v0.1.8

What's Changed

• GRPO trainer: Minimal initial integration by @nikg4 in #1482
• Update oumi infer to fall back to interactive mode if no input path is specified. by @taenin in #1483
• Add sample DDP/GCP config for GRPO trainer by @nikg4 in #1485
• Temporary fix for chat template issue with multimodal inference w/ in-process vLLM engine by @nikg4 in #1486
• [tiny] Update async_eval.yaml comments to reference correct class by @wizeng23 in #1488
• Fix a bug where overriding remote_params fails via the CLI (oumi infer) by @taenin in #1487
• Define GrpoParams under configs by @nikg4 in #1490
• Support more GRPO params by @nikg4 in #1491
• Minor updates to oumi env by @nikg4 in #1492
• Warn instead of error when device not found for MFU calculation by @wizeng23 in #1489
• Updated all CLI endpoints to support oumi:// prefix by @Spaarsh in #1468
• Fix chat template issue for nested content parts used for VLMs by @nikg4 in #1493
• Ctseng777/judge by @ctseng777 in #1474
• [Evaluation] Modularization & enabling custom evaluations by @kaisopos in #1484
• Update documentation formatting for BaseModel by @taenin in #1494
• Fix log_samples not propagating from eval_kwargs by @jgreer013 in #1496
• [Evaluation] Adding support for logging model samples for all backends by @kaisopos in #1499
• Support for deprecated input param ( evaluation_platform) by @kaisopos in #1500
• Limiting the AlpacaEval number of samples for quickstart by @kaisopos in #1501
• Add recurring tests to keep our test badges updated. by @taenin in #1498
• Add a schedule for our GPU, CPU, and doc tests by @taenin in #1503
• Update the GPU Tests badge to use results from main by @taenin in #1504
• vLLM version increment by @nikg4 in #1502
• Minor logging improvements by @nikg4 in #1505
• [Evaluation] Save Utils: Moving, fixes, and unit tests by @kaisopos in #1506
• Update sample GRPO script to validate num_generations by @nikg4 in #1509
• Resolve warning about --dispatch batches deprecated param by @nikg4 in #1510
• [Evaluation] Re-enabling evaluations with Math Hard (leaderboard_math_hard) by @kaisopos in #1511
• Update docker image and build script by @oelachqar in #1508
• Add Qwen QwQ Lora config by @wizeng23 in #1514
• Add QwQ eval/infer configs by @wizeng23 in #1515
• [Evaluation] Instantiating an inference engine (if needed) when running custom evaluations by @kaisopos in #1513
• Switch eval yaml configs to use evaluation_platform by @wizeng23 in #1516
• Mark BaseMapDataset as typing.Sized by @nikg4 in #1517
• VLM collator refactor by @nikg4 in #1512

Full Changelog: v0.1.7...v0.1.8

v0.1.7

What's Changed

• Update the RemoteInferenceEngine to appropriately handle openai format batch prediction endpoints. by @taenin in #1472
• Fix local models to not break the registry. by @taenin in #1476
• Create an inference config for Claude Sonnet 3.7 by @taenin in #1479
• Add notebook for fine-tuning MiniMath-R1-1.5B by @jgreer013 in #1480
• [Evaluation] Migrate LM Harness integration point from simple_evaluate to evaluate by @kaisopos in #1455
• [tiny]Update trl to 0.14 by @wizeng23 in #1478

Full Changelog: v0.1.6...v0.1.7

v0.1.6

What's Changed

• Update RemoteParams to no longer require an API URL. by @taenin in #1452
• [Tiny] Update default training params for Qwen2-VL-2B-Instruct by @optas in #1454
• [Tiny] Add more warnings for "special" requirements of Qwen2.5-VL by @optas in #1453
• Minor cleanup of oumi fetch by @taenin in #1463
• Support for multi-image VLM training by @nikg4 in #1448
• Remove a temp workaround in pad_sequences on the left side by @nikg4 in #1464
• [tiny] Add warning that Oumi doesn't support Intel Macs by @wizeng23 in #1467
• VLM-related logging improvements by @nikg4 in #1469
• Fix Oumi launcher to be able to run on RunPod and Lambda by @wizeng23 in #1470
• Enable pre-release install for uv in pyproject.toml by @wizeng23 in #1466

