m7n/nomic-embed-philosophy-triplets_v9

SentenceTransformer based on nomic-ai/nomic-embed-text-v1

This is a sentence-transformers model finetuned from nomic-ai/nomic-embed-text-v1. 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

Model Description

• Model Type: Sentence Transformer
• Base model: nomic-ai/nomic-embed-text-v1
• Maximum Sequence Length: 8192 tokens
• Output Dimensionality: 768 tokens
• Similarity Function: Cosine Similarity

Model Sources

• Documentation: Sentence Transformers Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 8192, 'do_lower_case': False}) with Transformer model: NomicBertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Normalize()
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("m7n/nomic-embed-philosophy-triplets_v9")
# Run inference
sentences = [
    'compare the remarks in On Habituations (Eth.: this does not appear in the Kuhn edition; it is edited by Muller in SM 2 9-31): As the hegemonic soul has capabilities (dunameis) directed towards all the technai, it is necessary that there is one (sc. dunamis) with which we understand consequence and conflict, and another with which we remember; and we are cleverer in respect of the first mentioned, but more retentive in respect of the second. (Eth. 4, = SM 2 25) The distinction between intellectual sharpness and retentive ability is an ancient commonplace: see in particular the Hippocratic text On Regimen 1 35. 12 PHP V 601; cf. Symp.Diff. VII 55-6; and see MM X 635-6: a second part of the soul belongs to us not in virtue of our growing or being alive, but because we are animals, it is located in the heart and is the source of the innate heat; the arteries are the conduits for this source, which has many names: it is called the living power (dunamis z6tike), the spirited power (dunamis thumoeides), the living soul, and the spirited soul. 200 So Galen draws on a diverse range of previous theories in order to construct his own account; but it is important to stress that the theory that results is no mere haphazard porridge of badly-digested and ill-assorted scraps from the table of his predecessors. He is not afraid to take issue with them on matters of substantial importance, and to take issue with them in his',
    'own characteristically, indeed uniquely, polemical style. For instance, he will not say, as Plato does, that any part of the soul is demonstrably immortal. Indeed, in his pronouncements on the matter he exhibits an admirable caution, unwilling to commit himself with any degree of certainty on matters which he views as being by their very nature resistant to secure demonstration."3 Thus at PHP V 791-2, Galen writes: Plato said that the cause who made us, the demiurge who fashioned the universe, commanded his children to make the human race by taking . .. the substance (ousia) of the immortal soul from him and adding it to what was generated. But we must realise that there is no formal similarity between proving and positing the fact that we were made in accordance with the providence of some god . . ., and knowing the substance of the maker or even of our own soul ... [T]he statements of the most divine Plato about the substance of our soul . .. and still more all that he says about our whole body, extend only to plausibility and reasonableness (achri tou pithanou kai eikotos). That last remark is important (indeed Galen, perhaps excessively charitably, takes Plato himself to be committed to it by his remarks about the eikos muthos [Tim. 29c-d]). There is a class of things about which we can at best speculate, and most particularly these are the preserve of the philosophers: In philosophy it is not surprising',
    "Hippocrates and Plato but in spite of its title, this has virtually nothing in common with the placita of Aëtius, nor, one presumes, with the phusikai doxai of Theophrastus. Galen's avowed aim is to show, by means of selective quotation and juxtaposition, the fundamental agreement between his two great masters of his title, in particular in regard to the structure and function of the soul, and at the expense in particular of the Stoics and Peripatetics. Mario Vegetti contributes a lucid and learned piece,5 which seeks to understand the work in terms of its intended audience avowedly not the sectarians of either medicine or philosophy, but rather generally educated and interested men of good will competent in the rudiments of logic; and he supposes that the wealth of quotation in which Galen indulges, even from works which he himself says are still readily available, indicates that his audience is not supposed to be made up of specialists. This may be right; but Galen loved to quote and to parade his learning, and while it is true that in this work the citations are longer than usual, this may simply be due (as Galen himself says) to the verbosity of those (principally Chrysippus) he intends to refute. Sooner or later, everyone who reads Galen (or even simply makes use of him as a source for other writers) has to get to grips with the question of how careful and trustworthy a reporter of other people's views he is and different scholars have proposed rather different",
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Triplet

