huang2244/adl-hw2-reranker

CrossEncoder based on BAAI/bge-reranker-large

This is a Cross Encoder model finetuned from BAAI/bge-reranker-large using the sentence-transformers library. It computes scores for pairs of texts, which can be used for text reranking and semantic search.

Model Details

Model Description

• Model Type: Cross Encoder
• Base model: BAAI/bge-reranker-large
• Maximum Sequence Length: 384 tokens
• Number of Output Labels: 1 label

Model Sources

• Documentation: Sentence Transformers Documentation
• Documentation: Cross Encoder Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Cross Encoders on Hugging Face

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 CrossEncoder

# Download from the 🤗 Hub
model = CrossEncoder("cross_encoder_model_id")
# Get scores for pairs of texts
pairs = [
    ["user: I've been using those green PMD discs and the salicylic peel combo I bought, but my skin still feels rough. Not sure I trust these strong acids.\nassistant: Got it, those can be pretty intense. Maybe your skin needs a break from harsh exfoliation.\nuser: Yeah, and I tried that 70% glycolic peel too—felt like it burned. I'm skeptical any new peel would be different.\nassistant: Totally understand. Not all peels are that aggressive; some use natural enzymes instead.\nuser: Maybe, but I'm wary of fruit acids too. What would actually be gentle but still help with sun damage?", "IQ Natural 1oz Glycolic Acid Pumpkin Mask (AHA Enzyme) Spa Quality\nCategory: Beauty Description: Glycolic acid is perhaps the best-known of a group of chemicals called fruit acids or alpha-hydroxy acids (AHA). It is derived from sugar cane, so it can be considered a natural product. Citric acid from oranges and other citrus fruits also fall under the same classification as glycolic acid. Glycolic acid works as an exfoliating agent because of its high acidity but easy solubility. Glycolic Acid goes under the damaged upper layers of skin and destroys the 'glue' which holds dead skin to the surface. As this dead skin is chemically burned off, the other ingredients carry the individual flakes away and a water rinse neutralizes the remaining acid. The result is a much-smoother skin surface and a more youthful appearance. A secondary benefit is glycolic acid's ability to draw moisturizers into the newly-exfoliated skin surface."],
    ["user: I liked the Wallace gold candy cane and Reed & Barton bell I bought before, but I'm unsure what holiday ornament format fits this year.\nassistant: Do you prefer a bell shape or another form, and should it be silver or gold plated for the tree or mantel?\nuser: A bell sounds nice. I'd like silver plated and something with the year and holiday plants embossed.", "SWAROVSKI Christmas Ornament, 2021 Annual Edition, Festive, Large, Champagne Crystal\nBrand: SWAROVSKI Category: Home & Kitchen > Home Décor Products > Home Décor Accents > Hanging Ornaments Features: Size: 3 1/8 x 2 1/4 x 3/8 inches The tradition of a Swarovski Annual Edition Ornament continues into its 30th year with this festive crystal creation by Michele Blasilli. The ornament has undergone a remarkable transformation to celebrate its special anniversary, adding contemporary appeal to its timeless design. The star has been expertly cut in an innovative, new asymmetric style to sparkle with 67 gold-tone crystal facets. The red velvet ribbon and brushed champagne gold-tone metal accents are fresh and modern. Exclusively available in 2021, its metal tag – in a round shape – is engraved with the year and embellished with a small crystal. Perfect for hanging on a decorative tree or in a window. Decoration object. Not a toy. Not suitable for children under 15. Description: The tradition of a Swarovski Annual Edition Ornament continues into its 30th year with this festive crystal creation by Michele Blasilli. The ornament has undergone a remarkable transformation to celebrate its special anniversary, adding contemporary appeal to its timeless design. The star has been expertly cut in an innovative, new asymmetric style to sparkle with 67 gold-tone crystal facets. The red velvet ribbon and brushed champagne gold-tone metal accents are fresh and modern. Exclusively available in 2021, its metal tag – in a round shape – is engraved with the year and embellished with a small crystal. Perfect for hanging on a decorative tree or in a window. Decoration object. Not a toy. Not suitable for children under 15. Details: Color: 2021 Annual Edition - Champagne Brand: SWAROVSKI Occasion: Christmas Material: Glass Theme: Religious Style: 2021 Annual Editions Target Audience: womens Included Components: Cord And String Item Dimensions LxWxH: 2.44 x 1.5 x 3.07 inches Shape: Round Mounting Type: Hanging Item Package Quantity: 1 Unit Count: 1 Count Is Customizable: No Batteries Required?: No Item Weight: 1.28 ounces Scent: Unscented Number of Pieces: 1 Product Dimensions: 2.44 x 1.5 x 3.07 inches Department: womens Manufacturer: Swarovski Item model number: 5583847 Best Sellers Rank: {'Home & Kitchen': 139934, 'Decorative Hanging Ornaments': 189} Assembly required: No"],
    ["user: I've been loving that peach lip butter I got from you guys, but now I need something for my dry, damaged hair.\nassistant: Sure, what kind of hair care are you looking for?\nuser: My hair is chemically treated and brittle, so I want a moisturizing shampoo with argan oil.\nassistant: Got it, I can find something that fits that.\nuser: Great, what do you suggest for that?", 'Rejuvenating Shampoo Unisex Shampoo by Macadamia, 33.8 Ounce\nCategory: Beauty Description: Macadamia Natural Oil Rejuvenating Shampoo is a moisturising shampoo for all hair types and is excellent for dry, damaged hair. The Rejuvenating Shampoo gently cleanses the hair while weightlessly replenishing moisture and protecting the hair from environmental elements.'],
