SentenceTransformer based on pkqinys/shawhin-clip-title-thumbnail-embeddings

This is a sentence-transformers model finetuned from pkqinys/shawhin-clip-title-thumbnail-embeddings on the shawhin-yt-title-thumbnail-pairs dataset. It maps sentences & paragraphs to a None-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: pkqinys/shawhin-clip-title-thumbnail-embeddings
• Maximum Sequence Length: None tokens
• Output Dimensionality: None dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • shawhin-yt-title-thumbnail-pairs

Model Sources

• Documentation: Sentence Transformers Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): CLIPModel()
)

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("sentence_transformers_model_id")
# Run inference
sentences = [
    'The Hugging Face Transformers Library | Example Code + Chatbot UI with Gradio',
    'How to Evaluate (and Improve) Your LLM Apps',
    'How to Improve LLMs with Tools (ft. OpenAI Agents SDK)',
]
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

Triplet

• Datasets: yt-title-thumbnail-train and yt-title-thumbnail-valid
• Evaluated with TripletEvaluator

Metric	yt-title-thumbnail-train	yt-title-thumbnail-valid
cosine_accuracy	1.0	0.9375

Training Details

Training Dataset

shawhin-yt-title-thumbnail-pairs

• Dataset: shawhin-yt-title-thumbnail-pairs at 24578b7
• Size: 75 training samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 75 samples:

  	anchor	positive	negative
  type	PIL.JpegImagePlugin.JpegImageFile	string	string
  details		min: 8 tokens mean: 15.52 tokens max: 27 tokens	min: 8 tokens mean: 15.49 tokens max: 27 tokens

• Samples:

  anchor	positive	negative
  	A Practical Introduction to Large Language Models (LLMs)	Prompt Engineering: How to Trick AI into Solving Your Problems
  	How to Build a Notion AI Agent (in 18 minutes)	4 Ways to Measure Fat Tails with Python (+ Example Code)
  	Context Engineering Explained (5 Practical Tips)	Why I Quit My $150,000 Data Science Job

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Evaluation Dataset

shawhin-yt-title-thumbnail-pairs

• Dataset: shawhin-yt-title-thumbnail-pairs at 24578b7
• Size: 16 evaluation samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 16 samples:

  	anchor	positive	negative
  type	PIL.JpegImagePlugin.JpegImageFile	string	string
  details		min: 8 tokens mean: 14.5 tokens max: 22 tokens	min: 9 tokens mean: 13.56 tokens max: 18 tokens

• Samples:

  anchor	positive	negative
  	Principal Component Analysis (PCA)	Introduction & Example (Python) Code
  	5 Reasons Why Every Data Scientist Should Consider Freelancing	Text Embeddings, Classification, and Semantic Search (w/ Python Code)
  	5 AI Projects You Can Build This Weekend (with Python)	The Wavelet Transform

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: epoch
• per_device_train_batch_size: 4
• per_device_eval_batch_size: 4
• learning_rate: 1e-05
• num_train_epochs: 4

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: epoch
• prediction_loss_only: True
• per_device_train_batch_size: 4
• 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: 1e-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: 4
• max_steps: -1
• lr_scheduler_type: linear
• 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: 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: 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}
• 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
• 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
• use_liger_kernel: False
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Click to expand