Full Changelog: v0.1.5...v0.1.6

v0.1.5

What's Changed

• Fix the remainder of our configs by @wizeng23 in #1356
• Adopt new Llama 3.1 HF names by @wizeng23 in #1357
• Define OUMI_USE_SPOT_VM env var and start using it to override use_spot param by @xrdaukar in #1359
• Support HuggingFaceM4/Docmatix dataset by @vishwamartur in #1342
• [nit] update default issue names by @oelachqar in #1367
• Update sft_datasets.md by @penfever in #1349
• Have GitHub Trending image hyperlink to GitHub Trending page by @wizeng23 in #1370
• Update the link for the trending banner. by @taenin in #1371
• Move code to disable caching in model.config to a helper function by @xrdaukar in #1378
• Update transformers version to 4.48 by @wizeng23 in #1372
• Update notebooks to improve their Colab experience by @wizeng23 in #1380
• Add proper labels and types to new Bugs and Feature Requests. by @taenin in #1383
• Upgrade omegaconf to 2.4.0dev3 by @wizeng23 in #1384
• Support HuggingFaceM4/the_cauldron dataset by @vishwamartur in #1366
• Update our FAQ for tips about installing oumi on Windows by @taenin in #1385
• Cleanup HuggingFaceM4/Docmatix and HuggingFaceM4/the_cauldron multimodal datasets by @xrdaukar in #1387
• Remove uneeded env vars from job configs by @wizeng23 in #1390
• Remove transformer version override for HuggingFaceTB/SmolVLM-Instruct in launcher script by @xrdaukar in #1388
• [Small Refactor] Moving the inference engine def outside the inference config by @kaisopos in #1395
• Evaluation - LM Harness: Adding vLLM support by @kaisopos in #1379
• Remove Docmatix dataset references from docstrings VLM config examples by @xrdaukar in #1397
• Fixed broken link in Oumi - A Tour.ipynb notebook by @ciaralema in #1398
• Fix broken links in notebooks. by @taenin in #1402
• Create a client for communicating with a Slurm node via SSH. by @taenin in #1389
• [tiny] Remove references to missing job configs in README by @wizeng23 in #1404
• Train+Inference with Qwen 2.5 VL (3B) by @optas in #1396
• Add a Slurm cluster and cloud to the oumi launcher. by @taenin in #1406
• Move pretokenize script from scripts/pretokenize/ to scripts/datasets/pretokenize/ by @xrdaukar in #1412
• Create a script to save Conversation-s from SFT datasets into .jsonl file by @xrdaukar in #1413
• [Evaluation] LM Harness refactor by @kaisopos in #1410
• Update save_conversations tool by @xrdaukar in #1421
• [SambaNova] Integrate SambaNova Systems to oumi inference by @ctseng777 in #1415
• [Μinor] Equating Qwen's 2.5 chat-template to version's 2.0 by @optas in #1419
• Add requirements header to configs and clean them up by @wizeng23 in #1411
• Updated oumi infer to support CLI argument for system prompt by @Spaarsh in #1422
• [Evaluation] LM Harness remote server support by @kaisopos in #1414
• [Feature] Add Tulu3 SFT Mixture Dataset Support by @bwalshe in #1381
• Support Multimodal inference with multiple images and PDF-s in NATIVE engine by @xrdaukar in #1424
• Update notebooks to run on Colab by @wizeng23 in #1423
• Add calm recipe. by @taenin in #1425
• Update VLM sample oumi infer -i commands by @xrdaukar in #1428
• Provide example show to start SGLang server using Docker by @xrdaukar in #1429
• Multi-image support in SGLang inference engine by @xrdaukar in #1426
• Calm readme by @emrecanacikgoz in #1432
• WildChat-50M Reproduction by @penfever in #1433
• Add WildChat support by @penfever in #1348
• Create pad_to_max_dim_and_stack() function in torch_utils by @xrdaukar in #1435
• use deterministic by @penfever in #1434
• Additional HF trainer parameters for config by @penfever in #1436
• Set a better default for vllm inference GPU usage. by @taenin in #1437
• Added fetch command and modified infer command to resolve oumi:// by @Spaarsh in #1439
• Require an inference config for oumi infer. by @taenin in #1443
• Make the tulu3 unit tests hermetic. by @taenin in #1446
• Add 2 more sample PDF-s with 1 and 2 pages under testdata/pdfs by @xrdaukar in #1427
• Enable ability to override list values in config via CLI by @wizeng23 in #1430
• Renamed CALM to CoALM by @jgreer013 in #1450
• Add support for Docmatix dataset to multimodal training script by @xrdaukar in #1449
• Update oumi launch status to show clusters with no running jobs. by @taenin in #1451