• Dataset: nomic
• Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.974
dot_accuracy	0.026
manhattan_accuracy	0.964
euclidean_accuracy	0.974
max_accuracy	0.974

Triplet

• Dataset: nomic
• Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.9755
dot_accuracy	0.0245
manhattan_accuracy	0.976
euclidean_accuracy	0.9755
max_accuracy	0.976

Training Details

Training Dataset

Unnamed Dataset

• Size: 10,000 training samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 271 tokens mean: 331.92 tokens max: 584 tokens	min: 265 tokens mean: 331.75 tokens max: 565 tokens	min: 267 tokens mean: 330.76 tokens max: 518 tokens

• Samples:

  anchor	positive	negative
  a mutual belief among all the pushers that I also am taking part in the pushing. This kind of joint activity should be regarded as a case of coaction (or a very weak kind of joint action) rather than "proper" joint action (see Tuomela, 1992, Chapter 2 for discussion). 3. Joint Intention As is generally accepted, a (mere) personal intention involves making up one's mind. Similarly, joint intentions such as we-intentions (to be discussed below) involve the participants' having jointly resolved or made up their minds or made a joint plan concerning what to do jointly. As was remarked above, "proper" joint actions are based on (at least mutually believed) agreement, either explicit or implicit. It should be obvious that if we-intentions are to have proper joint actions as their "satisfiers", they must also be based on agreement-making. But, as WHAT IS COOPERATION? 89 said, we will accept below that joint intentions can concern also joint actions in a wider sense. In joint-intention formation each agent accepts for himself: "I ought to participate in our doing X together". This acceptance here means that the agent (at least dimly) recognizes the existence of a joint plan to perform X and accordingly commits himself to performing X together with the others. A joint intention on conceptual grounds leads to each agent's acceptance of "I will participate in, or contribute to, our doing X", based on his acceptance of "We will do X" (the standard ex? pression for	joint intentions or "group-intentions", viz, we-intentions and standing group-intentions, as argued in Tuomela, 1992). Let us consider the central notion of we-intention in some more detail. We-intentions are action-generating joint intentions that agents have in situations of joint action, e.g., when they jointly intend to carry a table jointly. The content of a we-intention can be taken to be something like "to do X jointly" or "we to do X jointly". A we intention involves the intention to perform one's part of the joint action. We can say roughly that a member A? of a collective G ("we" for At) we-intends to do X if and only if At (i) intends to do his part of X (as his part of X), (ii) has a belief to the effect that the joint action opportunities for an intentional performance of X will obtain; and, furthermore, (iii) believes that there is (or will be) a mutual belief among the participating members of G or at least among those participants who do their parts of X intentionally as their parts of X to the effect that the joint action opportunities for an intentional performance of X will ohtain.1 Next consider a schema of practical reasoning that a we-intending agent is required to satisfy. This schema also serves to account for the commitments the participants of a joint action have towards other participants (cf. Tuomela, 1984, Chapter 2, Tuomela and Miller, 1988, and Tuomela, 1992, Chapter 3): (W)(i) We will do X (ii) X cannot be performed by us	action is to give a central place to the appropriate kind of intention that is relevant to collective action. This kind of intention is commonly called a 'collective', 'joint', or 'group' intention, and accounts of this form of intention have been offered by Michael Bratman [5], Margaret Gilbert [13, 12, 14], Raimo Tuomela [24, 25, 30, 23, 28, 27, 26], John Seaxle [19, 18], Seamus Miller [16], J. David Velleman [32], and others.3 Those who follow the strategy of analyzing collective action by giving an analysis of collective intention4 typically assume that once the analysis of collective intention is in place, the analysis of collective action will follow immediately.5 One challenge facing an account of collective intention is that there is an extremely wide variety of collective intentions, which may have various necessary and sufficient conditions, depending upon the specific circumstances. For example, we may correctly say that 'Russell and Whitehead had the intention to write the Principia Mathematics and that 'the angry mob had the intention to storm the Bastille'. But the level of coordination, planning, and 3Elsewhere, we have offered a game-theoretic characterisation of the concept of collective intention [9]. 4 In what follows, we shall use the term 'collective intention' neutrally to refer to the entire range of intentions attributable to groups. Such intentions have been called 'group intentions', 'joint intentions', and 'we-intentions'. 5 The strategy of analyzing collective action by giving an account of collective
