m7n/nomic-embed-philosophy-triplets_v3

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_v3")
# Run inference
sentences = [
    'reflective conceptual scheme of the the observer\'s world. So, even though we can differentiate by analysis the conditions of our awareness, there is no actual awareness short of conceptual incorporation into a functional and rational world. Not only is there no thinking without signs; there is no experience short of thirdness. This does not imply that such experiences are whole and complete relative to consciousness. That for both Hegel and Peirce never happens short of the Absolute ? however different their visions ofthat Absolute may have been. What it does mean is that there is no direct access to any reality as it presents itself either the Cartesian mental or the Cartesian physical, that can warrant any claims for an epistemological direct realism. That thought is always mediated by sym bols gives the basis for the "critical" in Peirce\'s own critical common-sen sism, and puts him in line with Santayana and Sellars (perhaps Wilfred as well as Roy Wood) for critical realism. While this is the case epistemologically, Peirce wanted to maintain that such an account requires ontologically the reality of all three pheno menological categories. He chided Hegel for dwelling on thirdness to the exclusion of secondness and firstness, treating the triadic form as "a mere RECONSTRUCTING DECONSTRUCTION 377 fashion of dress." "Had Hegel, instead of regarding the first two stages with his smile of contempt, held on to them as independent or distinct elements of the triune Reality, pragmaticists might have looked up to him as the great vindicator of their',
    'truth" (CP5.436). Peirce maintained that they are all three as ontologically required as are firsts and seconds required for triadic relations in logic. Reaction and feeling are implicit in the continuity of habit, actuality and quality in the generality of law, and interpretation presupposes facts with which to come to terms. Since Peirce, the neo-posi tivists have tried to reduce thirds to seconds, clearly without success. Among the reactions to that pristine scientism have been the current waves of neohegelian fervour that attempt to move the other way, by absorbing feeling and action into interpretation. Peirce\'s ontological realism is based on logical conditions for phenomenology , and thus resists the reductionis ts urges in both directions. These epistemological and ontological realisms are grounded in the dynamical object. Peirce understood dynamis much as Aristotle understood the material condition of change, as serving a conditional limit and as suggesting a teleological fulfillment. Peirce thought it conditional because it was interactive with other circumstances of action and experience. The determination of it as a limit to action and experience depended upon how it related to other conditions in its circumstantial context, and that can be dealt with in thought and action only in an actualized and conceptualized form. None the less, the dynamical object stands over against thought and action, and it requires coming to terms with it to make sense of the world and to act in it. Ultimately, were we to consider it as some sort of surd',
    'logic of relations; although what the intimate connection is between these studies and his metaphysics Peirce leaves the reader to guess at random. 4 This approach has many obvious affiliations with the "presuppositionless \' studies of essences by Descartes, Husserl, and Santayana, although in the light of Peirce \'s devastating critique of the Augustinian-Cartesian philosophy in his earliest papers, too much can not be made of this similarity. And after the nature of Thirdness has been discussed it will be seen how full of presuppositions Peirce \'s analysis really is. 368 JOURNAL OF PHILOSOPHY Let us turn to the categories themselves. The Category of Firstness indicates the aspect of freshness, life, freedom, variety, and spontaneity in the phaneron. Thus it would seem that Firstness presents the phase of atomicity, specificity, or sheer quality in experience. A quality is just what it is, and in its sheer immediacy is utterly disconnected with any other quality. On the other hand Peirce regards Firstness as a "pure nature" or quality to which embodiment and localization are essentially foreign (1.303). A quality is eternal, independent of time or of any realization (1.420); it is simply a "may-be," whose nature is that it might be a specific moment in a phaneron. Nevertheless Firstness is not a general or universal, since to be a universal involves reference to embodiment and localization (1.304). It is not easy to reconcile these various characterizations of Firstness. Peirce seems to shuttle back and forth between regarding Firstness',
