Instructions to use dinushiTJ/nz-research-commons-embedding-gemma-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use dinushiTJ/nz-research-commons-embedding-gemma-v2 with sentence-transformers:

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("dinushiTJ/nz-research-commons-embedding-gemma-v2")

sentences = [
    "maori_origin",
    "title: Cenozoic sedimentary and volcanic rocks of New Zealand: A reference volume of lithology, age and paleoenvironments with maps (PMAPs) and database.\n\nauthors: Kamp, Peter J.J.\n\nsubjects: paleoenvironment\n\nabstract: This volume presents descriptive geological data and text about each Cenozoic sedimentary and volcanic geological unit to formation and member level (in some cases) exposed on land in New Zealand, including their lithology, stratigraphic age and inferred environment of deposition or emplacement. These data are illustrated as two types of PMAPS: a present-day paleoenvironment map of New Zealand; and as restored paleoenvironment maps, one for each million years from 65 Ma to the present. These information and data underpin the development of a new Cenozoic paleogeographical model of New Zealand.\n\ntext: Cenozoic sedimentary and volcanic rocks of New Zealand: A reference volume of lithology, age and paleoenvironments with maps (PMAPs) and database. This volume presents descriptive geological data and text about each Cenozoic sedimentary and volcanic geological unit to formation and member level (in some cases) exposed on land in New Zealand, including their lithology, stratigraphic age and inferred environment of deposition or emplacement. These data are illustrated as two types of PMAPS: a present-day paleoenvironment map of New Zealand; and as restored paleoenvironment maps, one for each million years from 65 Ma to the present. These information and data underpin the development of a new Cenozoic paleogeographical model of New Zealand.\n\nyear: 2015",
    "title: Ethnic disparities in breast cancer survival in New Zealand: which factors contribute?\n\nauthors: Tin Tin, Sandar\n\nsubjects: Breast cancer\n\nabstract: Background: New Zealand has major ethnic disparities in breast cancer survival with Maori (indigenous people) and Pacific women (immigrants or descended from immigrants from Pacific Islands) faring much worse than other ethnic groups. This paper identified underlying factors and assessed their relative contribution to this risk differential. \r\nMethods: This study involved all women who were diagnosed with primary invasive breast cancer in two health regions, covering about 40% of the national population, between January 2000 and June 2014. Maori and Pacific patients were compared with other ethnic groups in terms of demographics, mode of diagnosis, disease factors and treatment factors. Cox regression modelling was performed with stepwise adjustments, and hazards of excess mortality from breast cancer for Maori and Pacific patients were assessed. \r\nResults: Of the 13,657 patients who were included in this analysis, 1281 (9.4%) were Maori, and 897 (6.6%) were Pacific women. Compared to other ethnic groups, they were younger, more likely to reside in deprived neighbourhoods and to have co-morbidities, and less likely to be diagnosed through screening and with early stage cancer, to be treated in a private care facility, to receive timely cancer treatment, and to receive breast conserving surgery. They had a higher risk of excess mortality from breast cancer (age and year of diagnosis adjusted hazard ratio: 1.76; 95% CI: 1.51-2.04 for Maori and 1.97; 95% CI: 1.67-2.32 for Pacific women), of which 75% and 99% respectively were explained by baseline differences. The most important contributor was late stage at diagnosis. Other contributors included neighbourhood deprivation, mode of diagnosis, type of health care facility where primary cancer treatment was undertaken and type of loco-regional therapy. \r\nConclusions: Late diagnosis, deprivation and differential access to and quality of cancer care services were the key contributors to ethnic disparities in breast cancer survival in New Zealand. Our findings underscore the need for a greater equity focus along the breast cancer care pathway, with an emphasis on improving access to early diagnosis for Maori and Pacific women.\n\ntext: Ethnic disparities in breast cancer survival in New Zealand: which factors contribute? Background: New Zealand has major ethnic disparities in breast cancer survival with Maori (indigenous people) and Pacific women (immigrants or descended from immigrants from Pacific Islands) faring much worse than other ethnic groups. This paper identified underlying factors and assessed their relative contribution to this risk differential. \r\nMethods: This study involved all women who were diagnosed with primary invasive breast cancer in two health regions, covering about 40% of the national population, between January 2000 and June 2014. Maori and Pacific patients were compared with other ethnic groups in terms of demographics, mode of diagnosis, disease factors and treatment factors. Cox regression modelling was performed with stepwise adjustments, and hazards of excess mortality from breast cancer for Maori and Pacific patients were assessed. \r\nResults: Of the 13,657 patients who were included in this analysis, 1281 (9.4%) were Maori, and 897 (6.6%) were