README.md

metadata

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:82169
  - loss:MultipleNegativesRankingLoss
base_model: BAAI/bge-m3
widget:
  - source_sentence: can beef help reduce emissions
    sentences:
      - >-
        Toward "Better" Meat? Aligning Meat Sourcing Strategies with Corporate
        Climate and Sustainability Goals These studies each shed light on the
        quantitative effects of shifting production or sourcing from a
        conventional system to an alternative system.


        Because Poore and Nemecek’s (2018) database only captured studies
        published between 2000 and June 2016, we performed a literature review
        using similar search terms and study inclusion criteria to capture
        additional studies that were published through 2022. As Poore and
        Nemecek (2018) did, in some instances we performed adjustments to fill
        data gaps or make results more comparable between studies (e.g.,
        estimating land use using data included in a study, making assumptions
        to estimate impacts from the animals’ full life cycle). See Appendix A
        for more details on our approach to adding in more recent studies and
        Appendix B for the full list of “paired studies” included in our
        analysis below, as well as all adjustments made. The Glossary provides
        definitions of the various production systems.


        For each quantitative environmental indicator (e.g., GHG emissions, land
        use) in each “paired study,” we calculated the percent changes that
        occurred when shifting from the conventional system to the alternative
        production system.
      - >-
        Toward "Better" Meat? Aligning Meat Sourcing Strategies with Corporate
        Climate and Sustainability Goals Finally, there are more complex
        nutrient quality indices that could be used as denominators (FAO 2021;
        Katz-Rosene et al. 2023), but, since no consensus exists about which one
        is “best,” we have used the simpler denominator of protein. In sum, use
        of any of these alternative numerators and denominators would not change
        the main findings and recommendations of this report.


        4.  For GHG emissions, we removed land-use-change emissions from the
        estimates in Poore and Nemecek (2018), so as not to double-count with
        the “carbon opportunity costs” of agricultural land use.
      - >-
        Toward "Better" Meat? Aligning Meat Sourcing Strategies with Corporate
        Climate and Sustainability Goals Shift toward lower-emissions foods. As
        noted elsewhere in this report, because beef is an emissions-intensive
        food, shifting purchases and sales toward lower-emissions foods can help
        companies reduce scope 3 emissions.


        There is growing interest in improving grazing management to increase
        the amount of carbon sequestered in pasturelands, a practice often
        called “regenerative grazing.” Some proponents of regenerative grazing
        even suggest that by removing carbon from the atmosphere, soil carbon
        sequestration could fully offset GHG emissions from beef production,
        suggesting potentially “carbon neutral” or “carbon negative” beef. And
        while traditional life cycle assessments assumed that soil carbon stocks
        on agricultural lands were in equilibrium and did not include soil
        carbon stock changes in studies on agriculture’s environmental impacts,
        more recent studies have begun to incorporate soil carbon measurements,
        including several beef studies included in our review (Buratti et al.
        2017; Eldesouky et al. 2018; Stanley et al. 2018).
  - source_sentence: what is the npv for land restoration in latin america
    sentences:
      - >-
        Toward "Better" Meat? Aligning Meat Sourcing Strategies with Corporate
        Climate and Sustainability Goals Overall, this places fish and seafood
        at the lower end of the environmental impact spectrum for animal
        proteins (Gephart et al. 2021) but usually still higher than plant-based
        proteins.


        Similarly to terrestrial animal proteins, life cycle assessments of
        aquaculture (fish farming) have found that there are environmental
        trade-offs with intensification. When finfish and crustacean aquaculture
        systems move along the spectrum from more traditional extensive systems
        to more industrialized intensive systems, land use and water use per
        kilogram of fish declines, but water pollution and energy use per
        kilogram of fish grow (Bohnes et al. 2018; Waite et al. 2014; Hall et
        al. 2011). Effects on GHG emissions can be mixed under intensification
        due to the growth in energy use and land use for feeds balanced by the
        reduction in land use for ponds (Searchinger et al. 2019), and
        translation of land use into “carbon opportunity costs” can help better
        weigh these trade-offs. Aquaculture is also a significant user of wild
        fish as feed; more than 20 percent of total wild-caught fish catch in
        2020 went to “nonfood” uses—mostly for fishmeal and fish oil used in
        aquaculture operations (FAO 2022c).
      - >-
        Toward "Better" Meat? Aligning Meat Sourcing Strategies with Corporate
        Climate and Sustainability Goals ▪ The company first simulates a pure
        “less meat” strategy to reduce scope 3 emissions and carbon opportunity
        costs by a combined 25 percent. To do so, it finds that sourcing 50
        percent less beef, 20 percent less of other meats, and 15 percent less
        dairy—and shifting the purchases toward pulses, soy, and
        vegetables—achieves this 25 percent reduction in climate impacts.


