bge-large-eedi-2024 / README.md

Gurveer05

MAP@25: 0.3104430178362365

caf33fd verified about 1 year ago

preview code

raw

history blame contribute delete

22.1 kB

metadata

base_model: BAAI/bge-large-en-v1.5
library_name: sentence-transformers
pipeline_tag: sentence-similarity
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:2442
  - loss:MultipleNegativesRankingLoss
widget:
  - source_sentence: |-
      Construct:  Identify the order of rotational symmetry of a shape.

      Question:  Which shape has the lowest order of rotational symmetry?

      Options:
      A. Equilateral triangle
      B. Circle
      C. Square
      D. Trapezium

      Correct Answer: Trapezium

      Incorrect Answer: Circle
    sentences:
      - >-
        When multiplying decimals, divides by the wrong power of 10 when
        reinserting the decimal
      - Does not understand inequality notation
      - Does not know how to find order of rotational symmetry
  - source_sentence: >-
      Construct:  Solve three or more step linear equations, with the variable
      on one side, involving negative integers.


      Question:  Tom and Katie are discussing how to solve:

      (5 x / 3)-1=-2


      Tom says a correct next line of working could be:  5 x-3=-6 


      Katie says a correct next line of working could be:  (5 x / 3)=-3 


      Who is correct?


      Options:

      A. Only Tom

      B. Only Katie

      C. Both Tom and Katie

      D. Neither is correct


      Correct Answer: Only Tom


      Incorrect Answer: Neither is correct
    sentences:
      - Mixes up the value of two terms when substituting
      - Thinks there are 100 ml in a litre
      - >-
        Does not understand that when multiplying both sides of an equation by
        an amount every term must be multiplied by the same amount
  - source_sentence: >-
      Construct:  Identify corresponding angles.


      Question:  M  and  N  are the intersections of the line  X Y  with the
      lines  P Q  and  R S .

      Which angle is corresponding to angle QMY? A pair of parallel lines
      pointing up to the left. PQ and RS are the ends of the parallel lines. PQ
      is on the left of the diagram with Q being the top left.

      A red straight line, XY, crosses the parallel lines. X is on the left of
      the diagram.

      Line XY crosses line PQ at a point marked M.

      Line XY crosses line RS at a point marked N.

      The angle QMY is marked in red.


      Options:

      A. XMP

      B. SNY

      C. SNX

      D. XNR


      Correct Answer: SNY


      Incorrect Answer: XNR
    sentences:
      - Confuses corresponding and alternate angles
      - Estimates shares of a ratio instead of calculating
      - Misremembers the quadratic formula
  - source_sentence: |-
      Construct:  Factorise a quadratic expression in the form x² - c.

      Question:  Factorise this expression, if possible:
      (
      p^2-4
      ).

      Options:
      A. (p-2)(p+2)
      B. p(p-2)
      C. (p-2)(p-2)
      D. Does not
      factorise

      Correct Answer: (p-2)(p+2)

      Incorrect Answer: p(p-2)
    sentences:
      - Mixes up greater than and less than symbols
      - Does not know how to find the length of a line segment from coordinates
      - Does not recognise difference of two squares
  - source_sentence: >-
      Construct:  Solve quadratic equations using the quadratic formula where
      the coefficient of x² is not 1.


      Question:  Vera wants to solve this equation using the quadratic formula.

      (

      3 h^2-10 h+4=0

      )


      What should replace the circle?  (? pm square root of (?-?) / bigcirc).


      Options:

      A. 3

      B. 5

      C. 9

      D. 6


      Correct Answer: 6


      Incorrect Answer: 3
    sentences:
      - Misremembers the quadratic formula
      - When asked for a specific term in a sequence gives the term after
      - Does not know that vertically opposite angles are equal

SentenceTransformer based on BAAI/bge-large-en-v1.5

This is a sentence-transformers model finetuned from BAAI/bge-large-en-v1.5 on the csv dataset. 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-large-en-v1.5
Maximum Sequence Length: 512 tokens
Output Dimensionality: 1024 tokens
Similarity Function: Cosine Similarity
Training Dataset:
- csv

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': 512, 'do_lower_case': True}) with Transformer model: BertModel 
  (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("Gurveer05/bge-large-eedi-2024")
# Run inference
sentences = [
    'Construct:  Solve quadratic equations using the quadratic formula where the coefficient of x² is not 1.\n\nQuestion:  Vera wants to solve this equation using the quadratic formula.\n(\n3 h^2-10 h+4=0\n)\n\nWhat should replace the circle?  (? pm square root of (?-?) / bigcirc).\n\nOptions:\nA. 3\nB. 5\nC. 9\nD. 6\n\nCorrect Answer: 6\n\nIncorrect Answer: 3',
    'Misremembers the quadratic formula',
    'Does not know that vertically opposite angles are equal',
]
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]

Training Details

Training Dataset

csv

Dataset: csv
Size: 2,442 training samples
Columns: qa_pair_text and MisconceptionName
Approximate statistics based on the first 1000 samples:
qa_pair_text MisconceptionName
type string string
details
min: 40 tokens
mean: 102.66 tokens
max: 512 tokens

min: 4 tokens
mean: 15.26 tokens
max: 39 tokens

	qa_pair_text	MisconceptionName
type	string	string
details	min: 40 tokens mean: 102.66 tokens max: 512 tokens	min: 4 tokens mean: 15.26 tokens max: 39 tokens