]
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	0.972
dot_accuracy	0.028
manhattan_accuracy	0.968
euclidean_accuracy	0.972
max_accuracy	0.972

Triplet

• Dataset: nomic
• Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.971
dot_accuracy	0.029
manhattan_accuracy	0.9695
euclidean_accuracy	0.971
max_accuracy	0.971

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: 10 tokens mean: 255.88 tokens max: 655 tokens	min: 14 tokens mean: 255.08 tokens max: 490 tokens	min: 15 tokens mean: 270.3 tokens max: 562 tokens

• Samples:

  anchor	positive	negative
  finite equational basis. The extension of this theorem to the quasivariety case by J. Czelakowski, W. Dziobiak [5] and the first author [16] followed. Finally, K. A. Baker and J. Wang proved [2] that a variety V with DPSC and Vsi (Vfsi) strictly elementary has finite equational basis. The new proof of Baker's theorem presented in [2] is based on the above mentioned result of K. A. Baker and J. Wang, and the fact that each finitely generated congruence-distributive variety hats DPSC [2, Theorem 2]. In order to obtain a proof of Pigozzi's theorem one can easily carry over the latter fact to the quasivariety case (Proposition 7) and use our Theorem 2. Pigozzi's theorem was already reproved by W. Dziobiak in [6] and by M. Maróti, R. McKenzie in [12]. But, according to our knowledge, the proof presented here is the shortest one, as is the proof of Baker's theorem given in [2]. However, both these proofs are non-constructive. The results more general than Pigozzi's theorem are known nowadays. The first author of this article proved in [17] that a locally finite relatively congruence-distributive quasivariety Q has a finite quasi-equational basis provided the class Qfsi is strictly elementary. W. Dziobiak, M. Maróti, R. McKenzie and the first author proved [7] that each finitely generated relatively congruence-meet-semidistributive quasivariety has a finite quasiequational basis. It is a quasivariety version of Willard's theorem [20]. Finally, K. Palasinska proved in [18] that a finitely generated	protoalgebraic filter-distributive strict universal Horn class is finitely axiomatizable. 2. Toolbox We briefly recall basic facts from quasivariety theory we will need. A standard book about quasivarieties is [8] and about universal algebra are [4, 14]. One may also consult the classical position [11]. Definable Relative Principal Subcongruences 111 We assume that all considered algebras are in the same finite language. A quasivariety Q is a class of algebras defined by a set S of quasi-identities (strict universal Horn sentences). The set E is called a quasiequational basis of Q. Recall that a class is strictly elementary, or finitely axiomatizable, if it may be defined by a single sentence. Notice that a quasivariety Q has a finite quasi-equational basis if and only if Q is strictly elementary. The smallest quasivariety Q containing a given class Ç is SPPu(ö), where P denotes the product class operator, Pu denotes the ultraproduct class operator, and S denotes the subalgebra class operator [8, Corollary 2.3.4]. If G is a finite family of finite algebras, then Q = SP(£), and every such Q is called finitely generated. Let Q be a quasivariety. A congruence a on an algebra A is called a Q-congruence provided A/ a G Q. Note that A G Q if and only if the equality relation 0^ on A is a Q-congruence. The set Cong (A) of all Q-congruences of A forms an algebraic lattice which is a meet-subsemilattice of Con(A) [8, Corollary 1.4.11]. We use the symbol V2 for the lattice join in Cong(A). A nontrivial algebra S	CEP. It follows from this result that V(Qmn), (, n) 0 {(0, 0), (1, 0)}, does not have AP. Using Priestley duality we show in Section 3 that the class of finite quasi-Stone algebras has AP. We do not know whether the whole class QSA has AP. This problem was posed by H. P. Sankappanavar in [8]. In Section 5 we explain briefly how to obtain the free quasi-Stone algebra over a set S from the free Q-lattice over S. Priestley Duality for Quasi-Stone Algebras 85 The concepts and results on distributive lattices and universal algebra that are used here are well known and can be found in the books [1] and [3]. We assume that the reader is familiar with Priestley duality for distributive lattice (see [6] and [7]), the theory of Q-distributive lattices ([4] and [5]) and the theory of quasi-Stone algebras ([8]). 