        ▪ The company then explores a plausible scenario of shifting all chicken
        and egg purchases toward higherwelfare products. It uses Table 7 and
        selects points within the impact ranges to assume that free-range
        chicken and eggs could lead to 15 percent higher GHG emissions and 25
        percent higher land use (carbon opportunity costs) than conventional
        chicken. The company estimates that this would increase total climate
        impacts, but only slightly, since chicken and eggs represent a small
        amount of the company’s total climate impact. Under this scenario, total
        climate impacts are reduced versus the base year by “only” 24 percent
        instead of 25 percent. ▪
      - >-
        The Economic Case for Landscape Restoration in Latin America This
        implies an underestimation of benefits given that, in this form, the
        restoration scenario equation does not account for the remaining annual
        difference in net flow values between the degraded hectare that is
        restored and the same hectare left degraded for the years between full
        restoration and the end of the study’s overall assumed 50-year time
        horizon. The NPVs of all target hectares would have to be calculated for
        all 50 years, particularly in the cases of lightly and moderately
        degraded lands which have recovery periods under restoration (delimited
        in this equation by t, which are only 7 and 15 years, respectively).
  - source_sentence: what is meat sourcing strategy
    sentences:
      - >-
        Toward "Better" Meat? Aligning Meat Sourcing Strategies with Corporate
        Climate and Sustainability Goals CHAPTER 1 Introduction and context Meat
        and dairy production are responsible for a large proportion of global
        greenhouse gas (GHG) emissions. According to one widely cited estimate
        by the Food and Agriculture Organization of the United Nations (FAO),
        animal agriculture (including the agricultural production process and
        related land-use change) accounted for 14.5 percent of global GHG
        emissions in 2005, with beef production alone accounting for 6 percent
        of global emissions (Gerber and FAO 2013). Toward “Better” Meat?
        Aligning meat sourcing strategies with corporate climate and
        sustainability goals | 11


        More recent estimates for animal agriculture’s contribution to global
        emissions in 2010–15 are of a similar magnitude, ranging from 11 to 20
        percent (e.g., Poore and Nemecek 2018; Twine 2021; Xu et al. 2021; FAO
        2022a). Animal agriculture also accounted for more than 30 percent of
        global methane emissions in 2017 (CCAC and UNEP 2021).
      - >-
        Toward "Better" Meat? Aligning Meat Sourcing Strategies with Corporate
        Climate and Sustainability Goals Further work is necessary to gather
        publicly available data on other environmental, social, and economic
        attributes of “better meat,” such as for soil health, on-farm
        biodiversity, and agricultural livelihoods, to inform corporate
        decision-making. Similarly, better data are needed on alternative
        systems and practices related to fish and seafood production; these
        “blue foods” are important contributors to global food and nutrition
        security, but data are even scarcer for these food production systems
        than for terrestrial animal agriculture.


        In an ideal world, “better meat” production could lead to improvements
        across all sustainability goals; however, our analysis shows that
        companies with quantitative sustainability goals need to consider both
        co-benefits and trade-offs across all goals when designing their meat
        sourcing strategies. We also show that balancing these goals is
        eminently possible. This analysis also confirms the critical importance
        of shifting diets high in animal-based foods toward plant-based foods
        and alternative proteins to improve both environmental and animal
        welfare outcomes.
      - >-
        Toward "Better" Meat? Aligning Meat Sourcing Strategies with Corporate
        Climate and Sustainability Goals It is true that poultry has a lower
        climate impact per kilogram of protein than beef and lamb, and climate
        strategies may consider a shift in purchasing from beef toward chicken
        to continue to provide the same amount of meat to consumers while
        reducing GHG emissions. However, an important trade-off to consider from
        an animal welfare perspective is the number of animal lives per unit of
        protein produced. While alternative systems thought of as “better” might
        improve the quality of life of the animals to some degree, animal
        welfare experts also recognize the inherent value of all animals, and
        companies might choose to factor the number of animals slaughtered into
        their decision-making as a simple and easily understood indicator of
        animal welfare.


