v1.0.0

What's Changed

• Improve chaining speed by @maxjeblick in #426
• upd transformers and other dependencies by @pascal-pfeiffer in #428
• Improve Data Visualizations by @maxjeblick in #421
• Fix conversation chain format output by @maxjeblick in #434
• RLHF large rollout by @pascal-pfeiffer in #369
• update transformers==4.34.0 by @maxjeblick in #439
• Add parent_id_column to hash by @maxjeblick in #437
• fix data viz visibility by @maxjeblick in #442
• Cast missing env variables to String by @maxjeblick in #440
• Enable skipped tests by @maxjeblick in #444
• RLHF fixes for small samples size by @pascal-pfeiffer in #445
• Split Train-val data on parent id by @maxjeblick in #438
• Fix style errors by @maxjeblick in #455
• Causal Classification Problem Type by @psinger in #449
• deepspeed by @haqishen in #288
• Check config for issues by @maxjeblick in #454
• Fix differential lr plots by @maxjeblick in #457
• version = "1.0.0" by @pascal-pfeiffer in #462

Full Changelog: v0.2.1...v1.0.0

v0.2.1

What's Changed

• Fix: Load default user settings by @maxjeblick in #419
• Refactor: conversation chaining by @maxjeblick in #386
• Fix: perplexity metric in RLHF by @pascal-pfeiffer in #410
• Fix: HF push cpu_shard, soften value check by @pascal-pfeiffer in #413

Full Changelog: v0.2.0...v0.2.1

v0.2.0

What's Changed

• [DOCS] Document how to install custom packages by @sherenem in #405
• [DevOps] Update LLM packer scripts by @DinukaH2O in #406
• Auto lora layers by @psinger in #408
• add dataset import from azure by @pascal-pfeiffer in #407
• Packages and Lora fixes by @psinger in #411
• Enhancements to Secret Management by @maxjeblick in #364

Full Changelog: v0.1.1...v0.2.0

v0.1.1

What's Changed

• Contribution guide by @maxjeblick in #329
• Fix no gpu experiments by @maxjeblick in #360
• Fix markdown plotting by @maxjeblick in #371
• clean space for docker build by @pascal-pfeiffer in #375
• ignore warnings for loading config yaml by @maxjeblick in #373
• Hotfix filterwarnings by @maxjeblick in #376
• only kill parent process when failing in ddp mode by @pascal-pfeiffer in #378
• Utility function to use the CLI interface to publish models to Hugging Face by @diegomiranda02 in #300
• Fixes seq2seq Perplexity by @psinger in #385
• Update FAQ answer by @sherenem in #388
• [DevOps] Update packer scripts for policy requirements by @DinukaH2O in #391
• add integration tests for all problem types by @pascal-pfeiffer in #380
• H2O_WAVE_APP_ADDRESS=http://127.0.0.1:8756 to unblock port 8000 by @pascal-pfeiffer in #394
• heap (default off) by @pascal-pfeiffer in #389
• Readme custom packages by @psinger in #395
• v0.1.1 release version by @pascal-pfeiffer in #401

New Contributors

• @diegomiranda02 made their first contribution in #300
• @DinukaH2O made their first contribution in #391

Full Changelog: v0.1.0...v0.1.1

v0.1.0

What's Changed

• New Experiment Summary page by @fatihozturkh2o in #316
• cuda118 pytorch by @pascal-pfeiffer in #325
• HF friendly repo name by @fatihozturkh2o in #322
• Max/insights table view by @maxjeblick in #301
• embed new video into videos section for LLM studio by @shaunyogeshwaran in #332
• Update build-and-push-release.yml by @pascal-pfeiffer in #334
• Sequence to Sequence Problem Type by @psinger in #308
• upd packages by @pascal-pfeiffer in #341
• saving hf yaml separately by @fatihozturkh2o in #342
• Fixing generation code by @psinger in #348
• Remove pickle support by @maxjeblick in #351
• Fix get_size_str by @maxjeblick in #353
• add-CPU-RAM-requirements-docs by @shaunyogeshwaran in #344
• Fix tests and test workflow by @maxjeblick in #355
• Refactor rlhf by @maxjeblick in #328
• llama2 7b with int4 quantization as default by @pascal-pfeiffer in #367

New Contributors

• @shaunyogeshwaran made their first contribution in #332

Full Changelog: v0.0.6...v0.1.0

v0.0.6

What's Changed

• delete redundant code by @maxjeblick in #271
• Token IDs (again) by @psinger in #265
• end user shall use "make llmstudio" to run without wave checking for file changes by @pascal-pfeiffer in #269
• Fix neptune version by @psinger in #273
• Minor model template changes by @psinger in #274
• Minor doc improvements by @sherenem in #261
• dont compute np.mean(val_losses) twice by @Quetzalcohuatl in #275
• account correctly for day of year by @pascal-pfeiffer in #277
• Package upgrades and minor changes by @psinger in #291
• Removal of llama force slow by @psinger in #292
• Add data sanity checks by @maxjeblick in #289
• LLama2 config fix by @psinger in #294
• Set huggingface token when loading tokenizers by @maxjeblick in #296
• Use unk token for pad if available by @psinger in #298
• Refactor problem type by @maxjeblick in #293
• Add required URLs to LLM Studio installation guide by @sherenem in #310
• Show git version in error tab by @maxjeblick in #306
• Fix Runpod support / fix copy_config to use output dir config (#280) by @Glavin001 in #281

