Matjac5/MNLP_M3_RAG_MODEL_data_mixture_maths

1---2tags:3- sentence-transformers4- sentence-similarity5- feature-extraction6- generated_from_trainer7- dataset_size:2688618- loss:MultipleNegativesRankingLoss9base_model: Qwen/Qwen3-0.6B-Base10widget:11- source_sentence: 'There are seven thieves. They stole diamonds from a diamond merchant12    and ran away. While running, night sets in and they decide to rest in the jungle.13 14    When everybody was sleeping, two of them woke up and decided to divide the diamonds15    equally among themselves. But when they divided the diamonds equally, one diamond16    is left.17 18    So they woke up the 3rd thief and tried to divide the diamonds equally again but19    still one diamond was left. Then they woke up the 4th thief to divide the diamonds20    equally again, and again one diamond was left. This happened with the 5th and21    6th thief – one diamond was still left.22 23    Finally, they woke up the 7th thief and this time the diamonds were divided equally.24 25    How many diamonds did they steal in total?'26  sentences:27  - ''''28  - ''''29  - e30- source_sentence: 'praveen starts business with rs . 3220 and after 5 months , hari31    joins with praveen as his partner . after a year , the profit is divided in the32    ratio 2 : 3 . what is hari ’ s contribution in the capital ?'33  sentences:34  - s35  - '5'36  - '['37- source_sentence: 'Which of the following is material of choice in class V38 39    cavity with abfraction?'40  sentences:41  - '['42  - t43  - G44- source_sentence: A right circular cylinder has a height of 25 and a radius of 5.45    A rectangular solid with a height of 15 and a square base, is placed in the cylinder46    such that each of the corners of the solid is tangent to the cylinder wall. Liquid47    is then poured into the cylinder such that it reaches the rim. What is the volume48    of the liquid?49  sentences:50  - '5'51  - '['52  - '2'53- source_sentence: Cerebral angiography was performed by -54  sentences:55  - S56  - t57  - '2'58pipeline_tag: sentence-similarity59library_name: sentence-transformers60---61 62# SentenceTransformer based on Qwen/Qwen3-0.6B-Base63 64This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [Qwen/Qwen3-0.6B-Base](https://huggingface.co/Qwen/Qwen3-0.6B-Base). 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.65 66## Model Details67 68### Model Description69- **Model Type:** Sentence Transformer70- **Base model:** [Qwen/Qwen3-0.6B-Base](https://huggingface.co/Qwen/Qwen3-0.6B-Base) <!-- at revision 11214f7f3465775dcce23c3752ecea5a42ee0ddc -->71- **Maximum Sequence Length:** 128 tokens72- **Output Dimensionality:** 1024 dimensions73- **Similarity Function:** Cosine Similarity74<!-- - **Training Dataset:** Unknown -->75<!-- - **Language:** Unknown -->76<!-- - **License:** Unknown -->77 78### Model Sources79 80- **Documentation:** [Sentence Transformers Documentation](https://sbert.net)81- **Repository:** [Sentence Transformers on GitHub](https://github.com/UKPLab/sentence-transformers)82- **Hugging Face:** [Sentence Transformers on Hugging Face](https://huggingface.co/models?library=sentence-transformers)83 84### Full Model Architecture85 86```87SentenceTransformer(88  (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: Qwen3Model 89  (1): Pooling({'word_embedding_dimension': 1024, '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})90)91```92 93## Usage94 95### Direct Usage (Sentence Transformers)96 97First install the Sentence Transformers library:98 99```bash100pip install -U sentence-transformers101```102 103Then you can load this model and run inference.104```python105from sentence_transformers import SentenceTransformer106 107# Download from the 🤗 Hub108model = SentenceTransformer("sentence_transformers_model_id")109# Run inference110sentences = [111    'Cerebral angiography was performed by -',112    'S',113    '2',114]115embeddings = model.encode(sentences)116print(embeddings.shape)117# [3, 1024]118 119# Get the similarity scores for the embeddings120similarities = model.similarity(embeddings, embeddings)121print(similarities.shape)122# [3, 3]123```124 125<!--126### Direct Usage (Transformers)127 128<details><summary>Click to see the direct usage in Transformers</summary>129 130</details>131-->132 133<!--134### Downstream Usage (Sentence Transformers)135 136You can finetune this model on your own dataset.137 138<details><summary>Click to expand</summary>139 140</details>141-->142 143<!--144### Out-of-Scope Use145 146*List how the model may foreseeably be misused and address what users ought not to do with the model.*147-->148 149<!--150## Bias, Risks and Limitations151 152*What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.*153-->154 155<!--156### Recommendations157 158*What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.*159-->160 161## Training Details162 163### Training Dataset164 165#### Unnamed Dataset166 167* Size: 268,861 training samples168* Columns: <code>sentence_0</code> and <code>sentence_1</code>169* Approximate statistics based on the first 1000 samples:170  |         | sentence_0                                                                        | sentence_1                                                                      |171  |:--------|:----------------------------------------------------------------------------------|:--------------------------------------------------------------------------------|172  | type    | string                                                                            | string                                                                          |173  | details | <ul><li>min: 5 tokens</li><li>mean: 48.3 tokens</li><li>max: 128 tokens</li></ul> | <ul><li>min: 0 tokens</li><li>mean: 0.97 tokens</li><li>max: 1 tokens</li></ul> |174* Samples:175  | sentence_0                                                                                                                                                    | sentence_1     |176  |:--------------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------|177  | <code>A 1200 m long train crosses a tree in 120 sec, how much time will I take to pass a platform 1100 m long?</code>                                         | <code>'</code> |178  | <code>What is the opposite of rarefaction zones, where air molecules in waves are loosely packed?</code>                                                      | <code>[</code> |179  | <code>if w is 40 percent less than e , e is 40 percent less than y , and z is 46 percent less than y , then z is greater than w by what percent of w ?</code> | <code>%</code> |180* Loss: [<code>MultipleNegativesRankingLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#multiplenegativesrankingloss) with these parameters:181  ```json182  {183      "scale": 20.0,184      "similarity_fct": "cos_sim"185  }186  ```187 188### Training Hyperparameters189#### Non-Default Hyperparameters190 191- `per_device_train_batch_size`: 64192- `per_device_eval_batch_size`: 64193- `num_train_epochs`: 4194- `fp16`: True195- `multi_dataset_batch_sampler`: round_robin196 197#### All Hyperparameters198<details><summary>Click to expand</summary>199 200- `overwrite_output_dir`: False201- `do_predict`: False202- `eval_strategy`: no203- `prediction_loss_only`: True204- `per_device_train_batch_size`: 64205- `per_device_eval_batch_size`: 64206- `per_gpu_train_batch_size`: None207- `per_gpu_eval_batch_size`: None208- `gradient_accumulation_steps`: 1209- `eval_accumulation_steps`: None210- `torch_empty_cache_steps`: None211- `learning_rate`: 5e-05212- `weight_decay`: 0.0213- `adam_beta1`: 0.9214- `adam_beta2`: 0.999215- `adam_epsilon`: 1e-08216- `max_grad_norm`: 1217- `num_train_epochs`: 4218- `max_steps`: -1219- `lr_scheduler_type`: linear220- `lr_scheduler_kwargs`: {}221- `warmup_ratio`: 0.0222- `warmup_steps`: 0223- `log_level`: passive224- `log_level_replica`: warning225- `log_on_each_node`: True226- `logging_nan_inf_filter`: True227- `save_safetensors`: True228- `save_on_each_node`: False229- `save_only_model`: False230- `restore_callback_states_from_checkpoint`: False231- `no_cuda`: False232- `use_cpu`: False233- `use_mps_device`: False234- `seed`: 42235- `data_seed`: None236- `jit_mode_eval`: False237- `use_ipex`: False238- `bf16`: False239- `fp16`: True240- `fp16_opt_level`: O1241- `half_precision_backend`: auto242- `bf16_full_eval`: False243- `fp16_full_eval`: False244- `tf32`: None245- `local_rank`: 0246- `ddp_backend`: None247- `tpu_num_cores`: None248- `tpu_metrics_debug`: False249- `debug`: []250- `dataloader_drop_last`: False251- `dataloader_num_workers`: 0252- `dataloader_prefetch_factor`: None253- `past_index`: -1254- `disable_tqdm`: False255- `remove_unused_columns`: True256- `label_names`: None257- `load_best_model_at_end`: False258- `ignore_data_skip`: False259- `fsdp`: []260- `fsdp_min_num_params`: 0261- `fsdp_config`: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}262- `fsdp_transformer_layer_cls_to_wrap`: None263- `accelerator_config`: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}264- `deepspeed`: None265- `label_smoothing_factor`: 0.0266- `optim`: adamw_torch267- `optim_args`: None268- `adafactor`: False269- `group_by_length`: False270- `length_column_name`: length271- `ddp_find_unused_parameters`: None272- `ddp_bucket_cap_mb`: None273- `ddp_broadcast_buffers`: False274- `dataloader_pin_memory`: True275- `dataloader_persistent_workers`: False276- `skip_memory_metrics`: True277- `use_legacy_prediction_loop`: False278- `push_to_hub`: False279- `resume_from_checkpoint`: None280- `hub_model_id`: None281- `hub_strategy`: every_save282- `hub_private_repo`: None283- `hub_always_push`: False284- `gradient_checkpointing`: False285- `gradient_checkpointing_kwargs`: None286- `include_inputs_for_metrics`: False287- `include_for_metrics`: []288- `eval_do_concat_batches`: True289- `fp16_backend`: auto290- `push_to_hub_model_id`: None291- `push_to_hub_organization`: None292- `mp_parameters`: 293- `auto_find_batch_size`: False294- `full_determinism`: False295- `torchdynamo`: None296- `ray_scope`: last297- `ddp_timeout`: 1800298- `torch_compile`: False299- `torch_compile_backend`: None300- `torch_compile_mode`: None301- `include_tokens_per_second`: False302- `include_num_input_tokens_seen`: False303- `neftune_noise_alpha`: None304- `optim_target_modules`: None305- `batch_eval_metrics`: False306- `eval_on_start`: False307- `use_liger_kernel`: False308- `eval_use_gather_object`: False309- `average_tokens_across_devices`: False310- `prompts`: None311- `batch_sampler`: batch_sampler312- `multi_dataset_batch_sampler`: round_robin313 314</details>315 316### Training Logs317| Epoch  | Step  | Training Loss |318|:------:|:-----:|:-------------:|319| 0.1190 | 500   | 4.0939        |320| 0.2380 | 1000  | 3.7716        |321| 0.3571 | 1500  | 0.0           |322| 0.4761 | 2000  | 0.0           |323| 0.5951 | 2500  | 0.0           |324| 0.7141 | 3000  | 0.0           |325| 0.8331 | 3500  | 0.0           |326| 0.9522 | 4000  | 0.0           |327| 1.0712 | 4500  | 0.0           |328| 1.1902 | 5000  | 0.0           |329| 1.3092 | 5500  | 0.0           |330| 1.4282 | 6000  | 0.0           |331| 1.5473 | 6500  | 0.0           |332| 1.6663 | 7000  | 0.0           |333| 1.7853 | 7500  | 0.0           |334| 1.9043 | 8000  | 0.0           |335| 2.0233 | 8500  | 0.0           |336| 2.1423 | 9000  | 0.0           |337| 2.2614 | 9500  | 0.0           |338| 2.3804 | 10000 | 0.0           |339| 2.4994 | 10500 | 0.0           |340| 2.6184 | 11000 | 0.0           |341| 2.7374 | 11500 | 0.0           |342| 2.8565 | 12000 | 0.0           |343| 2.9755 | 12500 | 0.0           |344| 3.0945 | 13000 | 0.0           |345| 3.2135 | 13500 | 0.0           |346| 3.3325 | 14000 | 0.0           |347| 3.4516 | 14500 | 0.0           |348| 3.5706 | 15000 | 0.0           |349| 3.6896 | 15500 | 0.0           |350| 3.8086 | 16000 | 0.0           |351| 3.9276 | 16500 | 0.0           |352 353 354### Framework Versions355- Python: 3.11.13356- Sentence Transformers: 4.1.0357- Transformers: 4.52.4358- PyTorch: 2.6.0+cu124359- Accelerate: 1.7.0360- Datasets: 3.6.0361- Tokenizers: 0.21.1362 363## Citation364 365### BibTeX366 367#### Sentence Transformers368```bibtex369@inproceedings{reimers-2019-sentence-bert,370    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",371    author = "Reimers, Nils and Gurevych, Iryna",372    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",373    month = "11",374    year = "2019",375    publisher = "Association for Computational Linguistics",376    url = "https://arxiv.org/abs/1908.10084",377}378```379 380#### MultipleNegativesRankingLoss381```bibtex382@misc{henderson2017efficient,383    title={Efficient Natural Language Response Suggestion for Smart Reply},384    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},385    year={2017},386    eprint={1705.00652},387    archivePrefix={arXiv},388    primaryClass={cs.CL}389}390```391 392<!--393## Glossary394 395*Clearly define terms in order to be accessible across audiences.*396-->397 398<!--399## Model Card Authors400 401*Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.*402-->403 404<!--405## Model Card Contact406 407*Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.*408-->