New Contributors

• @vishwamartur made their first contribution in #1342
• @ctseng777 made their first contribution in #1415
• @Spaarsh made their first contribution in #1422
• @bwalshe made their first contribution in #1381
• @emrecanacikgoz made their first contribution in #1432

Full Changelog: v0.1.4...v0.1.5

v0.1.4

What's Changed

• Add memory cleanup calls in e2e integration tests by @xrdaukar in #1277
• Set up versioning for our documentation by @taenin in #1275
• Make qwen2-VL evaluation job pass by @xrdaukar in #1278
• Add multi-modal (vlm) notebook with Llama 11B by @optas in #1258
• Documentation: Inference -> List supported models by @kaisopos in #1279
• [tiny] update website link by @oelachqar in #1280
• Update all documentation links to the new doc URL by @taenin in #1281
• Update Oumi - A Tour.ipynb by @brragorn in #1282
• Documentation: Judge (minor edits) by @kaisopos in #1283
• Fix citation by @oelachqar in #1285
• Add Deepseek R1 1.5B/32B configs by @wizeng23 in #1276
• Misc eval configs cleanup by @xrdaukar in #1286
• [docs] Describe parallel evaluation by @xrdaukar in #1284
• Update microsoft/Phi-3-vision-128k-instruct training config by @xrdaukar in #1287
• Add Together Deepseek R1 inference config by @wizeng23 in #1289
• [minor] vlm notebook minor updates (doc referencing, freeze visual backbone) by @optas in #1288
• Add missing -m oumi evaluate argument in eval config by @xrdaukar in #1291
• [docs] Add more references to VL-SFT and SFT notebooks by @xrdaukar in #1293
• Eval config change for deepseek-ai/DeepSeek-R1-Distill-Llama-70B by @xrdaukar in #1292
• [notebooks] Update intro & installation instruction by @oelachqar in #1294
• Update notebook intros by @oelachqar in #1296
• [notebooks] Update installation instructions for colab by @oelachqar in #1297
• Add Apache license header to src/oumi/**/*.py by @wizeng23 in #1290
• Minor updates to VLM Multimodal notebook by @xrdaukar in #1299
• [docs] Add latest notebooks and update references by @oelachqar in #1300
• [tiny] Add docs auto-generated .rst files to gitignore by @wizeng23 in #1298
• [tiny] use GitHub link for header by @oelachqar in #1301
• [docs][tiny] update inference engines reference by @oelachqar in #1302
• Update README/docs to add new DeepSeek models by @wizeng23 in #1304
• [docs] Use pip install oumi over pip install . by @wizeng23 in #1305
• Tune VLM SFT configs by @xrdaukar in #1306
• Tune VLM configs for SmolVLM and Qwen2-VL by @xrdaukar in #1307
• Update config/notebook pip installs to use PyPI by @wizeng23 in #1308
• [tiny] upgrade torch version by @oelachqar in #1295
• Update logging and unit tests related to chat templates by @xrdaukar in #1311
• fix(docs): "interested by joining" to "interested in joining" by @CharlesCNorton in #1312
• Add HF_TOKEN instructions to Oumi Multimodal notebook by @xrdaukar in #1313
• Update configuration.md by @penfever in #1314
• remove duplicate keys in config example by @lucyknada in #1315
• [Notebooks] Update VLM notebook by @xrdaukar in #1317
• Update parasail_inference_engine.py by @jgreer013 in #1320
• Fix typo and update warning message for OUMI trainer by @xrdaukar in #1319
• [Notebooks] Add a note that a notebook kernel restart may be needed after pip install oumi by @xrdaukar in #1318
• Update Phi3 to support multiple images by @xrdaukar in #1321
• Add more detailed comment headers to YAML configs by @wizeng23 in #1310
• [Notebooks] Add a note to Tour notebook to restart kernel after the first pip install by @xrdaukar in #1327
• Tweak --mem-fraction-static param in sample SGLang configs by @xrdaukar in #1328
• Disallow using DatasetParams field names as keys in DatasetParams.dataset_kwargs by @xrdaukar in #1324
• Support dataset_name_override dataset_kwarg by @xrdaukar in #1188
• Add an util and a test marker for HF token by @xrdaukar in #1329
• Update llama3-instruct chat template to align with the original models template by @xrdaukar in #1326
• chore: update launcher.sh by @eltociear in #1333
• [Notebooks] Minor improvements in VLM and CNN notebooks by @xrdaukar in #1335
• Update VLM cluster names in sample commands by @xrdaukar in #1336
• Update our README and docs with the github trending badge. by @taenin in #1340
• Update README.md - Add DeepSeek to supported models by @mkoukoumidis in #1343
• Update index.md - Add DeepSeek to supported models by @mkoukoumidis in #1344
• Update "GPU Tests" status badge in README page by @xrdaukar in #1345