New Contributors

• @Quetzalcohuatl made their first contribution in #275

Full Changelog: v0.0.5...v0.0.6

v0.0.5

What's Changed

• export cache dir to save downloaded backbone. by @osiire in #212
• Generation config fix by @psinger in #248
• add missing cfg by @pascal-pfeiffer in #252
• Customize data_folder & output_folder with environment variables (#216) by @Glavin001 in #217
• Add missing content to the Experiments section by @oshi98 in #249
• Improve sampling for chained samples by @psinger in #254
• Fix limit_chained_samples by @maxjeblick in #256
• Hotfix to avoid deletion of datasets by @maxjeblick in #262
• System prompt by @psinger in #250
• update transformers and other dependencies by @pascal-pfeiffer in #266
• add: safe serialization while pushing to hf by @shivance in #221
• introduce "H2O_LLM_STUDIO_WORKDIR" by @pascal-pfeiffer in #268

New Contributors

• @osiire made their first contribution in #212
• @Glavin001 made their first contribution in #217
• @shivance made their first contribution in #221

Full Changelog: v0.0.4...v0.0.5

v0.0.4

What's Changed

• Default settings adjustment by @psinger in #154
• Manual loss & chained setting by @psinger in #142
• support azure endpoints for openai api by @pascal-pfeiffer in #147
• update transformers to 4.30.1 by @maxjeblick in #158
• fix for additional pretraining and more detailed FAQ by @pascal-pfeiffer in #155
• switch back to default chat behaviour by @pascal-pfeiffer in #159
• enable train progress bar logging for sub 1 epoch evals by @pascal-pfeiffer in #162
• added perplexity as metric by @pascal-pfeiffer in #157
• personalize chatbot by @pascal-pfeiffer in #161
• hotfix by @pascal-pfeiffer in #164
• fix for custom chatbot_name and chatbot_author by @pascal-pfeiffer in #165
• Add documentation about updating H2O LLM Studio by @maxjeblick in #169
• Pp/rlhf by @pascal-pfeiffer in #152
• Pp/typing by @pascal-pfeiffer in #173
• Allow to change branch for downloading huggingface models by @haqishen in #168
• Use huggingface auth token for downloading models by @maxjeblick in #167
• Dataset refactoring by @maxjeblick in #177
• fix style command by @maxjeblick in #178
• Fix auth token for falcon models by @maxjeblick in #181
• Package update by @psinger in #176
• Run llmstudio docker image as llmstudio user instead of root user by @tomkraljevic in #163
• Dtype & HF Push Changes by @psinger in #182
• Use cache changes by @psinger in #188
• Add falcon peft target modules by @maxjeblick in #166
• Def loss in dictionary by @psinger in #193
• GPT metric endpoint by @psinger in #195
• Refactoring by @maxjeblick in #192
• Make gpu id for chat configurable by @maxjeblick in #184
• Update wave by @maxjeblick in #198
• Update README.md by @psinger in #199
• Push official docs site for LLM Studio by @sherenem in #156
• Fix documentation base URL by @sherenem in #201
• Diverse adjustments by @psinger in #207
• Fix link to documentation in README by @tmm1 in #202
• rlhf batches by @pascal-pfeiffer in #197
• Add doc issue template by @sherenem in #204
• check available disk space by @haqishen in #194
• Wave streaming by @maxjeblick in #205
• Checkpoint changes by @psinger in #222
• add torch uint8 for disk check by @haqishen in #224
• Fix tokenizer length by @psinger in #231
• Fix chat bug by @maxjeblick in #230
• cast to float before converting to numpy by @pascal-pfeiffer in #234
• make lora slider max values adjustable by @pascal-pfeiffer in #236
• Bitsandbytes fixes by @psinger in #226
• version bump to 0.0.4 and default batch_size 2 by @pascal-pfeiffer in #225

New Contributors

• @haqishen made their first contribution in #168
• @tomkraljevic made their first contribution in #163
• @tmm1 made their first contribution in #202

Full Changelog: v0.0.3...v0.0.4

v0.0.3

What's Changed

• update readme to include support for installing nvidia drivers if req… by @jfarland in #114
• chat bubbles switched, user on right side by @pascal-pfeiffer in #121
• [DevOps] Fix Snyk project issue by @ChathurindaRanasinghe in #115
• FAQ by @pascal-pfeiffer in #122
• replace(".", "-") by @pascal-pfeiffer in #124
• Fix local download of model by @maxjeblick in #126
• [DevOps] Fix condition issue for snyk test & snyk monitor by @ChathurindaRanasinghe in #127
• fix bug when loading from config and remote code by @pascal-pfeiffer in #132
• Max/4bit by @maxjeblick in #131
• model card by @pascal-pfeiffer in #138
• GPT eval progress by @pascal-pfeiffer in #139
• Proper DDP Eval Progress by @psinger in #141

New Contributors

• @jfarland made their first contribution in #114

Full Changelog: v0.0.2...v0.0.3

v0.0.2

What's Changed

• Github action for docker build and push to vovran by @lakinduakash in #86
• Tokenize rework by @psinger in #90
• [DevOps] Jenkins pipeline for creating cloud images. by @ChathurindaRanasinghe in #106
• [DevOps] Snyk Integration by @ChathurindaRanasinghe in #109
• HF Push and Download improvements by @psinger in #110
• Max/chat block by @maxjeblick in #85
• update version by @maxjeblick in #113

New Contributors

• @lakinduakash made their first contribution in #86

Full Changelog: v0.0.1...v0.0.2