    ['user: I want a toy for my toddler that helps with early learning and keeps them busy with hands-on play.\nassistant: Would you prefer something for letter and number recognition or more open-ended imaginative play? Also, do you need a storage box included?\nuser: Letter and number recognition sounds right, and yes, a storage box would be helpful for cleanup.\nassistant: Nice. Do you want a compact set sized for little hands, or a larger playset like the sand and water table you bought before?\nuser: Something compact and easy for little hands is better than a big outdoor setup.', "Infantino Discover and Play Soft Blocks Development Toy\nBrand: Infantino Category: Toys & Games > Baby & Toddler Toys > Early Development & Activity Toys > Sorting & Stacking Toys Features: Easy-to-grasp soft blocks complete with rattles crinkle Clacker ring Encourages development of gross mother skills Country Of Origin: China Description: Four soft activity block with fun on every side. Baby will love discovering silly sounds, interesting textures, bright colors and cute animal pals. Details: Item model number: 206-606Z Target gender: Unisex Number Of Items: 1 Style: Discover and Play (new) Batteries required: No Item Weight: 2.64 pounds Country/Region of origin: China Best Sellers Rank: {'Toys & Games': 14367, 'Toy Stacking Block Sets': 39, 'Sorting & Stacking Toys': 295, 'Pre-Kindergarten Toys': 7365}"],
    ["user: I need individual lash extensions for beginners. I already have a beginner kit and some under eye gel pads from before.\nassistant: Are your clients' natural lashes thicker and you mainly want to lengthen them, or are they sparse and you want a fuller look?\nuser: Their lashes are thick, so I just want to add length with a simple classic style.\nassistant: Do you prefer a soft lightweight lash that keeps its curl for months, or is easy pickup from the strip your main concern?\nuser: Soft and lightweight matters most, and I'd like the curl to last a long time.", "Ardell False Eyelashes Demi Wispies Black, 1 pack (5 pairs per pack)\nBrand: Ardell Category: Beauty & Personal Care > Tools & Accessories > Makeup Brushes & Tools > Eye > False Eyelashes & Adhesives > False Lashes Features: APPLY FAUX LASH EXTENSIONS LIKE A PRO: This multipack from Ardell includes a convenient tool that makes applying false lashes easy and precise. Apply your favorite false lashes in a snap with this innovative tool, the Ardell precision lash applicator that’s perfect for both the novice lash wearer and the aficionado. The rubber-coated grips offer superior hold and control as you place the lash on your lid and the ergonomic design fits comfortably in your hand. STAGGERED LENGTHS FOR A SUBTLE YET FLIRTY LOOK: They are shorter in the inner corners and longer at the outer corners, giving your natural lashes a sexy flared effect that many will find irresistible. Ardell Demi Demi Wispies Black exhibits delicate natural curls that add length and volume on your natural lashes without being obvious and are lightweight and perfect for users who are just starting to apply falsies and don’t want anything complicated. Fit to size, cut accordingly, and glue away! LUXURY FALSIES FOR LESS: The best way to achieve an ultra-glam, the dramatic look is with false eyelashes. Ardell Demi Wispies Black Lashes. These blinkers offer great quality and style both at an extremely affordable price! They will breathe life to your eyes and complement whatever makeup style you’re wearing. A good value for your money that lasts for a long time. To reuse, gently peel off the glue on the band and avoid using mascara. SAFE TO USE, 100% CRUELTY-FREE: Every pair of falsies and the lash adhesive within this pack are free from harmful ingredients that could irritate. All of Ardell’s products are 100% cruelty-free, have not been tested on animals. Each lash strand is made by hand to achieve the highest quality and is free from harmful ingredients so you can use them without having to worry about animals being harmed for your falsies! NOTICEABLE NATURAL LOOKING LASHES: The natural soft look to use for your everyday affairs! Take the bulky out of the equation with Ardell Demi Wispies! These falsies have delicate curves and multiple layers that add length and volume to your natural eyelashes! This design creates lash volume for a lush, eye-opening effect. Get instant intensity with a pretty, natural-looking wispy effect with this iconic style, Demi Falsies! Description: Ardell False Eyelashes Demi Wispies Black, 1 pack (5 pairs per pack) Portable And Ergonomic Featuring these strip lashes from Ardell®, easy to store and use! It comes in a multipack and each tray has a compact design that fits into your bag, which makes it easy to carry and travel-friendly. With this particular faux lash extensions, you will get the highest standard of quality and reliability as it does a terrific job of blending in with your natural eyelashes! High-Quality, Seamless Lash Look These faux eyelashes from Ardell® are incredibly lightweight and realistic looking, but they\\'re also attached"],
]
scores = model.predict(pairs)
print(scores.shape)
# (5,)