  unrestrictedly into its service. Characteristic of him was a certain harshness and a dogmatism rooted in a sense of personal righteousness and moral integrity; characteristic also was his sense of a high calling, a selfconsciousness suffused with pride and "heroism", and, along therewith, a certain lack of social imagination such as is requisite for an understanding of other people and for happy relations with them. Thus he was frequently unheeding whether of the feelings or of the 'Theodor Fritsch, Handbuch der Judenfrage: Die wichtigsten Tatsachen zur Beurteilung des jfidischen Volkes, 39th ed. (I8I to 200 thousand), Leipzig, Hammer-Verlag, 1935, pp. io and 5V8, respectively. (In view of recent political events in Germany it should be noted that, while the 39th edition of Fritsch's work was printed in 1935, the author died on September 8, 1933.) No. I.] Discussion rightful claims of others, and insensitive to the real value of their ideas and points of view.Jn consequence his exaggerations could reach a degree all but incredible. Those of his statements which were made in moments of emotional fervor or of an overpowering moral zeal must not be centered upon as crucial if one would acquire genuine knowledge as to that most comprehensive and tightly organized body of ideas and values which represent the real Fichte and constitute the soul of his philosophy. In interpreting Fichte it is important to remember also that his career was as turbulent as was his temperament. Moreover, the intellectual and cultural	influences that came in turn to play upon him during the years of his intense life were very diverse, and during this period epochal political and social events reached a culmination in the French Revolution and the Napoleonic conquests. Not strange, then, that even after the transformation which a study of the philosophy of Kant effected in Fichte's thought and objectives, there came further changes throughout the whole range of his ideas, whether social, political, ethical, religious, or metaphysical. To be sure it has been said of Fichte that the man "aus einem Gusz" inevitably propounded a philosophy "aus einem Stuck". If this means that Fichte, in comparison, for example, with Kant, was prompted by his very temperament, as well as by his conception of the requirements of reason, to exhibit the whole range of human thought and values as a single system, the statement is essentially true. Nevertheless it should be remembered that such unity as prevails in the Fichtean writings characterizes them severally rather than as a whole; and also that the unity in question is of the sort generated by a deep moral and religious passion rather than that issuing from a speculative imagination deployed by "cold thought" or the semblant activities of play or art. The earliest intimation we have of Fichte's attitude toward Jews and Judaism is in a sermon preached by him in Dubrenski on March 25, 1786, and thus in the twenty-fourth year of his life.4 After having contended that "God gives to all men certain incentives to	in that work. The essays have all the brilliance and learning that one associates with all of Professor Cassirer's work, and the English of the translation bears no marks of the German source. This doctoral dissertation undertakes to expound in some seventy pages the conception of God or the Absolute in the philosophy of Fichte. "Fichte has often been thought to have expressed himself so loosely that his precise meanings were hidden in vague generalities," Dr. Stine remarks. "A careful study of the works of Fichte, however, reveals that while the terminology varies, he was constantly seeking to express the same fundamental thoughts." There can be little doubt, on the evidence supplied in this thesis, that Dr. Stine has made a careful study of the works of Fichte. But while he may have succeeded in discerning the same fundamental thoughts behind the vague generalities, he has not succeeded in communicating them, to this reviewer at least. This is partly no doubt because the Idea of God, or the Absolute, is so central to Fichte's whole tortuously conceived philosophy that the attempt to expound it systematically in this brief compass was hazardous at best. A further difficulty is found in the method of exposition employed. Dr. Stine depends far too extensively on long quotations from Fichte to make his meaning plain, and these quotations are frequently more puzzling than illuminating. The terminology does indeed vary, and no effective attempt is made to chart any clear path through the tangled
