hmm404/siamese1.0

README.md

@@ -4,72 +4,74 @@ tags:
 - sentence-similarity
 - feature-extraction
 - generated_from_trainer
-- dataset_size:32351
-- loss:TripletLoss
+- dataset_size:49273
+- loss:ContrastiveLoss
 base_model: sentence-transformers/all-mpnet-base-v2
 widget:
-- source_sentence: No civilized society permits one human to intentionally harm or
-    take the life of another human without punishment, and abortion is no different.
+- source_sentence: Evil is separate from the creator and results from creations after
+    they are created
   sentences:
-  - There cannot be any dispute whatsoever that the fetus that a human woman is pregnant
-    with is a human also, and a member of the human species is defined as a human
-    being. Regardless of a "persons" \(a human and an individual\) age they are still
-    of the human species and biologically different and unique from the mother and
-    have life, these are scientific proven facts.
-  - Edwards syndrome can be detected when a baby is in the womb.  The condition causes
-    abnormalities and defects in multiple organs. The majority of children with this
-    condition will die due to a miscarriage. Those that are born, suffer from several
-    life-threatening medical problems. The small percentage of children with Edwards'
-    syndrome who live for longer than one year, often suffer from severe intellectual
-    disabilities. It is not unethical for a woman to abort a child with Edwards syndrome.
-  - Most countries have an army, including civilized ones. Citizens in the army that
-    kill others in the line of army duties are not punished.
-- source_sentence: Ants directly "milk" aphids for their honeydew), so while saying
-    no other animal drinks another's milk is true, other animals do actively drink
-    other animals secretions in a symbiotic way.
+  - The Creation of an omniscient and omnipotent God, after it's created, remains
+    in "his" control by definition. The time dimension is part of the world \(God's
+    creation\), so the consequences to the creation as time goes by are part of the
+    creation itself. There can be no separation. If evil exists, it must be part of
+    such a God's plan.
+  - There aren't sufficient grounds for believing that the Bible contains divine revelation.
+  - Higher occurrence rate of informal or 'donkey' voting.
+- source_sentence: Liquid democracy will slow down governments even more, as all citizens
+    have to ponder about each and every issue and on top, if they conclude that they
+    do not want to vote, then about all the many possible experts for each and every
+    issue.
   sentences:
-  - Dairy cows are well fed, protected from weather and disease, and generally are
-    far better off than their wild cousins.
-  - There is nothing symbiotic about humans' consumption of cows' milk, meaning their
-    relationship is simply exploitative.
-  - Football is associated with Chronic traumatic encephalopathy \(CTE\), which typically
-    entails behavioral problems, mood problems, and problems with thinking.
-- source_sentence: Reunification is highly unlikely anyway as no international great
-    power has an interest in it.
+  - Association with someone like Comey doesn't make a strong case for bias. Comey
+    himself maintained his objectivity long beyond what most people would be capable
+    of, until he was directly dragged into a necessarily political position by the
+    President himself during the private talk.
+  - Looking at the partial participation, growth and continued existence of both small
+    niches and huge groups on Reddit, Twitter, Facebook and tons of others, verified
+    accounts and admins on these systems are probably the best examples of a delegate
+    system. We can conclude that people will participate in the parts that interest
+    them and delegate the rest, resulting in the desired speed-up of democracy.
+  - There is some chicken-or-egg here. One might argue a welfare state leads to less
+    poverty, but another would argue welfare states are a luxury only affordable by
+    wealthy countries.
+- source_sentence: Renewable energy is a better option for replacing fossil fuels
+    than nuclear.
   sentences:
-  - Japan has no interest in Korean reunification as this would result in a large,
-    more powerful Korea. The two countries have had difficult diplomatic relations
-    for decades.
-  - After mosquitoes, humans are the second cause of humans' deaths. www.telegraph.co.uk
-  - The preferences of the great powers are not the determining factor here. Korean
-    unification will ultimately be a project of the two current governments.
-- source_sentence: Christ did not participate in any war. The logical answer for a
-    Christian is thus to not participate either.
+  - Pulling out supports the impression he frequently conveys to his base that he's
+    deadly serious about America First). They seem to approve.
+  - Renewable energy sources are cheaper than nuclear.
+  - Abolishing inheritance might lead to children not treating their parents as well
+    as they would have otherwise.
+- source_sentence: Throughout the play, Claudius indicates that he does not fully
+    believe that Hamlet is mad.
   sentences:
-  - While some experts consider that overpopulation and its pressure on resources
-    may not be the most pressing problem, many agree that aging populations are indeed
-    a concern. The strain on system that rely on pensions \(where the current workers
-    fund retirees through taxes\) can have dire results.
-  - One of the central teachings of Jesus in the New Testament is love, even towards
-    your enemy. Someone driven by this attitude will not fight in a war and kill humans.
-  - There were no wars to participate in during the lifetime of Jesus.
-- source_sentence: The use of laptops in classrooms impedes students' learning.
+  - Claudius admits that Hamlet's behaviour is more indicative of depression or melancholy
+    than madness "What he spake, though it lack'd form a little. Was not like madness."
+    \(Act 3, Scene i\)
+  - AI is a software.
+  - If people think that their car will kill them, it will not stop them from hacking
+    their car, even if it lowered their car's security. People tend to be less focused
+    on downstream consequences when dealing with immediate problems. Further, they
+    could justify it by saying that the probability that they would even be hacked
+    is incredibly low, and the marginal loss of security is worth it.
+- source_sentence: As race is socially constructed, it means different things to different
+    people. As a result, there is no clear line which courts can use to differentiate
+    what is and isn't whitewashing.
   sentences:
-  - The memory performance of students using laptops in the classroom is worse than
-    that of students not using laptops \(Hembrooke/Gay, p. 9; Wood et al, p. 369;
-    Risko et al., p. 281\).
-  - Surveys indicate that more than half of all students believe that they learn best
-    with the help of laptops and that they enjoy it the most compared to the use of
-    other devices or the use of no electronic devices at all \(Pearson, p. 29\).
-  - In no other area of medicine would it be acceptable to try and tout non-legitimate
-    treatments as viable options.
+  - For some, it would be whitewashing to have a lighter skinned black person portraying
+    a person who had a darker skin tone in real life. For others, this would not be
+    the case.
+  - Some goods may only be actualizable by allowing the possibility of evil.
+  - Energy for EVs has to be produced in often environmentally unfriendly power plants,
+    such as coal or gas.
 pipeline_tag: sentence-similarity
 library_name: sentence-transformers
 ---
 