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    "user: I've been using those green PMD discs and the salicylic peel combo I bought, but my skin still feels rough. Not sure I trust these strong acids.\nassistant: Got it, those can be pretty intense. Maybe your skin needs a break from harsh exfoliation.\nuser: Yeah, and I tried that 70% glycolic peel too—felt like it burned. I'm skeptical any new peel would be different.\nassistant: Totally understand. Not all peels are that aggressive; some use natural enzymes instead.\nuser: Maybe, but I'm wary of fruit acids too. What would actually be gentle but still help with sun damage?",
    [
        "IQ Natural 1oz Glycolic Acid Pumpkin Mask (AHA Enzyme) Spa Quality\nCategory: Beauty Description: Glycolic acid is perhaps the best-known of a group of chemicals called fruit acids or alpha-hydroxy acids (AHA). It is derived from sugar cane, so it can be considered a natural product. Citric acid from oranges and other citrus fruits also fall under the same classification as glycolic acid. Glycolic acid works as an exfoliating agent because of its high acidity but easy solubility. Glycolic Acid goes under the damaged upper layers of skin and destroys the 'glue' which holds dead skin to the surface. As this dead skin is chemically burned off, the other ingredients carry the individual flakes away and a water rinse neutralizes the remaining acid. The result is a much-smoother skin surface and a more youthful appearance. A secondary benefit is glycolic acid's ability to draw moisturizers into the newly-exfoliated skin surface.",
        "SWAROVSKI Christmas Ornament, 2021 Annual Edition, Festive, Large, Champagne Crystal\nBrand: SWAROVSKI Category: Home & Kitchen > Home Décor Products > Home Décor Accents > Hanging Ornaments Features: Size: 3 1/8 x 2 1/4 x 3/8 inches The tradition of a Swarovski Annual Edition Ornament continues into its 30th year with this festive crystal creation by Michele Blasilli. The ornament has undergone a remarkable transformation to celebrate its special anniversary, adding contemporary appeal to its timeless design. The star has been expertly cut in an innovative, new asymmetric style to sparkle with 67 gold-tone crystal facets. The red velvet ribbon and brushed champagne gold-tone metal accents are fresh and modern. Exclusively available in 2021, its metal tag – in a round shape – is engraved with the year and embellished with a small crystal. Perfect for hanging on a decorative tree or in a window. Decoration object. Not a toy. Not suitable for children under 15. Description: The tradition of a Swarovski Annual Edition Ornament continues into its 30th year with this festive crystal creation by Michele Blasilli. The ornament has undergone a remarkable transformation to celebrate its special anniversary, adding contemporary appeal to its timeless design. The star has been expertly cut in an innovative, new asymmetric style to sparkle with 67 gold-tone crystal facets. The red velvet ribbon and brushed champagne gold-tone metal accents are fresh and modern. Exclusively available in 2021, its metal tag – in a round shape – is engraved with the year and embellished with a small crystal. Perfect for hanging on a decorative tree or in a window. Decoration object. Not a toy. Not suitable for children under 15. Details: Color: 2021 Annual Edition - Champagne Brand: SWAROVSKI Occasion: Christmas Material: Glass Theme: Religious Style: 2021 Annual Editions Target Audience: womens Included Components: Cord And String Item Dimensions LxWxH: 2.44 x 1.5 x 3.07 inches Shape: Round Mounting Type: Hanging Item Package Quantity: 1 Unit Count: 1 Count Is Customizable: No Batteries Required?: No Item Weight: 1.28 ounces Scent: Unscented Number of Pieces: 1 Product Dimensions: 2.44 x 1.5 x 3.07 inches Department: womens Manufacturer: Swarovski Item model number: 5583847 Best Sellers Rank: {'Home & Kitchen': 139934, 'Decorative Hanging Ornaments': 189} Assembly required: No",
        'Rejuvenating Shampoo Unisex Shampoo by Macadamia, 33.8 Ounce\nCategory: Beauty Description: Macadamia Natural Oil Rejuvenating Shampoo is a moisturising shampoo for all hair types and is excellent for dry, damaged hair. The Rejuvenating Shampoo gently cleanses the hair while weightlessly replenishing moisture and protecting the hair from environmental elements.',