2. Duality The results of this section rely on ?2 of [4]. For a distributive lattice L with 0 and 1 we will denote its Priestley space by X. X can be represented as the set of prime filters of L ordered by set inclusion, the topology on X having as a sub-base the sets of the form {P E X: a 0 P} and {P C X: a E P} for each a E L. The map a 9-+ {P C I: a E P} is a lattice isomorphism of L onto the lattice of clopen increasing subsets of X, where join is set union, meet is
  London 2001 . 8 J. Milbank C. Pickstock G. Ward (eds.), Radical Orthodoxy. A New Theology, Routledge, London New York 1999; J. Milbank, Theology and Social Theory. Beyond Secular Reason, Blackwell Publishing, Maiden 20062, esp. chapt. 12, pp. 382-442; S. Zižek J. Milbank, The Monstrosity of Christ. Paradox or Dialectic?, MIT Press, Cambridge London 2009, pp. 110-233. The most notable figures in this context are: J. Derrida, The Gift of Death and Literature in Secret, trans, by D. Wills, University Of Chicago Press, Chicago 2007; Id., Violence and Metaphysics, in Writing and Difference, trans, by A. Bass, Routledge, London 1978, pp. 79-153; G. Deleuze, Difference and Repetition, trans, by P. Patton, Columbia University Press, New York 1994, pp. 5-11; G. Deleuze F. Guattari, A Thousand Plateaus: Capitalism and Schizophrenia, trans, by B. Massumi, University of Minnesota Press, Minneapolis 1987; M.C. Taylor, Altarity, University of Chicago Press, Chicago 1987; m! Dooley, The Politics of Exodus: S0ren Kierkegaard's Ethics of Responsibility, Fordham University Press, New York 2001. This list is provisional and highly selective. The historical-intellectual contours of poststructuralism are drawn by Johannes Willem Bertens in The Idea of the Postmodern. A History, Routledge, London New York 1995. 10 J. Caputo, The Prayers and Tears of Jacques Derrida: Religion without Religion, Indiana University Press, Bloomington 1997. 1 Id., On Religion, p. 119. 614 LEO STAN attack against established Christianity and its underlying critique of classical metaphysics12. Kierkegaard matters to Caputo because he discloses a universal horizon to religiosity, above and beyond the partial (and always contentious)	instantiations of particular religions. (2) Contrary to Caputo's efforts to universalize Kierkegaard's religious thought, a rather particularistic approach has been advanced by the Christian theologian John Milbank, one of the founders of Radical Orthodoxy. In a groundbreaking essay, titled The Sublime in Kierkegaard 13 , Milbank actually agrees with the poststructuralist reception and tries to describe to what extent Kierkegaard anticipates the postmodern image of the self. One of the reasons Milbank invokes in this direction is that the Kierkegaardian model of subjectivity remains somehow devoid of any universal significance14. For him, Kierkegaard's (religious) selfhood is «endlessly liable to fracture and postponement», «simultaneously preconstituted and deconstituted by a repetitious dynamic which permits only an illusory self-mastery»15. As 12 Moreover, Caputo relies on Kierkegaard's opposition to any rational grounding of religious experience with the subsequent privileging of the processual, finite, incommunicable, passionate, jocose, and thus always perspectivai side of the self's interaction with the divine. See Id., Instants, Secrets , and Singularities : Dealing Death in Kierkegaard and Derrida , in M. J. Matuštik M. Westphal (eds.), Kierkegaard in Posti Modernity, Indiana University Press, Bloomington Indianapolis 1995, pp. 216-238; Id., Radical Hermeneutics , Indiana University Press, Bloomington 1987, pp. 1 1-35, 92, 262, 292, 294; Id., How to Read Kierkegaard , W.W. Norton & Company, New York London 2008, pp. 3, 16, 55. Ironically enough, Caputo wishes to be part of the postmodernist reception of Kierkegaard, which, as already stated, prefers to insist on the nonand even	B ' 'provides the measure against the recurring idolatric tendencies across the spectrum of culture-spheres' (p. 12) a transcendence which is missed by theorists like