 # SentenceTransformer based on sentence-transformers/all-mpnet-base-v2
 
-This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [sentence-transformers/all-mpnet-base-v2](https://huggingface.co/sentence-transformers/all-mpnet-base-v2) on the train and test datasets. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.
+This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [sentence-transformers/all-mpnet-base-v2](https://huggingface.co/sentence-transformers/all-mpnet-base-v2). It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.
 
 ## Model Details
 
@@ -79,9 +81,7 @@ This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [s
 - **Maximum Sequence Length:** 384 tokens
 - **Output Dimensionality:** 768 dimensions
 - **Similarity Function:** Cosine Similarity
-- **Training Datasets:**
-    - train
-    - test
+<!-- - **Training Dataset:** Unknown -->
 <!-- - **Language:** Unknown -->
 <!-- - **License:** Unknown -->
 
@@ -119,9 +119,9 @@ from sentence_transformers import SentenceTransformer
 model = SentenceTransformer("sentence_transformers_model_id")
 # Run inference
 sentences = [
-    "The use of laptops in classrooms impedes students' learning.",
-    'The memory performance of students using laptops in the classroom is worse than that of students not using laptops \\(Hembrooke/Gay, p. 9; Wood et al, p. 369; Risko et al., p. 281\\).',
-    'Surveys indicate that more than half of all students believe that they learn best with the help of laptops and that they enjoy it the most compared to the use of other devices or the use of no electronic devices at all \\(Pearson, p. 29\\).',
+    "As race is socially constructed, it means different things to different people. As a result, there is no clear line which courts can use to differentiate what is and isn't whitewashing.",
+    'For some, it would be whitewashing to have a lighter skinned black person portraying a person who had a darker skin tone in real life. For others, this would not be the case.',
+    'Energy for EVs has to be produced in often environmentally unfriendly power plants, such as coal or gas.',
 ]
 embeddings = model.encode(sentences)
 print(embeddings.shape)
@@ -171,53 +171,29 @@ You can finetune this model on your own dataset.
 