Samples:

qa_pair_text	MisconceptionName
`Construct: Convert between cm³ and mm³. Question: 1 cm^3 is the same as _______ mm^3. Options: A. 10 B. 100 C. 1000 D. 10000 Correct Answer: 1000 Incorrect Answer: 10`	`Does not cube the conversion factor when converting cubed units`
`Construct: Write algebraic expressions with correct algebraic convention. Question: Which answer shows the following calculation using the correct algebraic convention? ( y x x+b x 3 ). Options: A. y x+b 3 B. x y+3 b C. y+3 b x D. 3 b x y Correct Answer: x y+3 b Incorrect Answer: 3 b x y`	`Multiplies all terms together when simplifying an expression`
`Construct: Write algebraic expressions with correct algebraic convention. Question: Which of the following is the correct way of writing: p divided by q , then add 3 using algebraic convention? Options: A. p q+3 B. (p / q)+3 C. (p / q+3) D. p-q+3 Correct Answer: (p / q)+3 Incorrect Answer: p-q+3`	`Has used a subtraction sign to represent division`

Loss: MultipleNegativesRankingLoss with these parameters:

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

Evaluation Dataset

csv

Dataset: csv
Size: 1,928 evaluation samples
Columns: qa_pair_text and MisconceptionName
Approximate statistics based on the first 1000 samples:
qa_pair_text MisconceptionName
type string string
details
min: 40 tokens
mean: 103.34 tokens
max: 512 tokens

min: 4 tokens
mean: 14.34 tokens
max: 40 tokens

	qa_pair_text	MisconceptionName
type	string	string
details	min: 40 tokens mean: 103.34 tokens max: 512 tokens	min: 4 tokens mean: 14.34 tokens max: 40 tokens

Samples:

qa_pair_text	MisconceptionName
`Construct: Multiply two decimals together with the same number of decimal places. Question: 0.4^2=. Options: A. 0.08 B. 0.8 C. 1.6 D. 0.16 Correct Answer: 0.16 Incorrect Answer: 0.8`	`Mixes up squaring and multiplying by 2 or doubling`
`Construct: Calculate the cube root of a number. Question: 3rd root of (8)=. Options: A. 2 . dot{6} B. 4 C. 64 D. 2 Correct Answer: 2 Incorrect Answer: 4`	`Halves when asked to find the cube root`
Construct: Calculate missing lengths of shapes by geometrical inference, where the lengths given are in the same units. Question: What is the area of the shaded section of this composite shape made from rectangles? A composite shape made from two rectangles that form an "L" shape. The base of the shape is horizontal and is 13cm long. The vertical height of the whole shape is 14cm. The horizontal width of the top part of the shape is 6cm. The vertical height of the top rectangle is 8cm. The right handed rectangle is shaded blue. Options: A. 48 cm^2 B. 104 cm^2 C. 42 cm^2 D. 56 cm^2 Correct Answer: 42 cm^2 Incorrect Answer: 48 cm^2	`Uses an incorrect side length when splitting a composite shape into parts`

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: 16
per_device_eval_batch_size: 16
gradient_accumulation_steps: 32
weight_decay: 0.01
num_train_epochs: 20
lr_scheduler_type: cosine_with_restarts
warmup_ratio: 0.1
fp16: True
load_best_model_at_end: 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: 16
per_device_eval_batch_size: 16
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 32
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 5e-05
weight_decay: 0.01
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 20
max_steps: -1
lr_scheduler_type: cosine_with_restarts
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: True
ignore_data_skip: False
fsdp: []
fsdp_min_num_params: 0
fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
fsdp_transformer_layer_cls_to_wrap: None
accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch
optim_args: None
adafactor: False
group_by_length: False
length_column_name: length
ddp_find_unused_parameters: None
ddp_bucket_cap_mb: None
ddp_broadcast_buffers: False
dataloader_pin_memory: True
dataloader_persistent_workers: False
skip_memory_metrics: True
use_legacy_prediction_loop: False
push_to_hub: False
resume_from_checkpoint: None
hub_model_id: None
hub_strategy: every_save
hub_private_repo: False
hub_always_push: False
gradient_checkpointing: True
gradient_checkpointing_kwargs: None
include_inputs_for_metrics: False
eval_do_concat_batches: True
fp16_backend: auto
push_to_hub_model_id: None
push_to_hub_organization: None
mp_parameters:
auto_find_batch_size: False
full_determinism: False
torchdynamo: None
ray_scope: last
ddp_timeout: 1800
torch_compile: False
torch_compile_backend: None
torch_compile_mode: None
dispatch_batches: None
split_batches: None
include_tokens_per_second: False
include_num_input_tokens_seen: False
neftune_noise_alpha: None
optim_target_modules: None
batch_eval_metrics: False
eval_on_start: False
eval_use_gather_object: False
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss
0.4183	2	1.2854	-
0.6275	3	-	1.0368
0.8366	4	1.0855	-
1.2549	6	0.7559	0.8548
1.6732	8	0.7032	-
1.8824	9	-	0.6840
2.0915	10	0.474	-
2.5098	12	0.3959	0.6023
2.9281	14	0.3279	-
3.1373	15	-	0.5576
3.3464	16	0.2164	-
3.7647	18	0.1991	0.4972
4.1830	20	0.1378	-
4.3922	21	-	0.5081
4.6013	22	0.1168	-
5.0196	24	0.0955	0.5000

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.1.1
Transformers: 4.44.2
PyTorch: 2.4.1+cu121
Accelerate: 0.34.2
Datasets: 2.19.2
Tokenizers: 0.19.1

Citation

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}

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}
}