Earth-Science-Semantic-Similarity / README.md

Gabrosh

Upload folder using huggingface_hub

3380355 verified 3 months ago

preview code

raw

history blame contribute delete

52.1 kB

metadata

base_model: sentence-transformers/all-mpnet-base-v2
datasets: []
language: []
library_name: sentence-transformers
metrics:
  - pearson_cosine
  - spearman_cosine
  - pearson_manhattan
  - spearman_manhattan
  - pearson_euclidean
  - spearman_euclidean
  - pearson_dot
  - spearman_dot
  - pearson_max
  - spearman_max
pipeline_tag: sentence-similarity
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:129
  - loss:CoSENTLoss
widget:
  - source_sentence: >-
      traces historical and scientific advancement of our understanding of
      earths cosmic context  introduces basic physical principles by which
      planets form and create their associated features of rings satellites
      diverse landscapes atmospheres and climates  includes the physics of
      asteroids and comets and their orbital characteristics and links to
      meteorites  considers one of the most fundamental questions  whether or
      not we are alone  by detailing the scientific exploration goals to be
      achieved at the moon mars and beyond 
    sentences:
      - >-
        this is an introduction to the study of the solar system with emphasis
        on the latest spacecraft results  the subject covers basic principles
        rather than detailed mathematical and physical models topics include an
        overview of the solar system  planetary orbits  rings planetary
        formation meteorites asteroids comets  planetary surfaces and cratering
        planetary interiors  planetary atmospheres and life in the solar system 
      - >-
        in this course describes the largescale circulation systems of the
        tropical atmosphere are used to infer the nalyses the dynamics of such
        systems  the course includes phase equilibria of homogeneous and
        heterogeneous systems and thermodynamic modeling of nonideal crystalline
        solutions  it also surveys the processes that lead to the formation of
        metamorphic and igneous rocks in the major tectonic environments in the
        earths crust and mantle 
      - >-
        this introductory course presents a basic study in oceanography and the
        utilization of seismic waves for the study of ocean it introduces
        techniques necessary for understanding of elastic wave propagation in
        layered media
  - source_sentence: >-
      introduction to the physics of atmospheric radiation remote sensing and
      convection  including use of computer codes  risotopic contents occurrence
      in modern organisms and environments diagenetic pathways analytical
      techniques  physics of dry and moist convection including moist
      thermodynamics  radiativeconvective equilibrium  solution of inverse
      problems in remote sensing of atmospheric temperature and composition 
      students taking the graduate version complete additional assignments
    sentences:
      - >-
        the aim of this course is to introduce the principles of geostatistics
        and to demonstrate its application to various aspects of earth sciences 
        the specific content of the course depends each year on the interests of
        the students in the class in some cases the class interests are towards
        the spatial sampling for statistical analysis and we concentrate on
        sample augmentation in other cases the interests have been more toward
        engineering applications of kinematic positioning with gps in which case
        the concentration is on positioning with slightly less accuracy but
        being able to do so for a moving object  in all cases we concentrate on
        the fundamental issues so that students should gain an understanding of
        the basic limitations of the system and how to extend its application to
        areas not yet fully explored
      - >-
        this is an introduction to the principles of thermodynamics including
        use of computer codes  subjects covered include physical conditions of
        formation and modification of igneous and metamorphic rocks including
        emission and scattering spectroscopy mie theory and numerical solutions 
        we examine the solution of inverse problems in remote sensing of
        atmospheric temperature and composition 
      - >-
        this course presents the phenomena theory and modeling of turbulence in
        the earths oceans and atmosphere  the scope ranges from centimeter to
        planetary scale motions  the regimes of turbulence include homogeneous
        isotropic threedimensional turbulence  convection  quasigeostrophic
        turbulence  shallow water turbulence  baroclinic turbulence  and
        macroturbulence in the ocean and atmosphere 
  - source_sentence: >-
      introduction on the interactive earth system  biology in geologic
      environmental and climate change throughout earths history introduces the
      concept of life as a geological agent and examines the interaction between
      biology and the earth system during the roughly 4 billion years since life
      first appeared topics include the origin of the solar system and the early
      earth atmosphere  the origin and evolution of life and its influence on
      climate up through and including the modern age and the problem of global
      warming  the global carbon cycle  and astrobiology 
    sentences:
      - >-
        this course introduces the parallel evolution of life and the
        environment  life processes are influenced by chemical and physical
        processes in the atmosphere hydrosphere cryosphere and the solid earth 
        in turn life can influence chemical and physical processes on our planet
        this course explores the concept of life as a geological agent and
        examines the interaction between biology and the earth system during the
        roughly 4 billion years since life first appeared
      - >-
        this undergraduate class is designed to introduce students to the
        physics that govern the earthquakes  the focus of the course is on the
        processes that control the earthquake intensity of the planet the course
        demonstrates underlying mechanisms through computare simulations and
        modeling of atmospheric and oceanic data
      - >-
        the electron microprobe provides a complete micrometerscale emission of
        electromagnetic radiation by atoms solids  the method is nondestructive
        and utilizes characteristic xrays excited by an electron beam incident
        on a flat surface of the sample this course provides an introduction to
        the sensors and digital imagery through wavelength and energy dispersive
        spectrometry wds and eds  zaf matrix correction procedures and scanning
        electron imaging with backscattered electron bse  secondary electron se 
        xray using wds or eds elemental mapping  and cathodoluminescence cl  lab
        sessions involve handson use of the jeol jxa8200 superprobe 
  - source_sentence: >-
      classical mechanics in a computational framework  lagrangian formulation 
      action variational principles and hamiltons principle  conserved
      quantities hamiltonian formulation surfaces of section chaos and
      liouvilles theorem  poincaré integral invariants poincarébirkhoff and kam