 ## Training Details
 
-### Training Datasets
+### Training Dataset
 
-#### train
+#### Unnamed Dataset
 
-* Dataset: train
-* Size: 25,880 training samples
-* Columns: <code>anchor</code>, <code>positive</code>, and <code>negative</code>
+* Size: 49,273 training samples
+* Columns: <code>sentence1</code>, <code>sentence2</code>, and <code>label</code>
 * Approximate statistics based on the first 1000 samples:
-  |         | anchor                                                                             | positive                                                                           | negative                                                                           |
-  |:--------|:-----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|
-  | type    | string                                                                             | string                                                                             | string                                                                             |
-  | details | <ul><li>min: 3 tokens</li><li>mean: 28.44 tokens</li><li>max: 186 tokens</li></ul> | <ul><li>min: 3 tokens</li><li>mean: 35.51 tokens</li><li>max: 148 tokens</li></ul> | <ul><li>min: 3 tokens</li><li>mean: 35.11 tokens</li><li>max: 169 tokens</li></ul> |
+  |         | sentence1                                                                          | sentence2                                                                          | label                                           |
+  |:--------|:-----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|:------------------------------------------------|
+  | type    | string                                                                             | string                                                                             | int                                             |
+  | details | <ul><li>min: 3 tokens</li><li>mean: 30.31 tokens</li><li>max: 126 tokens</li></ul> | <ul><li>min: 5 tokens</li><li>mean: 35.61 tokens</li><li>max: 180 tokens</li></ul> | <ul><li>0: ~54.80%</li><li>1: ~45.20%</li></ul> |
 * Samples:
-  | anchor                                                                                                                                                                                                                                                                                                     | positive                                                                                                                                                                                                                                   | negative                                                                                                                                                                                                                                                                                                 |
-  |:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
-  | <code>"Colorblindness" ignores the current realities of how someone's race affects them in daily life.</code>                                                                                                                                                                                              | <code>Minorities face disadvantages, solely due to their race, that majority races don't have to face.  These disadvantages cannot be ignored.</code>                                                                                      | <code>Out of scope from original stance. "Should society work towards...." meaning: is it a worthy end-goal. Outside of scope is current realities which this point is addressing.</code>                                                                                                                |
-  | <code>Taking fewer shifts at work may allow people to vote or attend a town hall or a school board meeting. This allows lower income employees time to be more politically engaged in their communities so that they are better able to represent their own interests in government. \(Berry 2012\)</code> | <code>Increased political engagement is good because it allows political forces to represent the people more easily. Increased political engagement will also help increase trust in government, leading to a better civil service.</code> | <code>Lower income employees lack political representation due to the lobbying power that being wealthy brings, not because they lack the time to be politically engaged.</code>                                                                                                                         |
-  | <code>The US has no long-standing practice of executing in an intentionally cruel manner and repeatedly phased out execution methods that were deemed inhumane.</code>                                                                                                                                     | <code>Electrocution was specifically adopted with the belief and intention that it would be more humane and less painful.</code>                                                                                                           | <code>88% of all executions in the United States since 1976 were conducted via lethal injections. It remains the primary method in all American states and for the federal government. This shows that America has de-facto a long-standing practice of executing in an avoidably painful manner.</code> |
-* Loss: [<code>TripletLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#tripletloss) with these parameters:
+  | sentence1                                                                                                                                                                                                                                                                                                   | sentence2                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                    | label          |
+  |:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------|
+  | <code>Pope Francis has tried to encourage Catholic priests to give communion to some divorced and remarried couples, or to families where unmarried parents are cohabiting. These statements have contributed to division among Catholics.</code>                                                           | <code>Pope Francis has stated that the church should apologise for its "blessing of many weapons", specifically against women, children, the LGBT community, and the poor.</code>                                                                                                                                                                                                                                                                                                                                            | <code>1</code> |
+  | <code>Large areas of Earth are either not economically viable or are outright inhospitable without expensive long term terraforming projects to render them viable for proper habitation. As the population grows, supporting said population becomes exponentially more expensive per square meter.</code> | <code>Most likely people will retreat inland instead of creating 'water worlds', because it is easier and populations are well-established there.</code>                                                                                                                                                                                                                                                                                                                                                                     | <code>1</code> |
+  | <code>Environmental and social wellbeing are ultimately in the best interests of businesses. Given enough freedom, businesses don't need to compromise society or the environment to thrive. Preserving these things will help them in the long run.</code>                                                 | <code>Nobody starts a business intending it to fail. They all must utilize sustainable business models. Any damage caused would be avoided through the government allowing companies what they need, while encouraging socially beneficial business practices, and enforcing individual liberty in the case a business attempts to infringe upon it. That’s as opposed to simply punishing those who step outside some arbitrary boundary with taxes and regulations, which kill otherwise perfectly good businesses.</code> | <code>1</code> |
+* Loss: [<code>ContrastiveLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#contrastiveloss) with these parameters:
   ```json
   {
-      "distance_metric": "TripletDistanceMetric.COSINE",
-      "triplet_margin": 0.3
-  }
-  ```
-
-#### test
-
-* Dataset: test
-* Size: 6,471 training samples
-* Columns: <code>anchor</code>, <code>positive</code>, and <code>negative</code>
-* Approximate statistics based on the first 1000 samples:
-  |         | anchor                                                                             | positive                                                                           | negative                                                                           |
-  |:--------|:-----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|
-  | type    | string                                                                             | string                                                                             | string                                                                             |
-  | details | <ul><li>min: 7 tokens</li><li>mean: 27.97 tokens</li><li>max: 116 tokens</li></ul> | <ul><li>min: 6 tokens</li><li>mean: 34.26 tokens</li><li>max: 124 tokens</li></ul> | <ul><li>min: 7 tokens</li><li>mean: 35.75 tokens</li><li>max: 134 tokens</li></ul> |
-* Samples:
-  | anchor                                                                                                                                                                                                                                                                                                                        | positive                                                                                                                                                                                                 | negative                                                                                                                                                                                                                                                                                               |
-  |:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
-  | <code>Theft only applies to instances involving the individual loss of property that then unjustly enriches another. Tax does not unjustly enrich any individual because it benefits \(and is necessary for\) society as a whole. Thus it is merely the just payment of a debt that all individuals owe to each other.</code> | <code>Taxation is required for the government to act in the economy, whether that be to provide infrastructure or essential services such as welfare.</code>                                             | <code>Such a collective act can still meet the definition of theft; for instance if a government authorises its agents to take items from another country \(as in cases of historic looting\). Just because a collective benefits from an action does not necessarily justify that action.</code>      |
-  | <code>Admitting more refugees can have adverse effects on host economies. US economic interests in Europe could subsequently be negatively impacted.</code>                                                                                                                                                                   | <code>Admitting more refugees could increase the population of low skilled laborers. By the logic of the Stopler-Samuelson theorem, this would drive wages down for citizens of the host country.</code> | <code>Approximately half of the jobs available to workers in Europe can be automated mckinsey.com. Since technological advancement has greater potential for job displacement than refugee migration, policy makers should be wary of citing economic reasons against expanding refugee intake.</code> |
-  | <code>Private universities have historically been institutions of innovation in education, forcing them into an unfair marketplace would be a mistake.</code>                                                                                                                                                                 | <code>Public universities are controlled by the state making politically controversial courses of study more difficult.</code>                                                                           | <code>Plenty of innovation in education comes from public universities, they make up many of the examples of innovation cited in this report.</code>                                                                                                                                                   |
-* Loss: [<code>TripletLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#tripletloss) with these parameters:
-  ```json
-  {
-      "distance_metric": "TripletDistanceMetric.COSINE",
-      "triplet_margin": 0.3
+      "distance_metric": "SiameseDistanceMetric.COSINE_DISTANCE",
+      "margin": 0.5,
+      "size_average": true
   }
   ```
 
