0.3145 | 200 | - | 0.7573 |
313
+ | 0.4717 | 300 | - | 0.7748 |
314
+ | 0.6289 | 400 | - | 0.7888 |
315
+ | 0.7862 | 500 | 0.5153 | 0.8000 |
316
+ | 0.9434 | 600 | - | 0.8039 |
317
+ | 1.0 | 636 | - | 0.8044 |
318
+ | 1.1006 | 700 | - | 0.8065 |
319
+ | 1.2579 | 800 | - | 0.8167 |
320
+ | 1.4151 | 900 | - | 0.8164 |
321
+ | 1.5723 | 1000 | 0.445 | 0.8192 |
322
+ | 1.7296 | 1100 | - | 0.8225 |
323
+ | 1.8868 | 1200 | - | 0.8287 |
324
+ | 2.0 | 1272 | - | 0.8284 |
325
+ | 2.0440 | 1300 | - | 0.8281 |
326
+ | 2.2013 | 1400 | - | 0.8255 |
327
+ | 2.3585 | 1500 | 0.4102 | 0.8276 |
328
+ | 2.5157 | 1600 | - | 0.8305 |
329
+ | 2.6730 | 1700 | - | 0.8343 |
330
+ | 2.8302 | 1800 | - | 0.8301 |
331
+ | 2.9874 | 1900 | - | 0.8351 |
332
+ | 3.0 | 1908 | - | 0.8355 |
333
+ | 3.1447 | 2000 | 0.3904 | 0.8336 |
334
+ | 3.3019 | 2100 | - | 0.8319 |
335
+ | 3.4591 | 2200 | - | 0.8319 |
336
+ | 3.6164 | 2300 | - | 0.8308 |
337
+ | 3.7736 | 2400 | - | 0.8331 |
338
+ | 3.9308 | 2500 | 0.3741 | 0.8370 |
339
+ | 4.0 | 2544 | - | 0.8383 |
340
+ | 4.0881 | 2600 | - | 0.8369 |
341
+ | 4.2453 | 2700 | - | 0.8385 |
342
+ | 4.4025 | 2800 | - | 0.8368 |
343
+ | 4.5597 | 2900 | - | 0.8370 |
344
+ | 4.7170 | 3000 | 0.3643 | 0.8388 |
345
 
346
 
347
  ### Framework Versions
eval/CrossEncoderCorrelationEvaluator_validation_results.csv CHANGED
@@ -1,6 +1,6 @@
1
  epoch,steps,Pearson_Correlation,Spearman_Correlation
2
- 1.0,386,0.837652001216976,0.8392216545657446
3
- 2.0,772,0.8587273733508869,0.8579410042452386
4
- 3.0,1158,0.8640762414247022,0.8607296855959887
5
- 4.0,1544,0.8618951242992149,0.8596087038742567
6
- 5.0,1930,0.8643015819365643,0.8620805071202834
 
1
  epoch,steps,Pearson_Correlation,Spearman_Correlation
2
+ 1.0,636,0.8050961988795169,0.8044347672638916
3
+ 2.0,1272,0.8267567950795853,0.8284146931811501
4
+ 3.0,1908,0.8351882809975475,0.8355004054548
5
+ 4.0,2544,0.8381740944766652,0.8382614031363851
6
+ 5.0,3180,0.8368434817201468,0.8374989674723212
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6485ea4ee6cc3859ccff14222884fed3c2f23e283906df217bd6bd60d8377bff
3
  size 1223854204
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6dde1675c82135fb9296d9c990693ce3373c5982f7f01cd53a72fb674e86d82
3
  size 1223854204
training_info.txt CHANGED
@@ -1,5 +1,5 @@
1
  Base Model: Alibaba-NLP/gte-multilingual-reranker-base
2
- Training Samples: 12349
3
  Epochs: 5
4
  Batch Size: 32
5
  Learning Rate: 2e-05
 
1
  Base Model: Alibaba-NLP/gte-multilingual-reranker-base
2
+ Training Samples: 20347
3
  Epochs: 5
4
  Batch Size: 32
5
  Learning Rate: 2e-05