New Contributors

• @CharlesCNorton made their first contribution in #1312
• @lucyknada made their first contribution in #1315
• @eltociear made their first contribution in #1333

Full Changelog: v0.1.3...v0.1.4

v0.1.3

What's Changed

• Documentation: Judge | Custom Model page by @kaisopos in #1195
• [WIP] Add a notebook for using CNN with custom dataset by @xrdaukar in #1196
• [Cherrypick for launch] Evaluate: return dict of results by @kaisopos in #1197
• Configs Train/Infer/Eval and Llama 3.3v (70b) by @optas in #1200
• Adding an integration test for evaluation fn's output (see PR-1197) by @kaisopos in #1199
• [docs] Add more details and cross-references related to customization by @xrdaukar in #1198
• Define single_gpu test marker by @xrdaukar in #1201
• Native inference: Don't set min_p, temperature in GenerationConfig if sampling is disabled by @xrdaukar in #1202
• Update tests to make them runnable on GCP by @xrdaukar in #1203
• Add newline before pformat(train_config) by @xrdaukar in #1204
• GCP tests launcher script changes by @xrdaukar in #1205
• [Evaluation] Bug: serialization by @kaisopos in #1207
• [docs] Add inference snippet for together.ai and DeepSeek APIs by @oelachqar in #1208
• Exclude multi_gpu tests from GitHub GPU tests by @xrdaukar in #1210
• Update e2e tests to support multi-GPU machines by @xrdaukar in #1206
• Add wrappers for remote inference engines by @oelachqar in #1209
• Vision-Lang & Inference (including LoRA) by @optas in #1174
• [BugFix] Throw a runtime error for quantized models & inference=VLLM by @kaisopos in #1212
• Fix most job configs by @wizeng23 in #1213
• e2e tests update by @xrdaukar in #1216
• [Notebook] Evaluation with Oumi by @kaisopos in #1218
• gpt2: move include_performance_metrics param from script to yaml by @xrdaukar in #1217
• Simplify inference engine API by @oelachqar in #1214
• Move configs to experimental by @wizeng23 in #1215
• [docs] Update index page by @oelachqar in #1220
• Update ConsoleLogger to write to STDOUT by @xrdaukar in #1221
• Set use_spot to False in our JobConfigs by @wizeng23 in #1222
• Delete oumi[optional] install target by @wizeng23 in #1224
• Scaffolding and the first testcase for e2e evaluation tests by @xrdaukar in #1225
• [docs] Update inference engines doc page by @oelachqar in #1227
• Clean-up inference engine builder by @oelachqar in #1226
• [VLLM Engine] Enabling BitsAndBytes quantization by @kaisopos in #1223
• Add example distillation notebook by @jgreer013 in #1228
• Add a script to pre-download models for gpu_tests by @xrdaukar in #1231
• Fix multi-GPU inference integration test by @xrdaukar in #1229
• [tiny][docs] Update PEFT/LoRA content by @optas in #1233
• [BugFix] GGUF does not work with VLLM by @kaisopos in #1232
• Re-enable parallel evaluation for VLM-s by @xrdaukar in #1235
• Add multimodal exemplar dataset in our provided mini-datasets by @optas in #1234
• [Tiny] renaming a field name (init_lora_weights) by @optas in #1236
• Add more e2e evaluation tests by @xrdaukar in #1237
• Fix pyright breakage when vllm and llama_cpp are not installed by @taenin in #1240
• Update our oumi launch documentation. by @taenin in #1239
• Update index.md title for "Join the Community!" by @mkoukoumidis in #1242
• Update quickstart.md - nit for Oumi support request by @mkoukoumidis in #1241
• [VLLM Engine] Improve support for GGUF models (incl. auto-download) by @kaisopos in #1238
• Update README.md title to "Join the Community!" by @mkoukoumidis in #1243
• Update quickstart.md by @brragorn in #1251
• Update quickstart.md by @brragorn in #1253
• Update quickstart.md by @brragorn in #1252