        "Infantino Discover and Play Soft Blocks Development Toy\nBrand: Infantino Category: Toys & Games > Baby & Toddler Toys > Early Development & Activity Toys > Sorting & Stacking Toys Features: Easy-to-grasp soft blocks complete with rattles crinkle Clacker ring Encourages development of gross mother skills Country Of Origin: China Description: Four soft activity block with fun on every side. Baby will love discovering silly sounds, interesting textures, bright colors and cute animal pals. Details: Item model number: 206-606Z Target gender: Unisex Number Of Items: 1 Style: Discover and Play (new) Batteries required: No Item Weight: 2.64 pounds Country/Region of origin: China Best Sellers Rank: {'Toys & Games': 14367, 'Toy Stacking Block Sets': 39, 'Sorting & Stacking Toys': 295, 'Pre-Kindergarten Toys': 7365}",
        "Ardell False Eyelashes Demi Wispies Black, 1 pack (5 pairs per pack)\nBrand: Ardell Category: Beauty & Personal Care > Tools & Accessories > Makeup Brushes & Tools > Eye > False Eyelashes & Adhesives > False Lashes Features: APPLY FAUX LASH EXTENSIONS LIKE A PRO: This multipack from Ardell includes a convenient tool that makes applying false lashes easy and precise. Apply your favorite false lashes in a snap with this innovative tool, the Ardell precision lash applicator that’s perfect for both the novice lash wearer and the aficionado. The rubber-coated grips offer superior hold and control as you place the lash on your lid and the ergonomic design fits comfortably in your hand. STAGGERED LENGTHS FOR A SUBTLE YET FLIRTY LOOK: They are shorter in the inner corners and longer at the outer corners, giving your natural lashes a sexy flared effect that many will find irresistible. Ardell Demi Demi Wispies Black exhibits delicate natural curls that add length and volume on your natural lashes without being obvious and are lightweight and perfect for users who are just starting to apply falsies and don’t want anything complicated. Fit to size, cut accordingly, and glue away! LUXURY FALSIES FOR LESS: The best way to achieve an ultra-glam, the dramatic look is with false eyelashes. Ardell Demi Wispies Black Lashes. These blinkers offer great quality and style both at an extremely affordable price! They will breathe life to your eyes and complement whatever makeup style you’re wearing. A good value for your money that lasts for a long time. To reuse, gently peel off the glue on the band and avoid using mascara. SAFE TO USE, 100% CRUELTY-FREE: Every pair of falsies and the lash adhesive within this pack are free from harmful ingredients that could irritate. All of Ardell’s products are 100% cruelty-free, have not been tested on animals. Each lash strand is made by hand to achieve the highest quality and is free from harmful ingredients so you can use them without having to worry about animals being harmed for your falsies! NOTICEABLE NATURAL LOOKING LASHES: The natural soft look to use for your everyday affairs! Take the bulky out of the equation with Ardell Demi Wispies! These falsies have delicate curves and multiple layers that add length and volume to your natural eyelashes! This design creates lash volume for a lush, eye-opening effect. Get instant intensity with a pretty, natural-looking wispy effect with this iconic style, Demi Falsies! Description: Ardell False Eyelashes Demi Wispies Black, 1 pack (5 pairs per pack) Portable And Ergonomic Featuring these strip lashes from Ardell®, easy to store and use! It comes in a multipack and each tray has a compact design that fits into your bag, which makes it easy to carry and travel-friendly. With this particular faux lash extensions, you will get the highest standard of quality and reliability as it does a terrific job of blending in with your natural eyelashes! High-Quality, Seamless Lash Look These faux eyelashes from Ardell® are incredibly lightweight and realistic looking, but they\\'re also attached",
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Training Details