Habermas. Kierkegaard is often held up in this collection as a corrective to some twentieth-century excesses or distortions. Thus, Patricia Huntington argues that Heidegger's ontologizing of Kierkegaard 'depletes the latter's thought of its ethical import' (p. 44). C. Stephen Evans contrasts Kierkegaard's Christian therapy with Freud's fatalism and pessimism. Matustik traces the way in which Kierkegaard evades modernity's political categories, whilst Westphal claims that Levinas's ethical philosophy lacks Kierkegaard's hope for an ultimate reconciliation of self and Other. This is no simple hagiography of the Danish master, however. True, the essayists are by and large keen to dispel some images of Kierkegaard, and to stress his relevance to social ethics (Perkins, McBride, Evans, Matustik) and REVIEWS 23I as a partner in dialogue for feminist thinking (Berry, Lorraine). But there are notes of criticism too. Huntington points to Kierkegaard's failure to develop a critique of our inherited symbolic order a failure in which he can be supplemented by a reading of Derrida. Alison Leigh Brown writes of Kierkegaard's evasion of feminine imagery for the divine and a need to resist his logic of sacrifice. Jurgen Habermas's contribution criticizes Michael Theunissen for his Kierkegaardian equation of transcendence with a God relationship. Thus, in various ways, a critique of Kierkegaard's conservatism and other-worldliness is upheld. The interest of the collection lies partly in its tensions tensions between those who
  constructed opposition between Spinoza and Hegel. Where Zizek views Hegel's non-dualistic relational epistemology as a substantial improvement over Spinoza's purported dogmatic	account of a reality which is external to the perceiver, I argue that Hegel inherited such an epistemology from Spinoza. Ultimately, it is Spinoza who provides Hegel with	makes sense of things. In Section II, the discussion turns to Hegel's notion of a 'form of consciousness'. I argue that Hegel's notion of a 'form of consciousness' is best understood as comprising a worldview. The principal advantage of articulating stances in a Hegelian way is that such an interpretation explicitly details both the theoretical and affective attitudes

• 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: 23 tokens mean: 257.39 tokens max: 530 tokens	min: 23 tokens mean: 256.21 tokens max: 530 tokens	min: 19 tokens mean: 265.53 tokens max: 814 tokens

• Samples:

  anchor	positive	negative
  cause harm.6 The authors identify the line between reckless and negligent conduct as the line between awareness and unawareness of the risk one is creating.7 In their view, there is no morally significant difference between punishing someone who himself lacked awareness of a risk even if a reasonable person would have been aware ofthat risk and strict liability. They argue that in either case the actor may be punished without proof that she herself consciously disregarded an unjustifiable risk,8 and that there is no principled way of identifying the universe of risks to which one ought to advert.9 And because they assume that negligent behavior lacks the element of choice and therefore falls below the threshold for culpability, they deem it improperly consequentialist to consider whether punishing negligence would promote the penal goal of communicating and enforcing norms for future conduct.10 II. THE NATURE OF THE ARGUMENT AGAINST NEGLIGENCE What sort of argument are the authors making against negligence liability? 'Negligence' might simply be the evaluative label the authors place on the category of conduct that they consider morally non-culpable. The argument might run that criminal culpability requires a guilty mind, a guilty mind requires con6 Id. at p. 79. 7 Id. at p. 69. 8 Id. atDD. 81-82. 9 Id. at η. 76. 10 Id. at p. 78. 