Epoch	Step	Training Loss	Validation Loss	yt-title-thumbnail-train_cosine_accuracy	yt-title-thumbnail-valid_cosine_accuracy
0	0	-	-	1.0	0.9375
0.2	1	0.5678	-	-	-
0.4	2	0.8427	-	-	-
0.6	3	0.8475	-	-	-
0.8	4	0.9326	-	-	-
1.0	5	0.9615	1.7712	1.0	0.9375
0.1	1	0.1598	-	-	-
0.2	2	0.084	-	-	-
0.3	3	0.2701	-	-	-
0.4	4	0.4036	-	-	-
0.5	5	0.3907	-	-	-
0.6	6	0.4165	-	-	-
0.7	7	0.6304	-	-	-
0.8	8	0.3674	-	-	-
0.9	9	0.7016	-	-	-
1.0	10	0.2821	1.2305	-	-
1.1	11	0.2749	-	-	-
1.2	12	0.4927	-	-	-
1.3	13	0.2579	-	-	-
1.4	14	0.5595	-	-	-
1.5	15	0.5023	-	-	-
1.6	16	0.2853	-	-	-
1.7	17	0.44	-	-	-
1.8	18	0.5634	-	-	-
1.9	19	0.4209	-	-	-
2.0	20	0.0825	1.2300	-	-
0.0526	1	0.0705	-	-	-
0.1053	2	0.0566	-	-	-
0.1579	3	0.0493	-	-	-
0.2105	4	0.0315	-	-	-
0.2632	5	0.0866	-	-	-
0.3158	6	0.069	-	-	-
0.3684	7	0.3537	-	-	-
0.4211	8	0.122	-	-	-
0.4737	9	0.1622	-	-	-
0.5263	10	0.1299	-	-	-
0.5789	11	0.2733	-	-	-
0.6316	12	0.1462	-	-	-
0.6842	13	0.3868	-	-	-
0.7368	14	0.2076	-	-	-
0.7895	15	0.1524	-	-	-
0.8421	16	0.2408	-	-	-
0.8947	17	0.3624	-	-	-
0.9474	18	0.3834	-	-	-
1.0	19	0.2456	0.7823	-	-
1.0526	20	0.1087	-	-	-
1.1053	21	0.1406	-	-	-
1.1579	22	0.1029	-	-	-
1.2105	23	0.1475	-	-	-
1.2632	24	0.1049	-	-	-
1.3158	25	0.053	-	-	-
1.3684	26	0.3995	-	-	-
1.4211	27	0.1227	-	-	-
1.4737	28	0.2695	-	-	-
1.5263	29	0.2224	-	-	-
1.5789	30	0.0945	-	-	-
1.6316	31	0.1475	-	-	-
1.6842	32	0.4774	-	-	-
1.7368	33	0.0315	-	-	-
1.7895	34	0.1688	-	-	-
1.8421	35	0.3641	-	-	-
1.8947	36	0.1004	-	-	-
1.9474	37	0.1414	-	-	-
2.0	38	0.067	0.7837	-	-
2.0526	39	0.1435	-	-	-
2.1053	40	0.2146	-	-	-
2.1579	41	0.2392	-	-	-
2.2105	42	0.1392	-	-	-
2.2632	43	0.3378	-	-	-
2.3158	44	0.0715	-	-	-
2.3684	45	0.1154	-	-	-
2.4211	46	0.2249	-	-	-
2.4737	47	0.0407	-	-	-
2.5263	48	0.0414	-	-	-
2.5789	49	0.1295	-	-	-
2.6316	50	0.0922	-	-	-
2.6842	51	0.077	-	-	-
2.7368	52	0.4554	-	-	-
2.7895	53	0.0699	-	-	-
2.8421	54	0.0663	-	-	-
2.8947	55	0.2612	-	-	-
2.9474	56	0.1907	-	-	-
3.0	57	0.1049	0.7811	-	-
3.0526	58	0.1027	-	-	-
3.1053	59	0.2408	-	-	-
3.1579	60	0.0248	-	-	-
3.2105	61	0.2142	-	-	-
3.2632	62	0.1579	-	-	-
3.3158	63	0.0789	-	-	-
3.3684	64	0.0668	-	-	-
3.4211	65	0.1484	-	-	-
3.4737	66	0.3956	-	-	-
3.5263	67	0.1063	-	-	-
3.5789	68	0.4022	-	-	-
3.6316	69	0.5607	-	-	-
3.6842	70	0.0283	-	-	-
3.7368	71	0.0781	-	-	-
3.7895	72	0.248	-	-	-
3.8421	73	0.08	-	-	-
3.8947	74	0.2495	-	-	-
3.9474	75	0.1528	-	-	-
4.0	76	0.2234	-	-	-

Framework Versions

• Python: 3.11.13
• Sentence Transformers: 3.3.1
• Transformers: 4.48.0
• PyTorch: 2.3.0
• Accelerate: 1.10.1
• Datasets: 4.1.1
• Tokenizers: 0.21.4

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}
}