Training Dataset

Unnamed Dataset

• Size: 38,640 training samples
• Columns: query, passage, and label
• Approximate statistics based on the first 1000 samples:

  	query	passage	label
  type	string	string	float
  details	min: 181 characters mean: 441.87 characters max: 872 characters	min: 45 characters mean: 1244.02 characters max: 3624 characters	min: 0.0 mean: 0.11 max: 1.0

• Samples:

  query	passage	label
  user: I've been using those green PMD discs and the salicylic peel combo I bought, but my skin still feels rough. Not sure I trust these strong acids. assistant: Got it, those can be pretty intense. Maybe your skin needs a break from harsh exfoliation. user: Yeah, and I tried that 70% glycolic peel too—felt like it burned. I'm skeptical any new peel would be different. assistant: Totally understand. Not all peels are that aggressive; some use natural enzymes instead. user: Maybe, but I'm wary of fruit acids too. What would actually be gentle but still help with sun damage?	IQ Natural 1oz Glycolic Acid Pumpkin Mask (AHA Enzyme) Spa Quality Category: Beauty Description: Glycolic acid is perhaps the best-known of a group of chemicals called fruit acids or alpha-hydroxy acids (AHA). It is derived from sugar cane, so it can be considered a natural product. Citric acid from oranges and other citrus fruits also fall under the same classification as glycolic acid. Glycolic acid works as an exfoliating agent because of its high acidity but easy solubility. Glycolic Acid goes under the damaged upper layers of skin and destroys the 'glue' which holds dead skin to the surface. As this dead skin is chemically burned off, the other ingredients carry the individual flakes away and a water rinse neutralizes the remaining acid. The result is a much-smoother skin surface and a more youthful appearance. A secondary benefit is glycolic acid's ability to draw moisturizers into the newly-exfoliated skin surface.	0.0
  user: I liked the Wallace gold candy cane and Reed & Barton bell I bought before, but I'm unsure what holiday ornament format fits this year. assistant: Do you prefer a bell shape or another form, and should it be silver or gold plated for the tree or mantel? user: A bell sounds nice. I'd like silver plated and something with the year and holiday plants embossed.	SWAROVSKI Christmas Ornament, 2021 Annual Edition, Festive, Large, Champagne Crystal Brand: SWAROVSKI Category: Home & Kitchen > Home Décor Products > Home Décor Accents > Hanging Ornaments Features: Size: 3 1/8 x 2 1/4 x 3/8 inches The tradition of a Swarovski Annual Edition Ornament continues into its 30th year with this festive crystal creation by Michele Blasilli. The ornament has undergone a remarkable transformation to celebrate its special anniversary, adding contemporary appeal to its timeless design. The star has been expertly cut in an innovative, new asymmetric style to sparkle with 67 gold-tone crystal facets. The red velvet ribbon and brushed champagne gold-tone metal accents are fresh and modern. Exclusively available in 2021, its metal tag – in a round shape – is engraved with the year and embellished with a small crystal. Perfect for hanging on a decorative tree or in a window. Decoration object. Not a toy. Not suitable for children under 15. Description: The tradition ...	0.0
  user: I've been loving that peach lip butter I got from you guys, but now I need something for my dry, damaged hair. assistant: Sure, what kind of hair care are you looking for? user: My hair is chemically treated and brittle, so I want a moisturizing shampoo with argan oil. assistant: Got it, I can find something that fits that. user: Great, what do you suggest for that?	Rejuvenating Shampoo Unisex Shampoo by Macadamia, 33.8 Ounce Category: Beauty Description: Macadamia Natural Oil Rejuvenating Shampoo is a moisturising shampoo for all hair types and is excellent for dry, damaged hair. The Rejuvenating Shampoo gently cleanses the hair while weightlessly replenishing moisture and protecting the hair from environmental elements.	0.0