Pacific women. Compared to other ethnic groups, they were younger, more likely to reside in deprived neighbourhoods and to have co-morbidities, and less likely to be diagnosed through screening and with early stage cancer, to be treated in a private care facility, to receive timely cancer treatment, and to receive breast conserving surgery. They had a higher risk of excess mortality from breast cancer (age and year of diagnosis adjusted hazard ratio: 1.76; 95% CI: 1.51-2.04 for Maori and 1.97; 95% CI: 1.67-2.32 for Pacific women), of which 75% and 99% respectively were explained by baseline differences. The most important contributor was late stage at diagnosis. Other contributors included neighbourhood deprivation, mode of diagnosis, type of health care facility where primary cancer treatment was undertaken and type of loco-regional therapy. \r\nConclusions: Late diagnosis, deprivation and differential access to and quality of cancer care services were the key contributors to ethnic disparities in breast cancer survival in New Zealand. Our findings underscore the need for a greater equity focus along the breast cancer care pathway, with an emphasis on improving access to early diagnosis for Maori and Pacific women.\n\nyear: 2018",
    "title: Command-line Instrument Control and Measurement Tools\n\nauthors: Ho, Cheng-Lin (Daniel)\n\nsubjects: VISA\n\nabstract: The main purpose of this thesis is to automate laboratory instruments with\r\ncommand lines. The created commands allow users to control instruments and\r\nrecord measured values into a file, by using industrial standards, these commands\r\ncan communicate to the instruments through different interfaces. The users may\r\nutilize these commands to execute an experiment remotely and records values\r\nautomatically. These commands made experiments flexible in location, reduce\r\nthe time spent for recording experiment values, and reduce any possible human\r\nerrors. The recorded values can be accepted by well-known analysis software\r\npackages such as MATLAB, for further processing as a file. These commands\r\nprovide the users with convenience and personal safety, when executing\r\nlaboratory experiments with modern laboratory instruments.\r\nThe target of this thesis is producing a set of commands that allow users to control,\r\nread, and record, the common instruments used on an electronics workbench.\r\nThese devices include power supplies, digital multimeters, function generators,\r\nand oscilloscopes. The produced commands allow users to establish two\r\nworkbenches with these common laboratory instruments. The created\r\ncommands were written in C language in combination with the test and\r\nmeasurement industrial standard to ensure the compatibility with instruments\r\nfrom different vendors. This thesis also provides the readers with required\r\nbackground knowledge that is related to the usage and development of these\r\ncommands. The target readers of this thesis are the graduates of Bachelor of\r\nElectrical and Electronics Engineering or higher.\r\n\r\nWith these command sets, researchers do not need to spend a long time\r\ncontrolling instruments and recording results from instrument manually. Apart\r\nfrom setting up the hardware, they can simply enter the command, to get the result\r\nthey require.\n\ntext: Command-line Instrument Control and Measurement Tools The main purpose of this thesis is to automate laboratory instruments with\r\ncommand lines. The created commands allow users to control instruments and\r\nrecord measured values into a file, by using industrial standards, these commands\r\ncan communicate to the instruments through different interfaces. The users may\r\nutilize these commands to execute an experiment remotely and records values\r\nautomatically. These commands made experiments flexible in location, reduce\r\nthe time spent for recording experiment values, and reduce any possible human\r\nerrors. The recorded values can be accepted by well-known analysis software\r\npackages such as MATLAB, for further processing as a file. These commands\r\nprovide the users with convenience and personal safety, when executing\r\nlaboratory experiments with modern laboratory instruments.\r\nThe target of this thesis is producing a set of commands that allow users to control,\r\nread, and record, the common instruments used on an electronics workbench.\r\nThese devices include power supplies, digital multimeters, function generators,\r\nand oscilloscopes. The produced commands allow users to establish two\r\nworkbenches with these common laboratory instruments. The created\r\ncommands were written in C language in combination with the test and\r\nmeasurement industrial standard to ensure the compatibility with instruments\r\nfrom different vendors. This thesis also provides the readers with required\r\nbackground knowledge that is related to the usage and development of these\r\ncommands. The target readers of this thesis are the graduates of Bachelor of\r\nElectrical and Electronics Engineering or higher.\r\n\r\nWith these command sets, researchers do not need to spend a long time\r\ncontrolling instruments and recording results from instrument manually. Apart\r\nfrom setting up the hardware, they can simply enter the command, to get the result\r\nthey require.\n\nyear: 2008"
]
embeddings = model.encode(sentences)

similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [4, 4]

Notebooks
Google Colab
Kaggle

SentenceTransformer based on google/embeddinggemma-300m

This is a sentence-transformers model finetuned from google/embeddinggemma-300m. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, classification, clustering, and more.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: google/embeddinggemma-300m
Maximum Sequence Length: 2048 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity
Supported Modality: Text

Model Sources

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

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'transformer_task': 'feature-extraction', 'modality_config': {'text': {'method': 'forward', 'method_output_name': 'last_hidden_state'}}, 'module_output_name': 'token_embeddings', 'architecture': 'Gemma3TextModel'})
  (1): Pooling({'embedding_dimension': 768, 'pooling_mode': 'mean', 'include_prompt': True})
  (2): Dense({'in_features': 768, 'out_features': 3072, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity', 'module_input_name': 'sentence_embedding', 'module_output_name': 'sentence_embedding'})
  (3): Dense({'in_features': 3072, 'out_features': 768, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity', 'module_input_name': 'sentence_embedding', 'module_output_name': 'sentence_embedding'})
  (4): 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("dinushiTJ/nz-research-commons-embedding-gemma-v2")
# Run inference
queries = [
    'maori_origin',
]
documents = [
    'title: Understanding cultural relationships: Whānau, whanaungatanga and Māori student attainment of university entrance in a mainstream secondary school in Aotearoa, New Zealand\n\nauthors: McGill, Kristin\n\nsubjects: Cultural relationships\n\nabstract: This study seeks to understand the importance of cultural relationships in supporting Māori student achievement of University Entrance.  This research is based on the stories of five female ākonga Māori, all of whom completed five years of secondary education, and their whānau.  It looks deeply into their relational experiences of whanaungatanga and whānautanga with their school, and the impact this had on their academic achievement of NCEA Level 3 and University Entrance. \r\n\r\nThe results highlight the importance of culturally grounded transformative praxis and the risk of attempting to incorporate culturally located principles such as whānau and whanaungatanga into a schooling context, while still operating within historical hegemonic frameworks.\n\ntext: Understanding cultural relationships: Whānau, whanaungatanga and Māori student attainment of university entrance in a mainstream secondary school in Aotearoa, New Zealand This study seeks to understand the importance of cultural relationships in supporting Māori student achievement of University Entrance.  This research is based on the stories of five female ākonga Māori, all of whom completed five years of secondary education, and their whānau.  It looks deeply into their relational experiences of whanaungatanga and whānautanga with their school, and the impact this had on their academic achievement of NCEA Level 3 and University Entrance. \r\n\r\nThe results highlight the importance of culturally grounded transformative praxis and the risk of attempting to incorporate culturally located principles such as whānau and whanaungatanga into a schooling context, while still operating within historical hegemonic frameworks.\n\nyear: 2023',