@@ -347,30 +323,42 @@ You can finetune this model on your own dataset.
 ### Training Logs
 | Epoch  | Step  | Training Loss |
 |:------:|:-----:|:-------------:|
-| 0.1236 | 500   | 0.2134        |
-| 0.2473 | 1000  | 0.1983        |
-| 0.3709 | 1500  | 0.1925        |
-| 0.4946 | 2000  | 0.1868        |
-| 0.6182 | 2500  | 0.1808        |
-| 0.7418 | 3000  | 0.1722        |
-| 0.8655 | 3500  | 0.166         |
-| 0.9891 | 4000  | 0.1682        |
-| 1.1128 | 4500  | 0.1304        |
-| 1.2364 | 5000  | 0.1239        |
-| 1.3600 | 5500  | 0.1224        |
-| 1.4837 | 6000  | 0.1248        |
-| 1.6073 | 6500  | 0.1204        |
-| 1.7310 | 7000  | 0.1191        |
-| 1.8546 | 7500  | 0.1142        |
-| 1.9782 | 8000  | 0.1169        |
-| 2.1019 | 8500  | 0.0802        |
-| 2.2255 | 9000  | 0.0641        |
-| 2.3492 | 9500  | 0.0662        |
-| 2.4728 | 10000 | 0.0696        |
-| 2.5964 | 10500 | 0.0624        |
-| 2.7201 | 11000 | 0.0617        |
-| 2.8437 | 11500 | 0.0641        |
-| 2.9674 | 12000 | 0.06          |
+| 0.0812 | 500   | 0.0309        |
+| 0.1623 | 1000  | 0.0281        |
+| 0.2435 | 1500  | 0.0279        |
+| 0.3247 | 2000  | 0.0272        |
+| 0.4058 | 2500  | 0.0257        |
+| 0.4870 | 3000  | 0.0272        |
+| 0.5682 | 3500  | 0.0265        |
+| 0.6494 | 4000  | 0.0261        |
+| 0.7305 | 4500  | 0.0252        |
+| 0.8117 | 5000  | 0.0255        |
+| 0.8929 | 5500  | 0.0261        |
+| 0.9740 | 6000  | 0.0254        |
+| 1.0552 | 6500  | 0.023         |
+| 1.1364 | 7000  | 0.0223        |
+| 1.2175 | 7500  | 0.0219        |
+| 1.2987 | 8000  | 0.0209        |
+| 1.3799 | 8500  | 0.0213        |
+| 1.4610 | 9000  | 0.0205        |
+| 1.5422 | 9500  | 0.0209        |
+| 1.6234 | 10000 | 0.0205        |
+| 1.7045 | 10500 | 0.0205        |
+| 1.7857 | 11000 | 0.0201        |
+| 1.8669 | 11500 | 0.0203        |
+| 1.9481 | 12000 | 0.0196        |
+| 2.0292 | 12500 | 0.0171        |
+| 2.1104 | 13000 | 0.0148        |
+| 2.1916 | 13500 | 0.0139        |
+| 2.2727 | 14000 | 0.0145        |
+| 2.3539 | 14500 | 0.0138        |
+| 2.4351 | 15000 | 0.0138        |
+| 2.5162 | 15500 | 0.0144        |
+| 2.5974 | 16000 | 0.0134        |
+| 2.6786 | 16500 | 0.0138        |
+| 2.7597 | 17000 | 0.014         |
+| 2.8409 | 17500 | 0.0133        |
+| 2.9221 | 18000 | 0.0136        |
 