• Loss: BinaryCrossEntropyLoss with these parameters:

  {
      "activation_fn": "torch.nn.modules.linear.Identity",
      "pos_weight": 7.0
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• gradient_accumulation_steps: 2
• learning_rate: 2e-05
• num_train_epochs: 1.0
• warmup_ratio: 0.1
• bf16: True
• dataloader_pin_memory: False
• gradient_checkpointing: True

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• prediction_loss_only: True
• per_device_train_batch_size: 8
• per_device_eval_batch_size: 8
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 2
• 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: 1.0
• 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: True
• 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: False
• 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: 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
• liger_kernel_config: None
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss
0.0207	50	1.9773
0.0414	100	1.0823
0.0621	150	1.0437
0.0828	200	1.013
0.1035	250	1.0523
0.1242	300	1.2126
0.1449	350	0.9172
0.1656	400	0.9736
0.1863	450	1.0456
0.2070	500	0.9589
0.2277	550	0.8505
0.2484	600	0.9519
0.2692	650	0.8379
0.2899	700	1.0454
0.3106	750	0.906
0.3313	800	0.8882
0.3520	850	0.9142
0.3727	900	0.9688
0.3934	950	0.8844
0.4141	1000	0.8167
0.4348	1050	0.9524
0.4555	1100	0.7877
0.4762	1150	0.8333
0.4969	1200	0.7308
0.5176	1250	0.8182
0.5383	1300	1.0125
0.5590	1350	0.9692
0.5797	1400	0.7449
0.6004	1450	0.8707
0.6211	1500	0.9595
0.6418	1550	0.7955
0.6625	1600	0.8751
0.6832	1650	0.7172
0.7039	1700	0.9637
0.7246	1750	0.7279
0.7453	1800	0.9455
0.7660	1850	0.7206
0.7867	1900	0.9473
0.8075	1950	0.771
0.8282	2000	0.6996
0.8489	2050	0.8529
0.8696	2100	0.8198
0.8903	2150	0.7267
0.9110	2200	0.7785
0.9317	2250	0.7944
0.9524	2300	0.6397
0.9731	2350	0.7383
0.9938	2400	0.9585

Framework Versions

• Python: 3.12.9
• Sentence Transformers: 5.1.0
• Transformers: 4.56.1
• PyTorch: 2.8.0+cu128
• Accelerate: 1.15.0
• Datasets: 5.0.1
• Tokenizers: 0.22.2

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",
}