    'title: Editorial: Pacific education: research and practice.\n\nauthors: Strachan, Jane\n\nsubjects: educational anthropology\n\nabstract: The article discusses various reports published within the issue including one by Tanya Wendt Samu on the call for teachers to be responsive to the diversities between group of learners as well as within groups of learners and another by Fran Cahill on the discussion of the difficulties that Samoan adolescents have in living within the traditional Samoan culture of home.\n\ntext: Editorial: Pacific education: research and practice. The article discusses various reports published within the issue including one by Tanya Wendt Samu on the call for teachers to be responsive to the diversities between group of learners as well as within groups of learners and another by Fran Cahill on the discussion of the difficulties that Samoan adolescents have in living within the traditional Samoan culture of home.\n\nyear: 2006',
    'title: An evaluation of Te Rau Puawai workforce 100: Perspectives of Te Rau Puawai bursars\n\nauthors: Nikora, Linda Waimarie\n\nsubjects: Maori students\n\nabstract: The Te Rau Puawai programme is an attempt to change the nature of the Maori\r\nmental health workforce. To do this, Maori with aspirations to work, or to continue to\r\nwork in the mental health workforce, are supported, financially and academically, to\r\ncomplete a tertiary qualification relevant to the field.\r\nTo evaluate the Te Rau Puawai programme, the Ministry of Health commissioned the\r\nMaori and Psychology Research Unit of the University of Waikato in July 2001. The\r\noverall aim of the evaluation was to provide the Ministry with a clearer understanding\r\nof the programme including: the perceived critical success factors, the barriers if any\r\nregarding Te Rau Puawai, the impact of the programme, the extent to which the\r\nprogramme may be transferable, gaps in the programme, and suggested\r\nimprovements.\r\nThe evaluation team set out to gather the experiences and perspectives of recipients of\r\nTe Rau Puawai services by asking all bursars to complete a questionnaire and\r\nvolunteer for follow up interviews or focus groups. Sixty two bursars responded to\r\nour questionnaire, and we complete focus group or individual follow up interviews\r\nwith 19 bursars.\n\ntext: An evaluation of Te Rau Puawai workforce 100: Perspectives of Te Rau Puawai bursars The Te Rau Puawai programme is an attempt to change the nature of the Maori\r\nmental health workforce. To do this, Maori with aspirations to work, or to continue to\r\nwork in the mental health workforce, are supported, financially and academically, to\r\ncomplete a tertiary qualification relevant to the field.\r\nTo evaluate the Te Rau Puawai programme, the Ministry of Health commissioned the\r\nMaori and Psychology Research Unit of the University of Waikato in July 2001. The\r\noverall aim of the evaluation was to provide the Ministry with a clearer understanding\r\nof the programme including: the perceived critical success factors, the barriers if any\r\nregarding Te Rau Puawai, the impact of the programme, the extent to which the\r\nprogramme may be transferable, gaps in the programme, and suggested\r\nimprovements.\r\nThe evaluation team set out to gather the experiences and perspectives of recipients of\r\nTe Rau Puawai services by asking all bursars to complete a questionnaire and\r\nvolunteer for follow up interviews or focus groups. Sixty two bursars responded to\r\nour questionnaire, and we complete focus group or individual follow up interviews\r\nwith 19 bursars.\n\nyear: 2002-05-01',
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# [1, 768] [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[ 0.9017, -0.1908,  0.8361]])