 
 ### Framework Versions
@@ -399,15 +387,17 @@ You can finetune this model on your own dataset.
 }
 ```
 
-#### TripletLoss
+#### ContrastiveLoss
 ```bibtex
-@misc{hermans2017defense,
-    title={In Defense of the Triplet Loss for Person Re-Identification},
-    author={Alexander Hermans and Lucas Beyer and Bastian Leibe},
-    year={2017},
-    eprint={1703.07737},
-    archivePrefix={arXiv},
-    primaryClass={cs.CV}
+@inproceedings{hadsell2006dimensionality,
+    author={Hadsell, R. and Chopra, S. and LeCun, Y.},
+    booktitle={2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)},
+    title={Dimensionality Reduction by Learning an Invariant Mapping},
+    year={2006},
+    volume={2},
+    number={},
+    pages={1735-1742},
+    doi={10.1109/CVPR.2006.100}
 }
 ```
 

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3e8a5771689729abff030256879319e0e8d174d9c026410d73bccf568ed0ae10
+oid sha256:36e88c6340b3eb3a8fdb0db2c1aa25591bc89c86bd3ca79da7bff4b6b7bfea28
 size 437967672

runs/Feb16_03-04-22_fd002087ebe1/events.out.tfevents.1739675063.fd002087ebe1.667.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab201d847dd13f6250fda40af30ceaddf79f49215a7d43974496d0259fad3ab6
+size 12347

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:286d7f67a76e9bf3a9c2f88dc037151c17c7bbd90f6450a7cc4c1eb942163d77
+oid sha256:20d502cd8247c5e637c7f4cf273aee02cb9cef7a21046455af267e6b7a831aa7
 size 5560