        Figure 5 shows the trade-off between climate and animal welfare
        indicators when shifting between animal-based foods, showing that the
        foods with the highest climate impact per kg of protein also require the
        fewest animals to be killed, and vice versa. For example, to produce a
        kg of protein, more than 100 times as many chickens need to be
        slaughtered compared to cows.
  - source_sentence: cost of restoring landscape
    sentences:
      - >-
        The Economic Case for Landscape Restoration in Latin America This report
        assesses the economic costs and benefits of landscape restoration in
        Latin America and the Caribbean by monetizing a set of benefits that
        could flow from 20 million hectares of restored lands. The introduction
        highlights some of the drivers and impacts of degradation in the Latin
        America and Caribbean region. The section that follows presents an
        overview of the method used to monetize the benefits of landscape
        restoration; detailed descriptions of the methodology and modeling
        approach are available in the annexes. Next, we present the results—the
        estimation of net economic benefits from restoration and the different
        values for biomes and degree of restoration. Finally, we suggest areas
        where future analysis could provide more location-specific financial
        estimates.


        Agriculture and forestry play an important role in the economy and
        social fabric of Latin America and the Caribbean
      - >-
        Getting Ready Include a more comprehensive analysis of the legal
        framework for tenure and existing conditions on-the-ground


        Discuss how tenure conflicts might be addressed as part of the REDD+
        strategy


        Discusses the ability of forest agencies to plan and implement forest
        management activities Considers the role of non-government stakeholders,
        including communities, in forest management Links identified governance
        challenges to proposed REDD+ strategy options and implementation
        framework


        The NPD provides an overview of recent efforts to improve forest
        management in RoC, e.g., through the FLEGT Voluntary Partnership
        Agreement (VPA), certification schemes, and improving the coverage and
        management of protected areas. According to the NPD, the FLEGT process
        identified numerous forest sector challenges that should be addressed as
        part of a REDD+ program, notably lack of forest administration capacity
        and the need to strengthen involvement of local populations in forest
        management decision-making. According to the NPD, over 4 million
        hectares of concessions have been developed since 2001, but the NPD does
        not discuss the role of the private sector in forest management
        activities in detail.
      - >-
        The Economic Case for Landscape Restoration in Latin America
        Nevertheless, because E&M activities will always require more than a
        single year to be fully implemented, the full per hectare cost should
        not be assigned to the first year of restoration alone, but rather to a
        number of initial years along the restoration time horizon. In the case
        of lightly degraded landscapes, the total cost/ha (from Tables 7 and 8)
        has been divided and assigned equally to the first four years (or
        roughly the first half) of the restoration time horizon. In the case of
        moderately degraded lands, the total cost has been subtracted from
        annual benefit flow values in equal annual tranches over the first 8
        years (again, roughly the first half of the restoration time horizon). 
        Finally, total costs for severely degraded lands are subtracted in equal
        annual amounts over the first 25 years of the restoration time horizon.
        Allocating costs over a 25-year time horizon has the effect of
        discounting costs relative to the benefits.
  - source_sentence: what is the wri meat initiative?
    sentences:
      - >-
        Pilot analysis of global ecosystems: Grassland ecosystems Although
        GLASOD was by necessity a somewhat subjective assessment it was
        extremely carefully prepared by leading experts in the field. It remains
        the only global database on the status of human-induced soil
        degradation, and no other data set comes as close to defining the extent
        of desertification at the global scale (UNEP 1997: V).
      - >-
        Toward "Better" Meat? Aligning Meat Sourcing Strategies with Corporate
        Climate and Sustainability Goals Toward “Better” Meat? Aligning meat
        sourcing strategies with corporate climate and sustainability goals


        WOR L D WOR L D R E S O U R C E S R E S O U R C E S I NS T I T U T E I
        NS T I T U T E


        RICHARD WAITE is the Acting Director for Agriculture Initiatives at WRI.


        is a doctoral student with Oxford University’s Environmental Change
        Institute and a former Research Analyst for WRI’s Food and Climate
        Programs.


        CLARA CHO is the Data Analyst for the Coolfood initiative at WRI.
        Contact: clara.cho@wri.org.


        We are pleased to acknowledge our institutional strategic partners that
        provide core funding to WRI: the Netherlands Ministry of Foreign
        Affairs, Royal Danish Ministry of Foreign Affairs, and Swedish
        International Development Cooperation Agency.