Training Details

Training Dataset

Unnamed Dataset

Size: 5,000 training samples
Columns: anchor, positive, and negative

Approximate statistics based on the first 100 samples:

	anchor	positive	negative
type	string	string	string
modality	text	text	text
details	min: 6 tokens mean: 6.75 tokens max: 8 tokens	min: 55 tokens mean: 702.62 tokens max: 1882 tokens	min: 93 tokens mean: 651.33 tokens max: 2048 tokens

Samples:

anchor	positive	negative
`maori_origin`	title: Bite authors: Mitchell, Tori subjects: poetry abstract: Bite is a collection of blunt confessional poetry draped in honesty, identity, and 21st century feminism, examined through lived experience with eating disorders, sexuality, trauma, disconnect from family and culture, love and loss, and what it means to be a girl growing up. The collection explores these raw concepts with vulnerability, through the eyes of ‘wolf girl’, in hopes that talking loudly about these topics will force the public to challenge the silence and stigma around them.

This thesis was inspired by Maori Mythology, and all of the Polynesian poets who showed me how important it is for women of colour to share their stories and experiences. It was also influenced by feminist artists Jenny Holzer, Tracey Emin, and Barbara Kruger, whose art embodies the content of these poems in a literal way: freedom of expression, stripped bare and unflinching. The poetry in this collection is confronting, visceral, and...	title: Remote coastal monitoring of beach usage on Tairua Beach authors: May-Stanley, Bridgette Petra subjects: Remote Sensing abstract: Research that gathers data from public reporting is susceptible to population bias, which arises from a lack of knowledge on the probability of a person’s ability to witness an event. Data collection based on public reporting is often used in coastal research relating to litter, stranded marine animals or bird spotting. It is biased by the probability of a person being in the vicinity, noticing and informing the appropriate organizations about the event. There is potential to improve population biased data by correcting for the probability of a person being present in a particular coastal vicinity. This thesis aims to better understand spatial and temporal beach usage at Tairua Beach in New Zealand. Building on existing literature, this research incorporates modern techniques to detect people on beaches from images taken every hour, during daylight...
`maori_origin`	title: Ka Mahuta, Ngāti Hauā and the importance of translation theory authors: Roa, Raukura subjects: Waiata Māori abstract: In this paper, I provide an English translation of one Māori waiata - Ka Mahuta – a waiata of Ngāti Hauā, along with a discussion of the relevance of translation theory and of culturally appropriate translation processes that fully involve those whose guidance, support, knowledge and understanding are of critical importance. text: Ka Mahuta, Ngāti Hauā and the importance of translation theory In this paper, I provide an English translation of one Māori waiata - Ka Mahuta – a waiata of Ngāti Hauā, along with a discussion of the relevance of translation theory and of culturally appropriate translation processes that fully involve those whose guidance, support, knowledge and understanding are of critical importance. year: 2003-09	title: Using the Internet to Enhance Teaching at The University of Waikato authors: Dewstow, Ross Albert subjects: e-learning abstract: The University of Waikato brought the Internet to New Zealand, was one of the first Universities in New Zealand to graduate students who had completed a bachelor's degree online, and recently won an award for innovative use of video software in an online classroom. The video software was created by a company that had its beginnings within the University. However, the use of the Internet for teaching and learning in the University has reached a plateau in the last few years, as measured by the daily page views of the online platform (Moodie, 2004), the number of courses taught online and staff teaching online remaining fairly constant. This thesis sets out to investigate why the use of online teaching at the University has not increased to a point where a majority of staff are using online teaching to at least supplement their classroom teaching.
Pr...
`maori_origin`	title: Pretty difficult: Implementing kaupapa Māori theory in English-medium secondary schools authors: Bishop, Russell subjects: Te Kotahitanga Project abstract: Developed in New Zealand some twenty years ago, kaupapa Māori has had a successful impact in education, notably in Māori-medium settings such as kōhanga reo, kura kaupapa Māori and wharekura. However, in mainstream educational settings, where the vast majority of Māori children continue to be educated, achievement disparities between Māori and their non-Māori peers persist. This article focuses on Te Kotahitanga, a large-scale kaupapa Māori school reform project that seeks to address educational disparities by improving the educational achievement of Māori students in mainstream schooling. Experiences with implementing Te Kotahitanga would suggest that reforming mainstream educational practices along kaupapa Māori lines is not easy. This article examines three main impediments encountered in attempts to implement the Te Ko...	title: Depositional record of historic lahars in the Whangaehu Gorge, Mt. Ruapehu authors: Graettinger, Alison Hollomon subjects: lahars abstract: Mt. Ruapehu is one of the most lahar prone volcanoes in the world, having both a crater lake and six small glaciers upon its 2797 m summit. The major outlet for the crater lake, the Whangaehu Gorge, has hosted over 46 historic lahars. However, the low preservation of debris flow deposits, as a result of frequent remobilisation on steep slopes, allows for the detailed description of only 9 lahar events over the last 150 years. Field investigation, historic aerial photos, two airborne LiDAR surveys and direct measurements have been utilised to describe the sedimentology, geomorphology and distribution of historic lahar deposits in the first 11 km of the Whangaehu Gorge. Inundation maps have been created for 1945, 1953, 1975, September 1995, October 1995, March 2007 and September 2007. Grain size distribution, componentry and geomorphol...