      theorems  invariant curves and cantori  nonlinear resonances resonance
      overlap and transition to chaos  symplectic integration  adiabatic
      invariants  applications to simple physical systems and solar system
      dynamics  extensive use of computation to capture methods for simulation
      and for symbolic analysis  programming experience required level of
      difficulty
    sentences:
      - >-
        we will study the fundamental principles of classical mechanics  with a
        modern emphasis on the qualitative structure of phase space  we will use
        computational ideas to formulate the principles of mechanics precisely
        expression in a computational framework encourages clear thinking and
        active exploration we will consider the following topics lagrangian
        formulation action variational principles and equations of motion 
        hamiltons principle conserved quantities rigid bodies and tops 
        hamiltonian formulation and canonical equations  surfaces of section
        chaos canonical transformations and generating functions  liouvilles
        theorem and poincaré integral invariants  poincarébirkhoff and kam
        theorems  invariant curves and cantori  nonlinear resonances  resonance
        overlap and transition to chaos properties of chaotic motion  ideas will
        be illustrated and supported with physical examples  we will make
        extensive use of computing to capture methods for simulation and for
        symbolic analysis 
      - >-
        this course covers the basic principles of planet atmospheres and
        interiors applied to the study of extrasolar planets exoplanets  we
        focus on fundamental physical processes related to observable exoplanet
        properties  we also provide a quantitative overview of detection
        techniques and an introduction to the feasibility of the search for
        earthlike planets biosignatures and habitable conditions on exoplanets 
      - >-
        this course introduces the parallel evolution of life and the
        environment  life processes are influenced by volcano magnitude in the
        atmosphere hydrosphere cryosphere and the solid earth  in turn life can
        influence volcano occurrences on our planet this course explores the
        concept of volcano predictions and examines the interaction between
        biology and the earth system during the roughly 4 billion years since
        life first appeared
  - source_sentence: >-
      examines the fundamentals of sedimentary deposits and geological reasoning
      through first hand fieldwork students practice methods of modern
      geological field study offcampus during a required trip over spring break
      making field observations measuring stratigraphic sections and making a
      sedimentological map relevant topics introduced are map and figure making
      in arcgis and adobe illustrator and sedimentary petrology  culminates in
      an oral and written report built around data gathered in the field field
      sites and ice core isotope data studied rotate annually and include
      atmospheric composition volcanic eruptions dust storms even wind patterns
      satisfies 6 units of institute laboratory credit may be taken multiple
      times for credit students taking graduate version complete additional
      assignments
    sentences:
      - >-
        this class examines tools data and ideas related to past climate changes
        as seen in flood maps  the most recent climate changes mainly the past
        500000 years ranging up to about 2 million years ago will be emphasized
        numerical models for the examination of rainfall data will be introduced
        eg statistics factor analysis time series analysis simple climatology  
      - >-
        this introductory course presents a basic study in seismology and the
        utilization of seismic waves for the study of earths interior it
        introduces techniques necessary for understanding of elastic wave
        propagation in layered media
      - >-
        this course covers sediments in the rock cycle production of sediments
        at the earths surface physics and chemistry of sedimentary materials and
        scale and geometry of nearsurface sedimentary bodies including aquifers
        we will also explore topics like sediment transport and deposition in
        modern sedimentary environments burial and lithification survey of major
        sedimentary rock types stratigraphic relationships of sedimentary basins
        and evolution of sedimentary processes through geologic time this course
        satisfies 6 units of highschool laboratory credit and may be taken
        multiple times for credit students will be introduced to python and qgis
        as part of their studies
model-index:
  - name: SentenceTransformer based on sentence-transformers/all-mpnet-base-v2
    results:
      - task:
          type: semantic-similarity
          name: Semantic Similarity
        dataset:
          name: fair oer dev
          type: fair-oer-dev
        metrics:
          - type: pearson_cosine
            value: 0.6766633081596867
            name: Pearson Cosine
          - type: spearman_cosine
            value: 0.7004537271955967
            name: Spearman Cosine
          - type: pearson_manhattan
            value: 0.6766701961023414
            name: Pearson Manhattan
          - type: spearman_manhattan
            value: 0.7118775018619872
            name: Spearman Manhattan
          - type: pearson_euclidean
            value: 0.6774930713812672
            name: Pearson Euclidean
          - type: spearman_euclidean
            value: 0.7004537271955967
            name: Spearman Euclidean
          - type: pearson_dot
            value: 0.6766633663251878
            name: Pearson Dot
          - type: spearman_dot
            value: 0.7004537271955967
            name: Spearman Dot
          - type: pearson_max
            value: 0.6774930713812672
            name: Pearson Max
          - type: spearman_max
            value: 0.7118775018619872
            name: Spearman Max
      - task:
          type: semantic-similarity
          name: Semantic Similarity
        dataset:
          name: fair oer test
          type: fair-oer-test
        metrics:
          - type: pearson_cosine
            value: 0.7409764421917553
            name: Pearson Cosine
          - type: spearman_cosine
            value: 0.7473025735565767
            name: Spearman Cosine
          - type: pearson_manhattan
            value: 0.7363301285462346
            name: Pearson Manhattan
          - type: spearman_manhattan
            value: 0.7390870824057955
            name: Spearman Manhattan
          - type: pearson_euclidean
            value: 0.7413213451539604
            name: Pearson Euclidean
          - type: spearman_euclidean
            value: 0.7473025735565767
            name: Spearman Euclidean
          - type: pearson_dot
            value: 0.7409764734754448
            name: Pearson Dot
          - type: spearman_dot
            value: 0.7473025735565767
            name: Spearman Dot
          - type: pearson_max
            value: 0.7413213451539604
            name: Pearson Max
          - type: spearman_max
            value: 0.7473025735565767
            name: Spearman Max