        The authors acknowledge the following individuals for their valuable
        guidance and critical reviews:
      - >-
        Getting Ready THE IMPORTANCE OF FOREST GOVERNANCE TO THE REDD+ READINESS
        PROCESS


        Strengthening forest governance will be an essential component of the
        activities implemented by countries seeking to achieve significant and
        lasting emission reductions through REDD+. Poor forest governance is
        often characterized by weak capacity to manage natural resources, lack
        of decision-maker accountability to impacted stakeholders, and lack of
        public access to information about the status and use of forest
        resources. Potential drivers of deforestation and forest
        degradation—such as illegal logging, unplanned forest conversion, and
        conflicts over access to land and resources—are often symptoms of weak
        forest governance. To develop effective national REDD+ strategies,
        governments need to better understand these challenges and develop
        measures to strengthen forest governance in ways that build the trust of
        domestic and international stakeholders.
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - cosine_accuracy@1
  - cosine_accuracy@3
  - cosine_accuracy@5
  - cosine_accuracy@10
  - cosine_precision@1
  - cosine_precision@3
  - cosine_precision@5
  - cosine_precision@10
  - cosine_recall@1
  - cosine_recall@3
  - cosine_recall@5
  - cosine_recall@10
  - cosine_ndcg@10
  - cosine_mrr@10
  - cosine_map@100
model-index:
  - name: SentenceTransformer based on BAAI/bge-m3
    results:
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: ir eval
          type: ir-eval
        metrics:
          - type: cosine_accuracy@1
            value: 0.34030612244897956
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.5389030612244898
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.6211734693877551
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.7122448979591837
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.34030612244897956
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.17963435374149658
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.12423469387755101
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.07122448979591836
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.34030612244897956
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.5389030612244898
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.6211734693877551
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.7122448979591837
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.5191028810993514
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.458020782717851
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.46727356494811056
            name: Cosine Map@100

SentenceTransformer based on BAAI/bge-m3

This is a sentence-transformers model finetuned from BAAI/bge-m3. It maps sentences & paragraphs to a 1024-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: BAAI/bge-m3
• Maximum Sequence Length: 8192 tokens
• Output Dimensionality: 1024 dimensions
• 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: XLMRobertaModel 
  (1): Pooling({'word_embedding_dimension': 1024, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, '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("collaborativeearth/bge-m3_wri")
# Run inference
sentences = [
    'what is the wri meat initiative?',
    'Toward "Better" Meat? Aligning Meat Sourcing Strategies with Corporate Climate and Sustainability Goals Toward “Better” Meat? Aligning meat sourcing strategies with corporate climate and sustainability goals\n\nWOR L D WOR L D R E S O U R C E S R E S O U R C E S I NS T I T U T E I NS T I T U T E\n\nRICHARD WAITE is the Acting Director for Agriculture Initiatives at WRI.\n\nis a doctoral student with Oxford University’s Environmental Change Institute and a former Research Analyst for WRI’s Food and Climate Programs.\n\nCLARA CHO is the Data Analyst for the Coolfood initiative at WRI. Contact: clara.cho@wri.org.\n\nWe are pleased to acknowledge our institutional strategic partners that provide core funding to WRI: the Netherlands Ministry of Foreign Affairs, Royal Danish Ministry of Foreign Affairs, and Swedish International Development Cooperation Agency.\n\nThe authors acknowledge the following individuals for their valuable guidance and critical reviews:',
    'Pilot analysis of global ecosystems: Grassland ecosystems Although GLASOD was by necessity a somewhat subjective assessment it was extremely carefully prepared by leading experts in the field. It remains the only global database on the status of human-induced soil degradation, and no other data set comes as close to defining the extent of desertification at the global scale (UNEP 1997: V).',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

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

Evaluation

Metrics

Information Retrieval

• Dataset: ir-eval
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.3403
cosine_accuracy@3	0.5389
cosine_accuracy@5	0.6212
cosine_accuracy@10	0.7122
cosine_precision@1	0.3403
cosine_precision@3	0.1796
cosine_precision@5	0.1242
cosine_precision@10	0.0712
cosine_recall@1	0.3403
cosine_recall@3	0.5389
cosine_recall@5	0.6212
cosine_recall@10	0.7122
cosine_ndcg@10	0.5191
cosine_mrr@10	0.458
cosine_map@100	0.4673

Training Details

Training Dataset

Unnamed Dataset

• Size: 82,169 training samples
• Columns: question and answer
• Approximate statistics based on the first 1000 samples:

  	question	answer
  type	string	string
  details	min: 4 tokens mean: 10.62 tokens max: 31 tokens	min: 53 tokens mean: 232.17 tokens max: 337 tokens