Loss: TripletLoss with these parameters:

{
    "distance_metric": "TripletDistanceMetric.COSINE",
    "triplet_margin": 0.3
}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 1
learning_rate: 2e-05
num_train_epochs: 1
warmup_ratio: 0.1
prompts: task: classification | query:

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
prediction_loss_only: True
per_device_train_batch_size: 1
per_device_eval_batch_size: 8
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: 2e-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: 1
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}
parallelism_config: None
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch_fused
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: None
hub_always_push: False
hub_revision: None
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_inputs_for_metrics: False
include_for_metrics: []
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
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
use_liger_kernel: False
liger_kernel_config: None
eval_use_gather_object: False
average_tokens_across_devices: False
prompts: task: classification | query:
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss
0.01	50	0.1511
0.02	100	0.0427
0.03	150	0.0149
0.04	200	0.0092
0.05	250	0.0502
0.06	300	0.013
0.07	350	0.0497
0.08	400	0.053
0.09	450	0.0499
0.1	500	0.016
0.11	550	0.0442
0.12	600	0.0422
0.13	650	0.0533
0.14	700	0.014
0.15	750	0.0075
0.16	800	0.0304
0.17	850	0.078
0.18	900	0.0116
0.19	950	0.0474
0.2	1000	0.0095
0.21	1050	0.0254
0.22	1100	0.0049
0.23	1150	0.0332
0.24	1200	0.024
0.25	1250	0.0124
0.26	1300	0.0275
0.27	1350	0.0517
0.28	1400	0.0344
0.29	1450	0.0162
0.3	1500	0.0269
0.31	1550	0.0234
0.32	1600	0.0124
0.33	1650	0.033
0.34	1700	0.007
0.35	1750	0.001
0.36	1800	0.0161
0.37	1850	0.027
0.38	1900	0.0057
0.39	1950	0.0097
0.4	2000	0.0087
0.41	2050	0.012
0.42	2100	0.0028
0.43	2150	0.0196
0.44	2200	0.0116
0.45	2250	0.0415
0.46	2300	0.0288
0.47	2350	0.0022
0.48	2400	0.0032
0.49	2450	0.0532
0.5	2500	0.0108
0.51	2550	0.0152
0.52	2600	0.0089
0.53	2650	0.0158
0.54	2700	0.0018
0.55	2750	0.006
0.56	2800	0.0021
0.57	2850	0.0098
0.58	2900	0.0038
0.59	2950	0.0104
0.6	3000	0.0181
0.61	3050	0.0114
0.62	3100	0.0049
0.63	3150	0.0074
0.64	3200	0.0122
0.65	3250	0.0094
0.66	3300	0.0153
0.67	3350	0.0212
0.68	3400	0.0
0.69	3450	0.0025
0.7	3500	0.0128
0.71	3550	0.0301
0.72	3600	0.018
0.73	3650	0.0339
0.74	3700	0.0059
0.75	3750	0.0018
0.76	3800	0.032
0.77	3850	0.0076
0.78	3900	0.0204
0.79	3950	0.0046
0.8	4000	0.0
0.81	4050	0.0295
0.82	4100	0.0042
0.83	4150	0.0168
0.84	4200	0.0232
0.85	4250	0.0002
0.86	4300	0.0
0.87	4350	0.0039
0.88	4400	0.0
0.89	4450	0.0079
0.9	4500	0.0177
0.91	4550	0.0301
0.92	4600	0.0246
0.93	4650	0.0029
0.94	4700	0.0273
0.95	4750	0.0
0.96	4800	0.0025
0.97	4850	0.0
0.98	4900	0.0018
0.99	4950	0.0324
1.0	5000	0.0105

Training Time

Training: 1.1 hours

Framework Versions

Python: 3.12.13
Sentence Transformers: 5.7.0
Transformers: 4.57.0.dev0
PyTorch: 2.10.0+cu128
Accelerate: 1.13.0
Datasets: 5.0.1
Tokenizers: 0.22.2

Additional Resources

Training and Finetuning Embedding Models with Sentence Transformers: the end-to-end guide for training or finetuning Sentence Transformer models.
Introduction to Matryoshka Embedding Models: variable-size embeddings that can be truncated with minimal quality loss.
Binary and Scalar Embedding Quantization for Significantly Faster & Cheaper Retrieval: post-training compression of embedding vectors.
Multimodal Embedding & Reranker Models with Sentence Transformers: use text, image, audio, and video models through the same API.
Training and Finetuning Multimodal Embedding & Reranker Models with Sentence Transformers: train multimodal embedding models, with a Visual Document Retrieval walkthrough.

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}
}