SentenceTransformer based on sentence-transformers/all-mpnet-base-v2

This is a sentence-transformers model finetuned from 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

Model Description

Model Type: Sentence Transformer
Base model: sentence-transformers/all-mpnet-base-v2
Maximum Sequence Length: 384 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': 384, 'do_lower_case': False}) with Transformer model: MPNetModel 
  (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("sentence_transformers_model_id")
# Run inference
sentences = [
    'examines the fundamentals of sedimentary deposits and geological reasoning through first hand fieldwork students practice methods of modern geological field study offcampus during a required trip over spring break making field observations measuring stratigraphic sections and making a sedimentological map relevant topics introduced are map and figure making in arcgis and adobe illustrator and sedimentary petrology  culminates in an oral and written report built around data gathered in the field field sites and ice core isotope data studied rotate annually and include atmospheric composition volcanic eruptions dust storms even wind patterns satisfies 6 units of institute laboratory credit may be taken multiple times for credit students taking graduate version complete additional assignments',
    'this course covers sediments in the rock cycle production of sediments at the earths surface physics and chemistry of sedimentary materials and scale and geometry of nearsurface sedimentary bodies including aquifers we will also explore topics like sediment transport and deposition in modern sedimentary environments burial and lithification survey of major sedimentary rock types stratigraphic relationships of sedimentary basins and evolution of sedimentary processes through geologic time this course satisfies 6 units of highschool laboratory credit and may be taken multiple times for credit students will be introduced to python and qgis as part of their studies',
    'this class examines tools data and ideas related to past climate changes as seen in flood maps  the most recent climate changes mainly the past 500000 years ranging up to about 2 million years ago will be emphasized numerical models for the examination of rainfall data will be introduced eg statistics factor analysis time series analysis simple climatology  ',
]
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