  To see why societies found it necessary to develop IP laws, consider property rights with respect to a book. In a world without copyright laws, if an author produces a book manuscript, he can only own the actual pages of the manu script and has no property rights to the text itself. When the author sells his book manuscript to the publisher, he has transferred all of his rights to the publisher with respect to the book. Likewise, if the publisher sells a copy of the book, he has transferred all of his rights to the buyer. Similar problems would arise for inventions in a world without patent laws and trademarks in a world without trademark laws. It would seem that people have important interests worth protecting that extend beyond their interests in tangible things; they also have interests in intangible things, such as the ideas written on paper or the information embodied in inventions. In order to recognize these interests in ideas and infor mation, it became necessary to create a new category of property, intellectual property.5 Intellectual properties, unlike tangible proper ties, require special legal protection because they are non-exclusive: two people can possess and use the same item of intellectual property without preventing each other from possessing or using it (Hettinger, 1989). For example, two people can both use and possess the Pluralistic Account of Intellectual Property same computer program, poem, manufacturing method, or the same song at the same time. IP laws allow people to gain exclusive control over objects that are non-exclusive (May, 1998). The reason why intellectual	properties are non exclusive is that information and ideas have no particular location in time and space: they are abstract objects.6 Although some writers, such as Barlow (1994), believe that it is impossible to protect IP in the information age, due to its ease of transmission, businesses and governments have developed many technological and legal methods for protecting IP, such as access restrictions, encryption, water-marking, licensing agreements, and infringement lawsuits (NAS, 2000). The most common legally recognized forms of IP are copyrights, patents, trademarks, and trade secrets (Miller and Davis, 2000; Foster and Shook, 1993). Many different countries have enacted laws and signed international treaties that define these rights and their scope. All of these laws distinguish between privately owned infor mation and information that is in the public domain. Copyrights give authors of original works the right to exclude others from copying those works without the author's permission; patents give inventors the right to exclude others from making, using, or commercializing their inventions without the inventor's permission; trademarks give the trademark holder the right to exclude others from using a symbol that he uses to distinguish his business or its products; and trade secrecy laws allow businesses to protect confidential business information (Paine, 1991). There are other types of IP in addition to these four traditional categories. Confidential, personal information, such as medical records or psychiatric records, financial data, credit reports, or purchasing data also can	on the analogy with tangible property and that recognizes the role of ideas in defining personality and social relationships. The combined effect of these assumptions is that trade secret law comes in for particularly serious criticism. It restricts methods of acquiring ideas (p. 35); it encourages secrecy (p. 36); it places unacceptable restrictions on employee mobility and technology transfer (p. 52); it can stifle competition (p. 50); it is more vulnerable to socialist objections (p. 52). In light of these deficiencies, Hettinger recommends that we consider the possibility of "eliminating most types of trade secrets entirely and letting patents carry a heavier load" (p. 49). He believes that trade secrets are undesirable in ways that copyrights and patents are not (p. 36). 6. Hettinger mentions trademark as another of our intellectual property institutions, along with our social sanction on plagiarism, but his central discussion focuses on copyright, patent, and trade secret concepts. Neither trademark principles nor the prohibition on plagiarism fits comfortably with his justification in terms of increasing the dissemination and use of ideas. Both are more closely related to giving recognition to the source or originator of ideas and products. 7. It may be helpful to think of two levels of justification: (i) an intermediate level consisting of objectives, purposes, reasons, and explanations for an institution or practice; and (2) an ultimate level linking those objectives and purposes to our most basic legitimating ideas such as the general good or individual liberty. Philosophers generally tend to be concemed with the