436 SUSAN Α. BANDES scious choice, negligence is by definition a lack of conscious choice, and hence negligence is non-culpable. Negligence might simply connote a lack of mens rea in	its most general sense: an absence of blameworthiness.11 Alternatively, it might connote a lack of mens rea in its more specific, elemental, sense: the absence of a conscious or subjective state of mind. In either formulation, the term 'negligence' makes no real contribution. It is essentially a legal conclusion serving the same purpose as the term 'non-culpable,' rather than a term identifying a type of conduct. This is the sort of argument that merges the descriptive task of identifying the mental state with the normative task of determining whether the mental state should be punishable.12 This does not strike me as a fair or plausible reading of the authors' argument. Rather, there is ample support for the conclusion that the authors are making a descriptive argument about the negligent state of mind as a fact in the world.13 The argument is that negligent actors do not make a conscious choice to risk or cause harm. Their actions are beyond their control.14 Moreover, 'character traits are not under actors' direct control, nor can we change our characters at will, at least not at any given moment in time'.15 This argument about the nature of negligence rests on two basic descriptive assumptions. First, it assumes that the negligent actor is unaware of the risk she is taking. Second, it assumes that at the time she undertakes 11 See Joshua Dressier, Understanding Criminal Law (5th ed. 2009), pp. 118-119 (distinguishing broad 'culpability' meaning of mens rea from narrow 'elemental' meaning of the term). 12	intention, focusing instead on corpo rate irresponsibility and negligence. He argues that corporate intention to do harm will not gen erally be identifiable. What could be more easily identifiable is a lack of awareness of risk, a lack of concern, perhaps a degree of irresponsibility. These "dispositions" can be readily absorbed by the culture of an organisation or part of an organisation, and can be extremely difficult for individuals to resist. In such a context the organ isation might be seen as having a moral equiva lent of civil liability, where we would not be looking for intention in order to hold the mis creant liable, but rather a lack of care leading to harmful consequences. Gibson does not argue that the concept of intention is erroneous, simply that the concept of negligence is more produc tive. However, shifting focus to corporate negli gence does not in fact remove the problem of corporate intention. A negligent corporation is one that can be held responsible for its failure to act prudently and with thought to the conse quences of its actions. To be held responsible for its negligent actions or inaction it must be capable of acting otherwise in this context, of acting prudently and far-sightedly. And prudent, far-sighted action involves intention, even if it is only the negative intention of avoiding certain outcomes. So a culpably negligent corporation must be capable of such intentional action. The problem of intention simply returns by another route, despite Gibson's efforts. A more fundamental rethinking of the issues is undertaken by Ouyang and
  clear the air and set the stage. Consider a green spot. Call F the fact of its being green. It is as "simple" a fact as I can think of. Yet, being a fact, F, like any other fact, has constituents which are things. In this sense no fact is "simple." (The familiar difference thus revealed between the uses of 'simple' as applied to facts and to things, respectively, will play a role in the last part.) Which thing or things are the constituents of F? Some answer that they are two, namely, an individual (the spot) and a character (the spot's color). The answer seems obvious; to me it is obvious. But, then, it has not been and still is not obvious to many. Up to a point, of course, everyone agrees that in being presented with F he is presented with the color. Beyond that point, though, some try to make distinctions. They might claim, for instance, that the individual and the character are not presented in quite the same way, only the former but not the latter being "wholly presented." Objections against calling the character a thing, as well as against calling an expression directly referring to it a name, as either unusual or confusing, may be a linguistic symptom of this attempt to distinguish. An extreme and explicit variant of it is to insist that F has only one constituent (which is a thing) and that this constituent is neither the spot nor its color but,	rather, the "colored spot." We have come upon the root of the realismnominalism controversy, a disagreement so fundamental that one might expect it to be relevant to almost any question a philosopher is likely to raise. I shall take it for granted that F has two constituents which are things and which are presented to us 20 INEFFABILITT AND ONTOLOGY whenever F is presented, namely, the