• Samples:

  question	answer
  what is the economic case of restoration	The Economic Case for Landscape Restoration in Latin America The Economic Case for Landscape Restoration in Latin America THE ECONOMIC CASE FOR LANDSCAPE RESTORATION IN LATIN AMERICA WALTER VERGARA, LUCIANA GALLARDO LOMELI, ANA R. RIOS, PAUL ISBELL, STEVEN PRAGER, RONNIE DE CAMINO Land use and land-use change are central to the economic and social fabric of Latin America and the Caribbean, and essential to the region’s prospects for sustainable development. Countries are realizing that now, more than ever, is the time for action. Eleven countries, three Brazilian states and several regional programs have already committed to restoring more than 27 million hectares of degraded land in Latin America—but can these ambitions become a reality while supporting good living standards and economic development?
  economic case of landscape restoration in latin america	The Economic Case for Landscape Restoration in Latin America The Economic Case for Landscape Restoration in Latin America THE ECONOMIC CASE FOR LANDSCAPE RESTORATION IN LATIN AMERICA WALTER VERGARA, LUCIANA GALLARDO LOMELI, ANA R. RIOS, PAUL ISBELL, STEVEN PRAGER, RONNIE DE CAMINO Land use and land-use change are central to the economic and social fabric of Latin America and the Caribbean, and essential to the region’s prospects for sustainable development. Countries are realizing that now, more than ever, is the time for action. Eleven countries, three Brazilian states and several regional programs have already committed to restoring more than 27 million hectares of degraded land in Latin America—but can these ambitions become a reality while supporting good living standards and economic development?
  what is lata-american landscape	The Economic Case for Landscape Restoration in Latin America Agriculture and forestry exports from Latin America represent about 13 percent of the global trade of food, feed, and fiber and account for a majority of employment outside large urban areas—numbers only expected to grow as Latin America is called upon to meet an increasing global demand for food. Yet, since the turn of the century, about 37 million hectares of natural forests, savannas and wetlands have been transformed to expand agriculture. Cumulative, unsustainable land-use practices have led to the degradation of about 300 million hectares, resulting in a reduction in yields and quality of production, and in losses in biomass content, soil quality, surface water hydrology, and biodiversity. Deforestation, land-use change, and unsustainable agricultural activities are also currently the largest drivers of climate change in the region, accounting for 56 percent of all greenhouse gas emissions. Today, while some progress ha...

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 32
• learning_rate: 1e-06
• num_train_epochs: 2
• warmup_ratio: 0.1
• fp16: True
• gradient_checkpointing: True
• 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: 32
• 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: 1e-06
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 2
• 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}
• tp_size: 0
• 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: None
• hub_always_push: False
• gradient_checkpointing: True
• 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
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	ir-eval_cosine_ndcg@10
-1	-1	-	0.4718
0.0389	100	0.7439	-
0.0779	200	0.6208	-
0.1168	300	0.4568	-
0.1558	400	0.3713	-
0.1947	500	0.3263	0.5004
0.2336	600	0.2722	-
0.2726	700	0.2521	-
0.3115	800	0.2541	-
0.3505	900	0.2348	-
0.3894	1000	0.2321	0.5090
0.4283	1100	0.2313	-
0.4673	1200	0.2195	-
0.5062	1300	0.2286	-
0.5452	1400	0.2188	-
0.5841	1500	0.2166	0.5115
0.6231	1600	0.2194	-
0.6620	1700	0.2006	-
0.7009	1800	0.1954	-
0.7399	1900	0.2157	-
0.7788	2000	0.2059	0.5154
0.8178	2100	0.203	-
0.8567	2200	0.1949	-
0.8956	2300	0.1943	-
0.9346	2400	0.206	-
0.9735	2500	0.2015	0.5175
1.0125	2600	0.1801	-
1.0514	2700	0.1867	-
1.0903	2800	0.1914	-
1.1293	2900	0.1827	-
1.1682	3000	0.1899	0.5165
1.2072	3100	0.1707	-
1.2461	3200	0.1872	-
1.2850	3300	0.1943	-
1.3240	3400	0.1854	-
1.3629	3500	0.1747	0.5182
1.4019	3600	0.1764	-
1.4408	3700	0.1866	-
1.4798	3800	0.1855	-
1.5187	3900	0.1782	-
1.5576	4000	0.1744	0.5181
1.5966	4100	0.1793	-
1.6355	4200	0.187	-
1.6745	4300	0.1907	-
1.7134	4400	0.1781	-
1.7523	4500	0.1825	0.5185
1.7913	4600	0.1981	-
1.8302	4700	0.1751	-
1.8692	4800	0.1824	-
1.9081	4900	0.1866	-
1.9470	5000	0.188	0.5191
1.9860	5100	0.1838	-

Framework Versions

• Python: 3.11.12
• Sentence Transformers: 4.1.0
• Transformers: 4.51.3
• PyTorch: 2.6.0+cu124
• Accelerate: 1.6.0
• Datasets: 2.14.4
• Tokenizers: 0.21.1

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",
}

MultipleNegativesRankingLoss

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply},
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}