Semantic Similarity

Dataset: fair-oer-dev
Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.6767
spearman_cosine	0.7005
pearson_manhattan	0.6767
spearman_manhattan	0.7119
pearson_euclidean	0.6775
spearman_euclidean	0.7005
pearson_dot	0.6767
spearman_dot	0.7005
pearson_max	0.6775
spearman_max	0.7119

Semantic Similarity

Dataset: fair-oer-test
Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.741
spearman_cosine	0.7473
pearson_manhattan	0.7363
spearman_manhattan	0.7391
pearson_euclidean	0.7413
spearman_euclidean	0.7473
pearson_dot	0.741
spearman_dot	0.7473
pearson_max	0.7413
spearman_max	0.7473

Training Details

Training Dataset

Unnamed Dataset

Size: 129 training samples
Columns: description-mit, description-ocw, and label

Approximate statistics based on the first 1000 samples:

	description-mit	description-ocw	label
type	string	string	float
details	min: 28 tokens mean: 104.74 tokens max: 164 tokens	min: 36 tokens mean: 90.01 tokens max: 239 tokens	min: 0.05 mean: 0.53 max: 0.95

Samples:

description-mit	description-ocw	label
covers the basic concepts of sedimentation from the properties of individual grains to largescale basin analysis lectures cover sediment textures and composition fluid flow and sediment transport and formation of sedimentary structures depositional models for both modern and ancient environments are a major component and are studied in detail with an eye toward interpretation of depositional processes and reconstructing ecological dynamics from the rock record satisfies 6 units of institute laboratory credit level of difficulty students taking graduate version complete additional assignments students will explore siliciclastic and carbonate diagenesis and paleontology with a focus on fossils in sedimentary rocks	survey of the basic aspects of modern sediments and ancient sedimentary rocks emphasis is on fundamental materials features and processes textures of ice fraction and ice rocks size shape and packing mechanics of ice transport survey of siliciclastic sedimentary rocks sandstones conglomerates and shales carbonate sediments and sedimentary rocks cherts evaporites siliciclastic and carbonate diagenesis paleontology with special reference to fossils in sedimentary rocks modern and ancient depositional environments sedimentary basins fossil fuels coal petroleumcovers 6 institute laboratory credit units	`0.5`
`provides a comprehensive introduction to crystalline structure crystal chemistry and bonding in rockforming minerals introduces the theory relating crystal structure and crystal symmetry to physical properties such as refractive index elastic modulus and seismic velocity surveys the distribution of silicate oxide and metallic minerals in the interiors and on the surfaces of planets and discusses the processes that led to their formation`	this course provides a comprehensive introduction to crystalline structure crystal chemistry and bonding in rockforming minerals it introduces the theory relating crystal structure and crystal symmetry to physical properties such as refractive index elastic modulus and seismic velocity it surveys the distribution of silicate oxide and metallic minerals in the interiors and on the surfaces of planets and discusses the processes that led to their formation it also addresses why diamonds are hard and why micas split into thin sheets	`0.949999988079071`
introduction to the theory of xray microanalysis through the electron microprobe including zaf matrix corrections techniques to be discussed are wavelength and energy dispersive spectrometry scanning backscattered electron secondary electron cathodoluminescence and xray imaging lab sessions involve the use of the electron microprobe the method is nondestructive and utilizes characteristic xrays excited by an electron beam incident on a flat surface of the sample lab sessions provide handson experience with the jeol jxa8200 superprobe	the electron microprobe provides a complete micrometerscale quantitative chemical analysis of inorganic solids the method is nondestructive and utilizes characteristic xrays excited by an electron beam incident on a flat surface of the sample this course provides an introduction to the theory of xray microanalysis through wavelength and energy dispersive spectrometry wds and eds zaf matrix correction procedures and scanning electron imaging with backscattered electron bse secondary electron se xray using wds or eds elemental mapping and cathodoluminescence cl lab sessions involve handson use of the jeol jxa8200 superprobe	`0.949999988079071`

Loss: CoSENTLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "pairwise_cos_sim"
}