individual and the character. But I shall also argue that F has three further constituents which in some sense are presented to us whenever F is. The accent in this paper is on these additional three-this excess, as it were, over either one or two constituents. This is the peculiar way in which, fundamental as it is, the difference between one and two, if I may so express myself, does not matter for my purpose. When I know that this is a green spot, I know also that (i) the spot is an individual, (2) the color is a character, and (3) the former exemplifies the latter (and not, perhaps, the latter the former). How could I know all this if it were not, in some sense, presented to me? To grasp the idea more firmly, consider for a moment a visual field containing two spots, one red, one green. When this field is presented to me, I also know which spot exemplifies which color and, for that matter, also that no spot or color exemplifies the other spot or color. How, to repeat,	cannot be opposed to, for they are part of Reality. Their true logical opposite is not Reality but the Absolute, and to it they are opposed as the finite to the Infinite, the relation being, if viewed from the side of appearances, on,e of total dependence; if viewed from the side of the Absolute, one of absoluteness, independence or irrelativity. (2) Whatever appears to be is.-By saying that a content is real, we mean in the first instance, that it is a logical function of the real system with which we come in contact in perceptive experience. Now this real system (the real world of common-sense) is itself an intellectual construction, and it owes its reality to its being a logical function of the absolute and all-inclusive system. If these positions are granted, it will follow that every appearance is real if every appearance in its integral and proper nature is a function of the absolute system. Since it is in the content of error and illusion, if anywhere, that we must look for unreal appearance, I propose to undertake an analysis which I hope will serve the double object of at once justifying this test of reality and of showing that every appearance satisfies it. Consider such cases as the following: the variations of colour which an object presents in different lights and at different distances; the blueness, for example, of the hills on the horizon, which we refuse to refer to them as their real colour, or the ghostlike appearance of a scarecrow at night to a belated wanderer,
  their end and in their fate,... willingly or not, everything that might be considered the stuff of psychology, kit and caboodle, will follow the path of the signifier. (Lacan 1988, pp. 43-44) 2 For convenience, Lacan's works shall be cited henceforth in the text with date and page number, for example (1948, p. 17); other citations shall be by name, date, and page number (unless the author is indicated in the immediately preceding text, in which case only date and page number will be cited). 3 This is Lacan's word, from the opening quotation above. 4 Human infants are, obviously, gendered, and thus usually referred to with pronouns such as "he or she" and "him or herself." However, since I will constantly be making pronominal reference to infants in my discussion of Lacan's account of the mirror stage, gendered but gender-neutral pronouns would become extremely awkward, making an already-complicated exposition even more muddled. (For example, the next sentence would read as follows: "The infant anticipates a not-yet-experienced coherence and unity of self as he or she recognizes him or herself in the unified body seen in the mirror.") Hence, when making pronominal reference to infants in this paper, I shall use "it" and "itself." This is not meant to dehumanize infants in any way (on the contrary), but rather to facilitate communication of complex ideas. & Springer The Alienating Mirror 21 1 Lacan the event in which a consciousness becomes a self-consciousness. Recognizing an as-yet unrealized unity in the body-image in a mirror and	recognizing this image as its own, the infant becomes aware of itself as a unity, and thus gains selfconsciousness. But Lacan's account diverges markedly here from Hegel's classic account of the acquisition of self-consciousness in the Phenomenology of Spirit. For Hegel, the transition from consciousness to self-consciousness cannot occur, as it does in Lacan's account, without reference to another consciousness.5 If self-consciousness or ego formation is only possible on the presupposition of certain kinds of social interaction, then the mirror stage is not the initial moment of ego-formation, as Lacan suggests, but rather the culmination or conclusion of an extended social process. In order to substantiate this objection, I will first have to present Lacan's account of the mirror stage and ego-formation in some detail. I will then be able to show why the infant's self-recognition in the mirror seems to be prefaced upon some sort of prior mutual-recognition with another subject. In conclusion, I will be able to identify some potential implications of this critique for Lacan's account of the subject and point toward several other promising directions for research. Lacan's account of ego-formation in the infant can be presented very briefly. The two papers in which Lacan himself presents his view are each a mere seven pages in length.6 The key idea, as we have already sketched, is that at a certain stage in an infant's development, when the infant first recognizes its image in a mirror, reacting with jubilation to the recognized image,	wholeness, marks his dynamics with 3 Hypatia Otherness throughout his psychic development. It is this simultaneous desire and fear, simultaneous dependence and disavowal, that shapes him into a split, violent subject warring always against the Otherness of his own embodiment. THE STUFF OF NIGHTMARES . . I want to look more carefully at these boundaries-boundaries between self and Other; and the projection of this internal, psychic Otherness onto boundaries between physical selves and others. I want to look more carefully at these boundaries and projections, at our cultural obsession with them, and at the violences that these produce. It is perhaps too easy in the telling of this story to turn one more time to Rene Descartes (1986) and lay this all at his doorstep-Descartes, the dualistic madman who forever severed our bodies from our minds, and consequently our selves from other selves. That story is perhaps too easy and too worn to illuminate these dynamics. In a more contemporary telling of the Cartesian dualism, Lacan portrays for us in the mirror stage the formation of a subject as an effect of the idealization of a body-in-control. To become a body-incontrol-or at least to pursue this as an ideal, which is necessary to become a legible and meaningful subject in western symbolics-the subject must clearly identify the rigid boundaries between itself and others, including that Other reflected back to it in the image of the mirror. The warring against his Otherness that marks this subject as forever split demands that he separate

• 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	-	-	0.932
0.04	100	0.0092	0.0087	0.934
0.08	200	0.0073	0.0078	0.95
0.12	300	0.0063	0.0069	0.952
0.16	400	0.0073	0.0060	0.956
0.2	500	0.0066	0.0056	0.958
0.24	600	0.0056	0.0060	0.958
0.28	700	0.0056	0.0059	0.954
0.32	800	0.006	0.0050	0.968
0.36	900	0.003	0.0045	0.968
0.4	1000	0.0035	0.0045	0.964
0.44	1100	0.0047	0.0041	0.968
0.48	1200	0.0039	0.0051	0.956
0.52	1300	0.0036	0.0043	0.964
0.56	1400	0.0042	0.0051	0.956
0.6	1500	0.0045	0.0041	0.97
0.64	1600	0.0026	0.0038	0.97
0.68	1700	0.0044	0.0037	0.968
0.72	1800	0.004	0.0043	0.97
0.76	1900	0.0039	0.0038	0.97
0.8	2000	0.0046	0.0043	0.97
0.84	2100	0.0042	0.0040	0.968
0.88	2200	0.0032	0.0040	0.972
0.92	2300	0.0047	0.0040	0.97
0.96	2400	0.004	0.0036	0.974
1.0	2500	0.0026	0.0035	0.978
1.04	2600	0.0024	0.0040	0.968
1.08	2700	0.001	0.0040	0.97
1.12	2800	0.0016	0.0039	0.972
1.16	2900	0.0011	0.0030	0.976
1.2	3000	0.0012	0.0034	0.976
1.24	3100	0.0009	0.0035	0.972
1.28	3200	0.0009	0.0037	0.97
1.32	3300	0.0004	0.0035	0.972
1.3600	3400	0.0003	0.0037	0.97
1.4	3500	0.0006	0.0038	0.97
1.44	3600	0.0008	0.0036	0.968
1.48	3700	0.0002	0.0036	0.968
1.52	3800	0.0002	0.0037	0.968
1.56	3900	0.0002	0.0038	0.968
1.6	4000	0.0003	0.0038	0.968
1.6400	4100	0.0003	0.0038	0.968
1.6800	4200	0.0003	0.0038	0.97
1.72	4300	0.0002	0.0038	0.97
1.76	4400	0.0005	0.0035	0.97
1.8	4500	0.0005	0.0034	0.974
1.8400	4600	0.0005	0.0039	0.964
1.88	4700	0.0002	0.0040	0.964
1.92	4800	0.0003	0.0040	0.964
1.96	4900	0.0002	0.0039	0.968
2.0	5000	0.0003	0.0038	0.968
2.04	5100	0.0001	0.0038	0.97
2.08	5200	0.0001	0.0037	0.97
2.12	5300	0.0001	0.0037	0.968
2.16	5400	0.0001	0.0036	0.972
2.2	5500	0.0002	-	0.971

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}
}