• Loss: TripletLoss with these parameters:

  {
      "distance_metric": "TripletDistanceMetric.COSINE",
      "triplet_margin": 0.05
  }
  

Evaluation Dataset

Unnamed Dataset

• Size: 500 evaluation samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 278 tokens mean: 331.45 tokens max: 489 tokens	min: 279 tokens mean: 332.39 tokens max: 468 tokens	min: 280 tokens mean: 330.43 tokens max: 471 tokens

• Samples:

  anchor	positive	negative
  with accepted beliefs, provided only there is enough of sentimental satisfaction in it to -compensate. If all the values which ideas may have are to count for truth and be simply summed and subtracted, then in any given .ease disproof by scientific or logical methods may be overbalanced by positive values of " subsequential utility " or tonic emotionality. On this precise point Dr. Schiller has nothing to say. He simply reiterates the " biological necessity " that all the idea = values shall determine the idea's "survival". This, so far as I know, has never been denied. The question is whether an idea that was ,contrary to sensible experience might not survive because its emotional, value outweighed the dissatisfaction at its contrariety to sensible experience; and whether therefore such an idea might not be " true " on Dr. Schiller's theory. Dr. Schiller accuses me of having attempted to " read a metaphysical meaning into a number of, pragmatic pronouncements which are clearly methodological ". But the texts from which I have cited in arguing that pragmatism of the Schiller type is subjectivistic bear such titles as The Ethical Basis of Metaphysics, 1 MIN, No. 86. Philosophy -and the Scientific TIzvestigation of a Future Life, and The Making of Reality. No pragmatist that I have ever read has confined either himself or his pragmatism to "methodological" considerations; and it is perfectly clear that Dr. Schiller himself does not mean to do	so, for he proceeds at once to present a third alternative which may save him from the necessity of deciding between idealism and realism.' This third possibility which I am accused of " ignoring " is "'the correlation of a mind-with-objects and objects-for-a-mind ". I have not ignored this possibility. I have recognised it as a very common formulation of idealism.2 The reader will observe that in the above formula the only constant is mind, or a relationship distinguished by the fact that mind must always be one of its terms. But waiving this point, how is one to discover the real importance of this " correlation " to whatever may take the place of object in it. How is one to determine the real, as distinguished from the methodological place of mind in the world? This is a fairly important question and it is the question at issue between idealism and realism. I cannot believe that it is "merely academic " and must therefore crave Dr. Schiller's indulgence further. Now as to "the Ego-centric Predicament ". Dr. Schiller admits that I give "much prominence " to it. But he must have discovered that fact by consulting the Index or the Table of Contents. 3 For he has not in the least understood the point, and most of what he attributes to me is flatly contradicted by the text. He suggests that I infer realism from the Ego-centric Predicament, whereas I. have invariably asserted that nothing can be argued from it. My central point is that it is	the name of idealism; it is also evident that they have really adopted the form and presuppositions of empirical realism, in spite of the 'mental' categories in which they have been expressed. Idealism has no more to do with 'ideas' interpreted merely as mental states than it has with any other type of 'independent' particulars; and the essence of idealism does not consist in the peculiar spiritual or mental character of what it regards as real, so much as in the element of universality which the real must exhibit whether it be mental or material, or whatever may be the subjective and particularist characters that may belong to it. A certain degree of 'independence' of abstract mind is therefore the condition of an element's being real, in the sense that its reality implies the possession of the character of objectivity upon which the truth of any judgment with respect to it rests. But this independence is not so much a matter of unique disparateness, as the author seems to imply, as of its embodying certain aspects of universality. But this merely suggests that, in the order of the real, the peculiarly mental fact, the state of consciousness in its particularist and subjective sense, holds a not very important place. Professor Moore's refutation would therefore have been important if it had relegated the mental state to its properly obscure place; but that would have contributed, not to a refutation of idealism, but to the necessity of a re-examination of
