DeBERTa3-s-CustomPoolin-toytest3-step1

SentenceTransformer based on microsoft/deberta-v3-small

This is a sentence-transformers model finetuned from microsoft/deberta-v3-small. It maps sentences & paragraphs to a 768-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: microsoft/deberta-v3-small <!-- at revision a36c739020e01763fe789b4b85e2df55d6180012 -->
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 tokens
• Similarity Function: Cosine Similarity

<!-- - **Training Dataset:** Unknown --> <!-- - **Language:** Unknown --> <!-- - **License:** Unknown -->

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: DebertaV2Model 
  (1): AdvancedWeightedPooling(
    (alpha_dropout_layer): Dropout(p=0.01, inplace=False)
    (gate_dropout_layer): Dropout(p=0.05, inplace=False)
    (linear_cls_pj): Linear(in_features=768, out_features=768, bias=True)
    (linear_cls_Qpj): Linear(in_features=768, out_features=768, bias=True)
    (linear_mean_pj): Linear(in_features=768, out_features=768, bias=True)
    (linear_attnOut): Linear(in_features=768, out_features=768, bias=True)
    (mha): MultiheadAttention(
      (out_proj): NonDynamicallyQuantizableLinear(in_features=768, out_features=768, bias=True)
    )
    (layernorm_output): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
    (layernorm_weightedPooing): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
    (layernorm_pjCls): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
    (layernorm_pjMean): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
    (layernorm_attnOut): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
  )
)

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("bobox/DeBERTa3-s-CustomPoolin-toytest3-step1")
# Run inference
sentences = [
    'can sweet potatoes cause itching?',
    'People with a true potato allergy may react immediately after touching, peeling, or eating potatoes. Symptoms may vary from person to person, but typical symptoms of a potato allergy include: rhinitis, including itchy or stinging eyes, a runny or stuffy nose, and sneezing.',
    'riding a bike does not cause pollution',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

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

<!-- ### Direct Usage (Transformers) <details><summary>Click to see the direct usage in Transformers</summary> </details> --> <!-- ### Downstream Usage (Sentence Transformers) You can finetune this model on your own dataset. <details><summary>Click to expand</summary> </details> --> <!-- ### Out-of-Scope Use *List how the model may foreseeably be misused and address what users ought not to do with the model.* -->

Evaluation

Metrics

Semantic Similarity

• Dataset: sts-test
• Evaluated with <code>EmbeddingSimilarityEvaluator</code>

Binary Classification

• Dataset: allNLI-dev
• Evaluated with <code>BinaryClassificationEvaluator</code>

Binary Classification

• Dataset: Qnli-dev
• Evaluated with <code>BinaryClassificationEvaluator</code>

<!-- ## Bias, Risks and Limitations *What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.* --> <!-- ### Recommendations *What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.* -->

Training Details

Training Dataset

Unnamed Dataset

• Size: 32,500 training samples
• Columns: <code>sentence1</code> and <code>sentence2</code>
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	<ul><li>min: 4 tokens</li><li>mean: 29.6 tokens</li><li>max: 369 tokens</li></ul>	<ul><li>min: 2 tokens</li><li>mean: 58.01 tokens</li><li>max: 437 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>The song ‘Fashion for His Love’ by Lady Gaga is a tribute to which late fashion designer?</code>	<code>Fashion Of His Love by Lady Gaga Songfacts Fashion Of His Love by Lady Gaga Songfacts Songfacts Gaga explained in a tweet that this track from her Born This Way Special Edition album is about the late Alexander McQueen. The fashion designer committed suicide by hanging on February 11, 2010 and Gaga was deeply affected by the tragic death of McQueen, who was a close personal friend. That same month, she performed at the 2010 Brit Awards wearing one of his couture creations and she also paid tribute to her late friend by setting the date on the prison security cameras in her Telephone video as the same day that McQueen's body was discovered in his London home.</code>
  <code>e. in solids the atoms are closely locked in position and can only vibrate, in liquids the atoms and molecules are more loosely connected and can collide with and move past one another, while in gases the atoms or molecules are free to move independently, colliding frequently.</code>	<code>Within a substance, atoms that collide frequently and move independently of one another are most likely in a gas</code>
  <code>Helen Lederer is an English comedian .</code>	<code>Helen Lederer ( born 24 September 1954 ) is an English : //www.scotsman.com/news/now-or-never-1-1396369 comedian , writer and actress who emerged as part of the alternative comedy boom at the beginning of the 1980s .</code>