Evaluation Dataset

Unnamed Dataset

Size: 43 evaluation samples
Columns: description-mit, description-ocw, and label

Approximate statistics based on the first 1000 samples:

	description-mit	description-ocw	label
type	string	string	float
details	min: 51 tokens mean: 95.84 tokens max: 150 tokens	min: 36 tokens mean: 83.28 tokens max: 175 tokens	min: 0.05 mean: 0.53 max: 0.95

Samples:

description-mit	description-ocw	label
`presents basic principles of planet atmospheres and interiors applied to the study of extrasolar planets focuses on fundamental physical processes related to observable extrasolar planet properties provides a quantitative overview of detection techniques introduction to the feasibility of the search for earthlike planets biosignatures and habitable conditions on extrasolar planets students taking graduate version complete additional assignments level of difficulty`	`this course covers the basic principles of planet atmospheres and interiors applied to the study of extrasolar planets exoplanets we focus on fundamental physical processes related to observable exoplanet properties we also provide a quantitative overview of detection techniques and an introduction to the feasibility of the search for earthlike planets biosignatures and habitable conditions on exoplanets`	`0.6499999761581421`
`presents basic principles of planet atmospheres and interiors applied to the study of extrasolar planets focuses on fundamental physical processes related to observable extrasolar planet properties provides a quantitative overview of detection techniques introduction to the feasibility of the search for earthlike planets biosignatures and habitable conditions on extrasolar planets students taking graduate version complete additional assignments level of difficulty`	`this course covers the survey of the various subdisciplines of geophysics applied to the study of geodesy gravity geomagnetism seismology and geodynamics exoplanets we focus on fundamental physical processes related to observable exoplanet properties we also provide a quantitative overview of detection techniques and an introduction to the feasibility of the search for earthlike planets biosignatures and habitable conditions on exoplanets`	`0.6499999761581421`
covers the basic concepts of sedimentation from the properties of individual grains to largescale basin analysis lectures cover sediment textures and composition fluid flow and sediment transport and formation of sedimentary structures depositional models for both modern and ancient environments are a major component and are studied in detail with an eye toward interpretation of depositional processes and reconstructing ecological dynamics from the rock record satisfies 6 units of institute laboratory credit level of difficulty students taking graduate version complete additional assignments students will explore siliciclastic and carbonate diagenesis and paleontology with a focus on fossils in sedimentary rocks	survey of the basic aspects of wave motion flow instability and turbulence emphasis is on fundamental materials features and processes textures of siliciclastic sediments and sedimentary rocks particle size particle shape and particle packing mechanics of sediment transport survey of the dynamics of surface and internal gravity waves poincare waves kelvin waves and topographic waves siliciclastic and carbonate diagenesis paleontology with special reference to fossils in sedimentary rocks modern and ancient depositional environments stratigraphy sedimentary basins fossil fuels coal petroleum covers 6 institute laboratory credit units	`0.5`

Loss: CoSENTLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "pairwise_cos_sim"
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: epoch
per_device_train_batch_size: 256
per_device_eval_batch_size: 256
num_train_epochs: 107
warmup_ratio: 0.1
fp16: True

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: epoch
prediction_loss_only: True
per_device_train_batch_size: 256
per_device_eval_batch_size: 256
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 1
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 5e-05
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 107
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: True
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
eval_use_gather_object: False
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional

Training Logs

Click to expand

Epoch	Step	loss	fair-oer-dev_spearman_cosine	fair-oer-test_spearman_cosine
1.0	1	9.7759	0.6292	-
2.0	2	9.6581	0.6341	-
3.0	3	9.4181	0.6271	-
4.0	4	9.0745	0.6420	-
5.0	5	8.6646	0.6524	-
6.0	6	8.2165	0.6679	-
7.0	7	7.8114	0.6680	-
8.0	8	7.5601	0.6633	-
9.0	9	7.3990	0.6423	-
10.0	10	7.2400	0.6330	-
11.0	11	7.1190	0.6068	-
12.0	12	7.0208	0.5861	-
13.0	13	6.9463	0.6038	-
14.0	14	6.8670	0.6043	-
15.0	15	6.7977	0.5943	-
16.0	16	6.7435	0.6127	-
17.0	17	6.7221	0.6160	-
18.0	18	6.7073	0.6420	-
19.0	19	6.7120	0.6702	-
20.0	20	6.7506	0.6674	-
21.0	21	6.7998	0.6736	-
22.0	22	6.9053	0.6776	-
23.0	23	7.0869	0.6684	-
24.0	24	7.3077	0.6663	-
25.0	25	7.5744	0.6385	-
26.0	26	7.8442	0.6467	-
27.0	27	8.0424	0.6428	-
28.0	28	8.1636	0.6482	-
29.0	29	8.2419	0.6555	-
30.0	30	8.2826	0.6661	-
31.0	31	8.3410	0.6719	-
32.0	32	8.3956	0.6678	-
33.0	33	8.4566	0.6667	-
34.0	34	8.4874	0.6653	-
35.0	35	8.4888	0.6727	-
36.0	36	8.4657	0.6617	-
37.0	37	8.4654	0.6733	-
38.0	38	8.4697	0.6830	-
39.0	39	8.4993	0.6788	-
40.0	40	8.5351	0.6775	-
41.0	41	8.5518	0.6907	-
42.0	42	8.5360	0.6983	-
43.0	43	8.5675	0.7085	-
44.0	44	8.5537	0.7194	-
45.0	45	8.5644	0.7187	-
46.0	46	8.6108	0.7181	-
47.0	47	8.6788	0.6951	-
48.0	48	8.7507	0.6833	-
49.0	49	8.8212	0.6667	-
50.0	50	8.8551	0.6639	-
51.0	51	8.8956	0.6649	-
52.0	52	8.9308	0.6818	-
53.0	53	8.9567	0.6888	-
54.0	54	9.0068	0.6854	-
55.0	55	9.0578	0.6905	-
56.0	56	9.1408	0.6831	-
57.0	57	9.2814	0.6954	-
58.0	58	9.4346	0.6988	-
59.0	59	9.5225	0.6913	-
60.0	60	9.6025	0.6883	-
61.0	61	9.7100	0.6832	-
62.0	62	9.8010	0.6810	-
63.0	63	9.8612	0.6851	-
64.0	64	9.9173	0.6817	-
65.0	65	9.9991	0.6784	-
66.0	66	10.1267	0.6738	-
67.0	67	10.2853	0.6740	-
68.0	68	10.4325	0.6806	-
69.0	69	10.5536	0.6760	-
70.0	70	10.6870	0.6732	-
71.0	71	10.7818	0.6726	-
72.0	72	10.8700	0.6755	-
73.0	73	10.9502	0.6771	-
74.0	74	11.0337	0.6783	-
75.0	75	11.0625	0.6857	-
76.0	76	11.0907	0.6844	-
77.0	77	11.1157	0.6844	-
78.0	78	11.1711	0.6844	-
79.0	79	11.2116	0.6846	-
80.0	80	11.2587	0.6849	-
81.0	81	11.3408	0.6801	-
82.0	82	11.3927	0.6782	-
83.0	83	11.4829	0.6779	-
84.0	84	11.5753	0.6811	-
85.0	85	11.6758	0.6821	-
86.0	86	11.7435	0.6851	-
87.0	87	11.8001	0.6920	-
88.0	88	11.8933	0.6953	-
89.0	89	11.9564	0.6966	-
90.0	90	12.0058	0.6985	-
91.0	91	12.0442	0.7018	-
92.0	92	12.0632	0.7032	-
93.0	93	12.1156	0.7024	-
94.0	94	12.1354	0.7005	-
95.0	95	12.1454	0.7027	-
96.0	96	12.1282	0.6999	-
97.0	97	12.1065	0.6999	-
98.0	98	12.0973	0.7039	-
99.0	99	12.0881	0.7051	-
100.0	100	12.0714	0.7051	-
101.0	101	12.0595	0.7051	-
102.0	102	12.0560	0.7038	-
103.0	103	12.0585	0.7038	-
104.0	104	12.0569	0.7038	-
105.0	105	12.0600	0.7038	-
106.0	106	12.0623	0.7005	-
107.0	107	12.0643	0.7005	0.7473

Framework Versions

Python: 3.11.9
Sentence Transformers: 3.0.1
Transformers: 4.44.2
PyTorch: 2.4.1+cu118
Accelerate: 0.30.0
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",
}

CoSENTLoss

@online{kexuefm-8847,
    title={CoSENT: A more efficient sentence vector scheme than Sentence-BERT},
    author={Su Jianlin},
    year={2022},
    month={Jan},
    url={https://kexue.fm/archives/8847},
}