  to Hesse therefore, must argue for this alternative descriptions thesis. Her strategy for establishing both conditions for realism is straightforward. She argues that neither competing theories nor alternative formalizations need be interpreted as being about different entities. With respect to the historical challenge, it is always possible, she claims, to translate a large number of the statements of an earlier theory into the statements of its historical successor ([2], p. 299). Similarly, with respect to the conceptual challenge, she claims that the statements of one formalization can be translated into the statements of the alternative formalization ([2], p. 296). The claim in both cases, however, is that partial or complete translation between theories or alternative formalizations indicates that the opposing theories are about the same world and that, rather than being ontologically incompatible, they are much like alternative natural languages which describe the same world. Inter-translatability is thus a mark of identity of reference. Translation is clearly the cornerstone of Hesse's realism and she exerts a strong effort in explaining the sense in which the statements within both historical theories and alternative formalizations can be translated. In both cases, she claims that the translation is achieved by recognizing certain identities of intensional reference ([2], pp. 296, 299) (henceforth IR) within the respective theories or formalizations. Her realism presupposes this translational concept and I will argue that despite the care that she spends in constructing it, there are enough	problems within it to justify its rejection. In the light of these problems, I will then attempt to redesign the concept so that it will allow for a realistic attitude toward science. The notion of IR is complex. One tempting method for explicating it would be to first lay out Hesse's definition of it and then clarify the concept via a detailed analysis of this definition. This route, however, is not a fruitful one; for although she states that Intensional reference is the relation which subsists between a descriptive predicate in a given language and a property of 446 REALISM AND INTENSIONAL REFEREN an object when the statement ascribing that predicate to that object is true. ([2], p. 62) she does nothing to clarify the myriad of epistemological and metaphysical problems imbedded within this definition. Furthermore, her own presentation of the concept of IR does not focus around this definition. Instead she begins by employing the via negativa; i.e., she distinguishes IR from a number of other classically semantic concepts. For although IR captures the meaning of many general terms used within science, she claims that this meaning cannot be identified either with the extension or the sense of those general terms. According to Hesse, IR cannot be identified with extension. For despite the fact that scientific theories do discriminate objects into classes, these classes are not adequately described as extensional, for they are not defined merely by the objects contained in them. They involve also what I shall call intensional reference, that is, they depend on recognitions of similarities and differences	the instrumental reliability of the methodology which scientists actually employ. 7 and According to the realist, existing theories provide approximate knowledge not only of relations between observables, but also of the unobservable structures which underlie obserbable phenomena. 8 The crucial link between the "unobservable structures" which the instrumentalist or strict empiricist does not want to buy and Boyd's insistence that one is forced into buying them is, I argue, a theory of reference. For Boyd seems to believe that the referent of a key term in a scientific theory is seldom or never ( ] , and, more importantly, that vocabularies employing different terms are employing the terms co-referentially despite what may appear, to the casual observer, to be differences in meaning. Now it seems clear that what Boyd has in mind as a theory of reference is, baldly, a causal theory of reference. Crudely speaking, a causal theory of reference would allow for the retention of extension over a period of time for natural kind terms, and would do the kind or work that Boyd would need to preserve the epistemic access which he claims to be a condition for epistemic success. At this point one would do well to remind oneself of the salient differences between causal theories of reference and what is referred to in the literature as the "classical" view. The latter is generally regarded as the direct descendant of a view associated with Frege and to some extent with Russell. On this view
