This is a sentence-transformers model finetuned from NovaSearch/stella_en_1.5B_v5. 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: NovaSearch/stella_en_1.5B_v5
Maximum Sequence Length: 512 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': 512, 'do_lower_case': False}) with Transformer model: PeftModelForFeatureExtraction 
  (1): Pooling({'word_embedding_dimension': 1536, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Dense({'in_features': 1536, 'out_features': 1024, 'bias': True, 'activation_function': 'torch.nn.modules.linear.Identity'})
)

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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
    "Subject: Adding and Subtracting with Decimals\nConstruct: Subtract decimals where the numbers involved have a different number of decimal places\nQuestion: \\( 0.55-0.2= \\)\nCorrectAnswer: \\( 0.35 \\)\nIncorrectAnswer: \\( 0.33 \\)\nIncorrectReason: The correct answer to the problem \\( 0.55 - 0.2 \\) is \\( 0.35 \\). This is correct because when you subtract \\( 0.2 \\) from \\( 0.55 \\), you are essentially performing the operation \\( 0.55 - 0.20 \\). This can be visualized as \\( 55 \\) hundredths minus \\( 20 \\) hundredths, which equals \\( 35 \\) hundredths, or \\( 0.35 \\).\n\nThe incorrect answer \\( 0.33 \\) likely stems from a common misconception or a calculation error. One possible reason for this mistake is a misunderstanding of decimal subtraction or a misinterpretation of the place values. For example, someone might incorrectly think that \\( 0.55 - 0.2 \\) is the same as \\( 0.55 - 0.22 \\), leading to \\( 0.33 \\). Alternatively, the error could be due to a simple arithmetic mistake, such as not properly aligning the decimal points during the subtraction process. It's important to ensure that the decimal points are aligned correctly and to understand the value of each digit in the decimal places.",
    'When subtracting decimals with a different number of decimals, subtracts one digit from more than one column',
    'Does not know that 7 and -7 are different',
]
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: val
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@25	0.7408
cosine_precision@50	0.0168
cosine_precision@100	0.009
cosine_precision@150	0.0063
cosine_precision@200	0.0048
cosine_recall@50	0.8372
cosine_recall@100	0.9002
cosine_recall@150	0.9381
cosine_recall@200	0.9587
cosine_ndcg@25	0.3986
cosine_mrr@25	0.3017
cosine_map@25	0.3017

Training Details

Training Dataset

Unnamed Dataset

Size: 8,562 training samples
Columns: anchor, positive, and negative
Approximate statistics based on the first 1000 samples:
anchor positive negative
type string string string
details
min: 190 tokens
mean: 331.9 tokens
max: 512 tokens
min: 4 tokens
mean: 13.79 tokens
max: 42 tokens
min: 4 tokens
mean: 15.33 tokens
max: 41 tokens

Samples:

anchor	positive	negative
Subject: Function Machines[object Object]Construct: Calculate the square of a number[object Object]Question: [object Object] What is the input of this function machine?[object Object]CorrectAnswer: ( 256 )[object Object]IncorrectAnswer: ( 8 )[object Object]IncorrectReason: The correct answer is ( 256 ) because the function machine involves taking the square root of the input to produce the output. Given that the output is ( 16 ), we need to find a number whose square root is ( 16 ). Mathematically, this means solving the equation ( \sqrt{x} = 16 ). Squaring both sides, we get ( x = 16^2 = 256 ). Therefore, the input must be ( 256 ).[object Object][object Object]The incorrect answer ( 8 ) likely stems from a common misconception. Someone might have seen the output ( 16 ) and thou...	`Mixes up squaring and multiplying by 2 or doubling`	`Confuses written 'teen' numbers with their corresponding single-digit number`
Subject: Function Machines[object Object]Construct: Calculate the square of a number[object Object]Question: [object Object] What is the input of this function machine?[object Object]CorrectAnswer: ( 256 )[object Object]IncorrectAnswer: ( 8 )[object Object]IncorrectReason: The correct answer is ( 256 ) because the function machine involves taking the square root of the input to produce the output. Given that the output is ( 16 ), we need to find a number whose square root is ( 16 ). Mathematically, this means solving the equation ( \sqrt{x} = 16 ). Squaring both sides, we get ( x = 16^2 = 256 ). Therefore, the input must be ( 256 ).[object Object][object Object]The incorrect answer ( 8 ) likely stems from a common misconception. Someone might have seen the output ( 16 ) and thou...	`Mixes up squaring and multiplying by 2 or doubling`	`When multiplying multiples of ten and the answer requires an extra digit, leaves off that extra digit`
Subject: Ratio and Proportion[object Object]Construct: Convert between currencies given an exchange rate[object Object]Question: Convert 350 Thai baht to Australian Dollars.\n1 Australian dollar = 25 Thai baht[object Object]CorrectAnswer: 14[object Object]IncorrectAnswer: 350[object Object]IncorrectReason: The correct answer is 14 Australian dollars. This is because the conversion rate given is 1 Australian dollar (AUD) equals 25 Thai baht (THB). To convert 350 THB to AUD, you divide 350 by 25, which equals 14 AUD. This calculation correctly reflects the exchange rate and the amount of money being converted.[object Object][object Object]The incorrect answer of 350 is likely due to a misunderstanding of the conversion process. Someone might have mistakenly thought that the amount in Thai baht is the same in Australian dollars, not taking into account the exchange rate. This error occurs when the person fails to apply the conversion factor, instead assuming that the currency values are equivalent without adjustment. This misconception can lead to significant errors in financial trans...	`Assumes a 1:1 conversion ratio between different currencies`	`Believes that the larger the divisor, the larger the answer.`

Loss: CachedMultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim" 4}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 16
per_device_eval_batch_size: 4
learning_rate: 0.001
num_train_epochs: 1.0
lr_scheduler_type: cosine
save_only_model: True
bf16: True
load_best_model_at_end: 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: 4
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: 0.001
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: cosine
lr_scheduler_kwargs: {}
warmup_ratio: 0.0
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: True
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: 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}
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: 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
eval_use_gather_object: False
average_tokens_across_devices: False
prompts: None
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs