Scaling Deep Contrastive Learning Batch Size under Memory Limited Setup
Paper
•
2101.06983
•
Published
•
1
This is a sentence-transformers model finetuned from huggingface/CodeBERTa-small-v1 on the soco_train_java dataset. 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.
SentenceTransformer(
(0): Transformer({'max_seq_length': 512, 'do_lower_case': False, 'architecture': 'RobertaModel'})
(1): Pooling({'word_embedding_dimension': 768, '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})
)
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("buelfhood/SOCO-Java-codeberta-cmnrl-triplets-ep1-bs512-lr2e-05-split0.1")
# Run inference
sentences = [
'import java.net.*;\nimport java.io.*;\n\npublic class BruteForce {\n private String strUserName;\n private String strURL;\n private int iAttempts;\n \n public BruteForce(String strURL,String strUserName) {\n this.strURL = strURL;\n this.strUserName = strUserName;\n this.iAttempts = 0 ;\n\n }\n \n public String getPassword(){\n URL u;\n String result ="";\n PassGenBrute PG = new PassGenBrute(3);\n URLConnection uc;\n String strPassword = new String();\n String strEncode;\n try{\n while (result.compareTo("HTTP/1.1 200 OK")!=0){\n \n strEncode = PG.getNewPassword();\n u = new URL(strURL);\n uc = u.openConnection();\n uc.setDoInput(true);\n uc.setDoOutput(true);\n strPassword = strEncode;\n strEncode = strUserName + ":" + strEncode;\n \n strEncode = new String(Base64.encode(strEncode.getBytes()));\n uc.setRequestProperty("Authorization"," " + strEncode);\n \n result = uc.getHeaderField(0);\n uc = null;\n u = null;\n iAttempts++;\n }\n\n }\n catch (Exception me) {\n System.out.println("MalformedURLException: "+me);\n }\n return(strPassword);\n }\n \n public int getAttempts(){\n return (iAttempts);\n };\n \n public static void main (String arg[]){\n timeStart = 0;\n timeEnd = 0;\n \n if (arg.length == 2) {\n BruteForce BF = new BruteForce(arg[0],arg[1]);\n System.out.println("Processing ... ");\n timeStart = System.currentTimeMillis();\n \n System.out.println("Password = " + BF.getPassword());\n timeEnd = System.currentTimeMillis();\n System.out.println("Total Time Taken = " + (timeEnd - timeStart) + " (msec)");\n System.out.println("Total Attempts = " + BF.getAttempts());\n }\n else {\n System.out.println("[Usage] java BruteForce <URL> <USERNAME>");\n\n }\n\n }\n}\n\nclass PassGenBrute {\n private char[] password;\n public PassGenBrute(int lenght) {\n password = new char[lenght];\n for (int i = 0; i < lenght; i++){\n password[i] = 65;\n }\n password[0]--;\n }\n \n public String getNewPassword()\n throws PasswordFailureException{\n password[0]++;\n\n try {\n for (int i=0; i<password.length ; i++){\n if (password[i] == 90) {\n password[i] = 97;\n }\n if (password[i] > 122) {\n password[i] = 65;\n password[i+1]++;\n }\n }\n }\n catch (RuntimeException re){\n throw new PasswordFailureException ();\n }\n return new String(password);\n }\n}\n\nclass PasswordFailureException extends RuntimeException {\n\n public PasswordFailureException() {\n }\n}',
'import java.net.*;\nimport java.io.*;\n\n\npublic class Dictionary {\n private String strUserName;\n private String strURL;\n private String strDictPath;\n private int iAttempts;\n\n \n public Dictionary(String strURL,String strUserName,String strDictPath) {\n this.strURL = strURL;\n this.strUserName = strUserName;\n this.iAttempts = 0 ;\n this.strDictPath = strDictPath;\n }\n \n\n public String getPassword(){\n URL u;\n String result ="";\n PassGenDict PG = new PassGenDict(3,strDictPath);\n URLConnection uc;\n String strPassword = new String();\n String strEncode;\n try{\n while (result.compareTo("HTTP/1.1 200 OK")!=0){\n \n strEncode = PG.getNewPassword();\n u = new URL(strURL);\n uc = u.openConnection();\n uc.setDoInput(true);\n uc.setDoOutput(true);\n strPassword = strEncode;\n strEncode = strUserName + ":" + strEncode;\n \n strEncode = new String(Base64.encode(strEncode.getBytes()));\n uc.setRequestProperty("Authorization"," " + strEncode);\n \n result = uc.getHeaderField(0);\n uc = null;\n u = null;\n iAttempts++;\n }\n\n }\n catch (Exception me) {\n System.out.println("MalformedURLException: "+me);\n }\n return(strPassword);\n }\n \n public int getAttempts(){\n return (iAttempts);\n };\n \n public static void main(String arg[]){\n timeStart = 0;\n timeEnd = 0;\n \n if (arg.length == 3) {\n Dictionary BF = new Dictionary(arg[0],arg[1],arg[2]);\n\n System.out.println("Processing ... ");\n timeStart = System.currentTimeMillis();\n System.out.println("Password = " + BF.getPassword());\n timeEnd = System.currentTimeMillis();\n System.out.println("Total Time Taken = " + (timeEnd - timeStart) + " (msec)");\n System.out.println("Total Attempts = " + BF.getAttempts());\n }\n else {\n System.out.println("[Usage] java BruteForce <URL> <USERNAME> <Dictionary path>");\n\n }\n\n }\n}\n\n\nclass PassGenDict {\n\n private char[] password;\n private String line;\n int iPassLenght;\n private BufferedReader inputFile;\n public PassGenDict(int lenght, String strDictPath) {\n try{\n inputFile = new BufferedReader(new FileReader(strDictPath));\n }\n catch (Exception e){\n }\n iPassLenght = lenght;\n }\n \n public String getNewPassword()\n throws PasswordFailureException{\n try {\n {\n line = inputFile.readLine();\n }while (line.length() != iPassLenght);\n\n }\n catch (Exception e){\n throw new PasswordFailureException ();\n }\n return (line);\n }\n}\n\nclass PasswordFailureException extends RuntimeException {\n\n public PasswordFailureException() {\n }\n}',
'import java.util.*;\nimport java.io.*;\nimport javax.swing.text.html.*;\n\n\npublic class WatchDog {\n\n public WatchDog() {\n\n }\n public static void main (String args[]) {\n DataInputStream newin;\n\n try{\n System.out.println("ishti");\n\n System.out.println("Downloading first copy");\n Runtime.getRuntime().exec("wget http://www.cs.rmit.edu./students/ -O oldfile.html");\n String[] cmdDiff = {"//sh", "-c", "diff oldfile.html newfile.html > Diff.txt"};\n String[] cmdMail = {"//sh", "-c", "mailx -s \\"Diffrence\\" \\"@cs.rmit.edu.\\" < Diff.txt"};\n while(true){\n Thread.sleep(24*60*60*1000);\n System.out.println("Downloading new copy");\n Runtime.getRuntime().exec("wget http://www.cs.rmit.edu./students/ -O newfile.html");\n Thread.sleep(2000);\n Runtime.getRuntime().exec(cmdDiff);\n Thread.sleep(2000);\n newin = new DataInputStream( new FileInputStream( "Diff.txt"));\n if (newin.readLine() != null){\n System.out.println("Sending Mail");\n Runtime.getRuntime().exec(cmdMail);\n Runtime.getRuntime().exec("cp newfile.html oldfile.html");\n\n }\n }\n\n }\n catch(Exception e){\n e.printStackTrace();\n }\n\n }\n\n}',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]
# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[ 1.0000, 0.9527, -0.2369],
# [ 0.9527, 1.0000, -0.2560],
# [-0.2369, -0.2560, 1.0000]])
anchor_code, positive_code, and negative_code| anchor_code | positive_code | negative_code | |
|---|---|---|---|
| type | string | string | string |
| details |
|
|
|
| anchor_code | positive_code | negative_code |
|---|---|---|
|
|
|
|
|
|
|
|
|
CachedMultipleNegativesRankingLoss with these parameters:{
"scale": 20.0,
"similarity_fct": "cos_sim",
"mini_batch_size": 32,
"gather_across_devices": false
}
anchor_code, positive_code, and negative_code| anchor_code | positive_code | negative_code | |
|---|---|---|---|
| type | string | string | string |
| details |
|
|
|
| anchor_code | positive_code | negative_code |
|---|---|---|
|
|
import java.net.; |
|
|
|
import java.io.; |
import java.io.; |
|
CachedMultipleNegativesRankingLoss with these parameters:{
"scale": 20.0,
"similarity_fct": "cos_sim",
"mini_batch_size": 32,
"gather_across_devices": false
}
per_device_train_batch_size: 512learning_rate: 2e-05num_train_epochs: 1warmup_ratio: 0.1fp16: Truebatch_sampler: no_duplicatesoverwrite_output_dir: Falsedo_predict: Falseeval_strategy: noprediction_loss_only: Trueper_device_train_batch_size: 512per_device_eval_batch_size: 8per_gpu_train_batch_size: Noneper_gpu_eval_batch_size: Nonegradient_accumulation_steps: 1eval_accumulation_steps: Nonetorch_empty_cache_steps: Nonelearning_rate: 2e-05weight_decay: 0.0adam_beta1: 0.9adam_beta2: 0.999adam_epsilon: 1e-08max_grad_norm: 1.0num_train_epochs: 1max_steps: -1lr_scheduler_type: linearlr_scheduler_kwargs: {}warmup_ratio: 0.1warmup_steps: 0log_level: passivelog_level_replica: warninglog_on_each_node: Truelogging_nan_inf_filter: Truesave_safetensors: Truesave_on_each_node: Falsesave_only_model: Falserestore_callback_states_from_checkpoint: Falseno_cuda: Falseuse_cpu: Falseuse_mps_device: Falseseed: 42data_seed: Nonejit_mode_eval: Falseuse_ipex: Falsebf16: Falsefp16: Truefp16_opt_level: O1half_precision_backend: autobf16_full_eval: Falsefp16_full_eval: Falsetf32: Nonelocal_rank: 0ddp_backend: Nonetpu_num_cores: Nonetpu_metrics_debug: Falsedebug: []dataloader_drop_last: Falsedataloader_num_workers: 0dataloader_prefetch_factor: Nonepast_index: -1disable_tqdm: Falseremove_unused_columns: Truelabel_names: Noneload_best_model_at_end: Falseignore_data_skip: Falsefsdp: []fsdp_min_num_params: 0fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}fsdp_transformer_layer_cls_to_wrap: Noneaccelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}parallelism_config: Nonedeepspeed: Nonelabel_smoothing_factor: 0.0optim: adamw_torch_fusedoptim_args: Noneadafactor: Falsegroup_by_length: Falselength_column_name: lengthddp_find_unused_parameters: Noneddp_bucket_cap_mb: Noneddp_broadcast_buffers: Falsedataloader_pin_memory: Truedataloader_persistent_workers: Falseskip_memory_metrics: Trueuse_legacy_prediction_loop: Falsepush_to_hub: Falseresume_from_checkpoint: Nonehub_model_id: Nonehub_strategy: every_savehub_private_repo: Nonehub_always_push: Falsehub_revision: Nonegradient_checkpointing: Falsegradient_checkpointing_kwargs: Noneinclude_inputs_for_metrics: Falseinclude_for_metrics: []eval_do_concat_batches: Truefp16_backend: autopush_to_hub_model_id: Nonepush_to_hub_organization: Nonemp_parameters: auto_find_batch_size: Falsefull_determinism: Falsetorchdynamo: Noneray_scope: lastddp_timeout: 1800torch_compile: Falsetorch_compile_backend: Nonetorch_compile_mode: Noneinclude_tokens_per_second: Falseinclude_num_input_tokens_seen: Falseneftune_noise_alpha: Noneoptim_target_modules: Nonebatch_eval_metrics: Falseeval_on_start: Falseuse_liger_kernel: Falseliger_kernel_config: Noneeval_use_gather_object: Falseaverage_tokens_across_devices: Falseprompts: Nonebatch_sampler: no_duplicatesmulti_dataset_batch_sampler: proportionalrouter_mapping: {}learning_rate_mapping: {}@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",
}
@misc{gao2021scaling,
title={Scaling Deep Contrastive Learning Batch Size under Memory Limited Setup},
author={Luyu Gao and Yunyi Zhang and Jiawei Han and Jamie Callan},
year={2021},
eprint={2101.06983},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
Base model
huggingface/CodeBERTa-small-v1