• Update quickstart.md by @brragorn in #1250
• [Minor refactor] Moving model caching to oumi.utils by @kaisopos in #1246
• Add more details to troubleshooting FAQ by @wizeng23 in #1249
• Update training_methods.md - Change compute requirement suggestions by @mkoukoumidis in #1245
• Update train.md - nit description change by @mkoukoumidis in #1244
• [docs] misc docs feedback by @oelachqar in #1248
• [tiny] Qwen2-VL activate experimental datapipes by @optas in #1247
• Update Oumi - A Tour.ipynb by @brragorn in #1254
• [docs] more docs feedback by @oelachqar in #1255
• Update supported_models.md by @penfever in #1256
• Rename experimental_use_torch_datapipes data param by @xrdaukar in #1257
• Add pypi release workflow using testpypi by @oelachqar in #1259
• Update workflow names by @oelachqar in #1262
• Update default idle_minutes_to_autostop to 1 hour. by @taenin in #1264
• update pypi release workflow to use trusted env by @oelachqar in #1265
• Add padding_side param to internal model config by @xrdaukar in #1260
• Documentation: Updates on Evaluation/Judge (based on Manos' feedback) by @kaisopos in #1261
• [tiny] less strict requirements by @oelachqar in #1266
• Add Deepseek R1 Distill Llama 8B/70B configs by @wizeng23 in #1263
• Update index.md to highlight beta stage by @mkoukoumidis in #1268
• Update README.md to highlight beta stage by @mkoukoumidis in #1267
• Disable pre-release packages by @oelachqar in #1270
• Update common_workflows.md - Clarify OpenAI is just an example by @mkoukoumidis in #1271
• Documentation: Evaluation page (update to highlight multi-modal) by @kaisopos in #1269
• Update launch.md by @taenin in #1272
• Add pypi release workflow by @oelachqar in #1273
• Documentation: Judge | minor edit (bold) by @kaisopos in #1274

Full Changelog: v0.1.2...v0.1.3

v0.1.2.3

What's Changed

• Re-enable parallel evaluation for VLM-s by @xrdaukar in #1235
• Add multimodal exemplar dataset in our provided mini-datasets by @optas in #1234
• [Tiny] renaming a field name (init_lora_weights) by @optas in #1236
• Add more e2e evaluation tests by @xrdaukar in #1237
• Fix pyright breakage when vllm and llama_cpp are not installed by @taenin in #1240
• Update our oumi launch documentation. by @taenin in #1239
• Update index.md title for "Join the Community!" by @mkoukoumidis in #1242
• Update quickstart.md - nit for Oumi support request by @mkoukoumidis in #1241
• [VLLM Engine] Improve support for GGUF models (incl. auto-download) by @kaisopos in #1238
• Update README.md title to "Join the Community!" by @mkoukoumidis in #1243
• Update quickstart.md by @brragorn in #1251
• Update quickstart.md by @brragorn in #1253
• Update quickstart.md by @brragorn in #1252
• Update quickstart.md by @brragorn in #1250
• [Minor refactor] Moving model caching to oumi.utils by @kaisopos in #1246
• Add more details to troubleshooting FAQ by @wizeng23 in #1249
• Update training_methods.md - Change compute requirement suggestions by @mkoukoumidis in #1245
• Update train.md - nit description change by @mkoukoumidis in #1244
• [docs] misc docs feedback by @oelachqar in #1248
• [tiny] Qwen2-VL activate experimental datapipes by @optas in #1247
• Update Oumi - A Tour.ipynb by @brragorn in #1254
• [docs] more docs feedback by @oelachqar in #1255
• Update supported_models.md by @penfever in #1256
• Rename experimental_use_torch_datapipes data param by @xrdaukar in #1257
• Add pypi release workflow using testpypi by @oelachqar in #1259
• Update workflow names by @oelachqar in #1262
• Update default idle_minutes_to_autostop to 1 hour. by @taenin in #1264
• update pypi release workflow to use trusted env by @oelachqar in #1265