• Loss: <code>GISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

Evaluation Dataset

Unnamed Dataset

• Size: 1,664 evaluation samples
• Columns: <code>sentence1</code> and <code>sentence2</code>
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2
  type	string	string
  details	<ul><li>min: 4 tokens</li><li>mean: 29.01 tokens</li><li>max: 367 tokens</li></ul>	<ul><li>min: 2 tokens</li><li>mean: 56.14 tokens</li><li>max: 389 tokens</li></ul>

• Samples:

  sentence1	sentence2
  <code>What planet did the voyager 1 spacecraft visit in 1980?</code>	<code>The Voyager 1 spacecraft visited Saturn in 1980. Voyager 2 followed in 1981. These probes sent back detailed pictures of Saturn, its rings, and some of its moons ( Figure below ). From the Voyager data, we learned what Saturn’s rings are made of. They are particles of water and ice with a little bit of dust. There are several gaps in the rings. These gaps were cleared out by moons within the rings. Gravity attracts dust and gas to the moon from the ring. This leaves a gap in the rings. Other gaps in the rings are caused by the competing forces of Saturn and its moons outside the rings.</code>
  <code>Diffusion Diffusion is a process where atoms or molecules move from areas of high concentration to areas of low concentration.</code>	<code>Diffusion is the process in which a substance naturally moves from an area of higher to lower concentration.</code>
  <code>Who had an 80s No 1 with Don't You Want Me?</code>	<code>The Human League - Don't You Want Me - YouTube The Human League - Don't You Want Me Want to watch this again later? Sign in to add this video to a playlist. Need to report the video? Sign in to report inappropriate content. Rating is available when the video has been rented. This feature is not available right now. Please try again later. Uploaded on Feb 27, 2009 Music video by The Human League performing Don't You Want Me (2003 Digital Remaster). Category</code>

• Loss: <code>GISTEmbedLoss</code> with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 768, '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()
  ), 'temperature': 0.025}
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 256
• lr_scheduler_type: cosine_with_min_lr
• lr_scheduler_kwargs: {'num_cycles': 0.5, 'min_lr': 3.3333333333333337e-06}
• warmup_ratio: 0.33
• save_safetensors: False
• fp16: True
• push_to_hub: True
• hub_model_id: bobox/DeBERTa3-s-CustomPoolin-toytest3-step1-checkpoints-tmp
• hub_strategy: all_checkpoints
• batch_sampler: no_duplicates

All Hyperparameters

<details><summary>Click to expand</summary>

• 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: 256
• 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: 5e-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: 3
• max_steps: -1
• lr_scheduler_type: cosine_with_min_lr
• lr_scheduler_kwargs: {'num_cycles': 0.5, 'min_lr': 3.3333333333333337e-06}
• warmup_ratio: 0.33
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• logging_nan_inf_filter: True
• save_safetensors: False
• 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}
• 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: True
• resume_from_checkpoint: None
• hub_model_id: bobox/DeBERTa3-s-CustomPoolin-toytest3-step1-checkpoints-tmp
• hub_strategy: all_checkpoints
• hub_private_repo: False
• hub_always_push: False
• gradient_checkpointing: False
• 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

</details>

Training Logs

<details><summary>Click to expand</summary>

The model card is truncated. Read the rest on Hugging Face.