  the logicistic crusade for clarity are revealed in the writings of Whitehead's former colleague and co-author of Principia Mathematica, Bertrand Russell, whose theory of logical atomism directly confronts the problem of vagueness. Russell goes a step beyond Frege and argues that the vagueness of general terms can be conquered using the symbolic methods of Principia Mathematica and the established results of exact science.3 But the upshot of his frontal assault on vagueness merely underscores a crucial question: just what do the precise symbolic structures of logic and mathematics, and the formal calculations of the exact sciences, actually contribute to the rationality of explanations? Russell's treatment of vagueness does, however, help bring out one important point. He says, for example, that . . . what we believe ourselves to know in philosophy is more doubtful than the detail of science, though perhaps not more doubtful than its most When one compares this remark with his claim that it would be a great mistake to suppose that vague knowledge must be false. On the contrary, a vague belief has a much better chance of being true than a precise one, because there are more possible facts that would verify it, (Russell 1923, p. 91) the question arises whether there are any instances of precise knowledge that do not embrace any extremely vague ideas. Perhaps the greatest degree of certainty and truth that so-called exact science proffers philosophy comes only in the form of very general, and hence extremely vague, facts about	nature. Russell does not pursue this line of thought, however, for he chooses to tie all the tricky questions associated with the big question of what logic and science really do for philosophy into one neat logicistical package. He thus begs the primary question of what an ideally rational explanation ought to look like. Logical reconstructions of certain fundamental notions, such as matter and mind, are required, he maintains, just because these vague notions are not amenable to the methods of exact science. Apart from the circularity of this move, it is also worth noting that despite his suspicion of ordinary concepts Russell enlists the general idea of event here, thus forcing into the open the question whether this fundamental notion may be intrinsically vague. As quantum mechanics (which is often touted as one of the most successful of modern physical theories) now assures us, it is not possible to achieve a precise, purely objective description of any physical event.4 Hence in so far as the notion of event is fundamental both to science and philosophy, the principal lesson for the rationalistic philosopher who wishes, like Russell, to take science seriously may be that vagueness is an inescapable element of every rational account of anything whatever, and not a disease of the understanding in need of a rigorous logicistical therapy. That it may be both simplistic and misleading to view the chief aim of philosophy as primarily a quest for perfect clarity and definiteness in fundamental concepts and their modes of	phenomena as "understood". In this way I argue that the quantum revolution should not be seens as implying an "endgame for understanding" but as an opening move in a process of axiological revision which would allow the rational inquirer to discourse on the understanding of quantum phenomena without "merely redefining terms to paper over our ignorance". (Cushing, 1991, p. 337) Thus to be a realist is to take up a certain stand with respect to the axiology of inquiry: the realist upholds the cognitive goal of "understanding". The pursuit of any cognitive goal is rational or irrational only against the context of other beliefs held by a rational inquirer. Thus, the goals it is rational for a scientist to pursue are at least partially a function of that inquirer's conception of the physical world. Thus axiological revision may well entail ontological reconstruction, a task that falls in what I call the "philosophy of nature". The analysis presented here will be organized around three central ontological reference points: properties, individuals, and relations, to each of which a section of this paper is dedicated. However, before turning to these topics, a few general points about the relation of ontology to natural science are in order. Realism , And The Philosophy Of Nature The scientist's felt need for an understanding of quantum phenomena is akin to the philosopher's well-known need to be delivered from puzzlement; some philosophers seek to solve the problem, while others try to dissolve it. The realist administers therapy by attempting to satisfy the yearning

• Loss: TripletLoss with these parameters:

  {
      "distance_metric": "TripletDistanceMetric.COSINE",
      "triplet_margin": 0.05
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 4
• per_device_eval_batch_size: 4
• learning_rate: 1e-05
• weight_decay: 0.01
• warmup_ratio: 0.1
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 4
• per_device_eval_batch_size: 4
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 1
• eval_accumulation_steps: None
• learning_rate: 1e-05
• weight_decay: 0.01
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 3
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.1
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• logging_nan_inf_filter: True
• save_safetensors: True
• save_on_each_node: False
• save_only_model: False
• restore_callback_states_from_checkpoint: False
• no_cuda: False
• use_cpu: False
• use_mps_device: False
• seed: 42
• data_seed: None
• jit_mode_eval: False
• use_ipex: False
• bf16: False
• fp16: False
• fp16_opt_level: O1
• half_precision_backend: auto
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• local_rank: 0
• ddp_backend: None
• tpu_num_cores: None
• tpu_metrics_debug: False
• debug: []
• dataloader_drop_last: False
• dataloader_num_workers: 0
• dataloader_prefetch_factor: None
• past_index: -1
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: False
• ignore_data_skip: False
• fsdp: []
• fsdp_min_num_params: 0
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• fsdp_transformer_layer_cls_to_wrap: None
• accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch
• optim_args: None
• adafactor: False
• group_by_length: False
• length_column_name: length
• ddp_find_unused_parameters: None
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• dataloader_pin_memory: True
• dataloader_persistent_workers: False
• skip_memory_metrics: True
• use_legacy_prediction_loop: False
• push_to_hub: False
• resume_from_checkpoint: None
• hub_model_id: None
• hub_strategy: every_save
• hub_private_repo: False
• hub_always_push: False
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• include_inputs_for_metrics: False
• eval_do_concat_batches: True
• fp16_backend: auto
• push_to_hub_model_id: None
• push_to_hub_organization: None
• mp_parameters:
• auto_find_batch_size: False
• full_determinism: False
• torchdynamo: None
• ray_scope: last
• ddp_timeout: 1800
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• dispatch_batches: None
• split_batches: None
• include_tokens_per_second: False
• include_num_input_tokens_seen: False
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• eval_on_start: False
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss	nomic_max_accuracy
0	0	-	-	0.92
0.04	100	0.0082	0.0093	0.926
0.08	200	0.0078	0.0083	0.926
0.12	300	0.0077	0.0076	0.934
0.16	400	0.0055	0.0067	0.944
0.2	500	0.0045	0.0060	0.954
0.24	600	0.008	0.0055	0.956
0.28	700	0.0044	0.0048	0.966
0.32	800	0.0057	0.0056	0.958
0.36	900	0.0033	0.0053	0.958
0.4	1000	0.0038	0.0051	0.958
0.44	1100	0.0033	0.0062	0.958
0.48	1200	0.0032	0.0057	0.95
0.52	1300	0.0038	0.0055	0.962
0.56	1400	0.0038	0.0048	0.964
0.6	1500	0.0048	0.0047	0.962
0.64	1600	0.0026	0.0047	0.966
0.68	1700	0.0033	0.0051	0.962
0.72	1800	0.0039	0.0054	0.962
0.76	1900	0.0028	0.0048	0.966
0.8	2000	0.0042	0.0046	0.97
0.84	2100	0.0043	0.0044	0.968
0.88	2200	0.0038	0.0044	0.968
0.92	2300	0.0032	0.0040	0.97
0.96	2400	0.0034	0.0042	0.97
1.0	2500	0.0041	0.0045	0.964
1.04	2600	0.002	0.0042	0.966
1.08	2700	0.0023	0.0039	0.97
1.12	2800	0.0027	0.0040	0.968
1.16	2900	0.0005	0.0040	0.972
1.2	3000	0.0002	0.0043	0.968
1.24	3100	0.0004	0.0042	0.966
1.28	3200	0.0002	0.0041	0.964
1.32	3300	0.0003	0.0042	0.97
1.3600	3400	0.0004	0.0040	0.968
1.4	3500	0.0002	0.0040	0.974
1.44	3600	0.0005	-	0.976

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.0.1
• Transformers: 4.42.4
• PyTorch: 2.3.1+cu121
• Accelerate: 0.32.1
• Datasets: 2.21.0
• Tokenizers: 0.19.1

Citation

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}

TripletLoss

@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}
}