Full Changelog: v0.1.2.2...v0.1.2.3

v0.1.2.0-alpha

What's Changed

• Update README.md - Better highlight features & nits by @mkoukoumidis in #995
• [tiny] update docstring and cleanup by @oelachqar in #1006
• Qwen2-VL: minor updates by @xrdaukar in #1000
• Update README.md - Describe Oumi's most common capabilities by @mkoukoumidis in #996
• Fix readme. by @taenin in #1009
• Updated our ascii logo by @taenin in #1008
• [docs] Update readme by @oelachqar in #1010
• Cleanup scripts by @oelachqar in #1011
• Cleanup experimental folder by @oelachqar in #1012
• Update lists of supported VLM-s in README and docs by @xrdaukar in #1014
• Freeze Python package versions by @xrdaukar in #1007
• Update blip2's chat template to use the "default" one by @xrdaukar in #1015
• Add docstrings how to start vLLM and SGLang servers for Llama-3.2-11B-Vision-Instruct by @xrdaukar in #1016
• Evaluation: bugfixing, corner case, unit tests by @kaisopos in #1003
• Configure asyncio_default_fixture_loop_scope to reduce pytest warnings by @xrdaukar in #1013
• Update the registry to load registered core values upon use. by @taenin in #1017
• Update default installation instructions to pypi by @taenin in #1018
• [tiny] Update debug datasets by @oelachqar in #1020
• [docs] Address misc docs feedback by @oelachqar in #1019
• [tiny] update pre-defined judges and docs by @oelachqar in #1021
• Parameterize e2e training test, and add config for Qwen2-VL by @xrdaukar in #1023
• Remove our docs password from the readme. by @taenin in #1024
• VLM docs update by @xrdaukar in #1025
• Fix loading registered pretrain datasets by @wizeng23 in #1005
• Update @requires_gpus test decorator to optionally specify min GPU memory requirement by @xrdaukar in #1029
• [tiny] Update GitHub workflows by @oelachqar in #1034
• Update BaseConfig.from_yaml to also support Path by @xrdaukar in #1026
• [tiny] Cleanup judge engine builder & fix circular dep by @oelachqar in #1035
• Create GPU GitHub Actions workflow by @oelachqar in #1004
• Add structured outputs support to gemini/vertex engines by @oelachqar in #1022
• [docs] Fix feedback on training and inference user guides by @oelachqar in #1037
• [docs][tiny] fix examples in inference guide by @oelachqar in #1038
• Add a sanity test for circular imports. by @taenin in #1030
• Resolve circular dependencies in Oumi by @taenin in #1039
• Move our circular dependency test to e2e to speed up GPU CI tests. by @taenin in #1040
• Add custom inference engine for gemini API by @oelachqar in #1036
• Define CLI in our quickstart. by @taenin in #1042
• Skip running GPU tests on low-risk code paths by @oelachqar in #1043
• Define more terms in our training docs. by @taenin in #1044
• Fix the broken python text snippet on the train page. by @taenin in #1045
• Fix the second python snippet in the train page. by @taenin in #1046
• [docs] Add Gemini to the list of supported inference API-s, and sort them by @xrdaukar in #1048
• Fix issues in most notebooks by @wizeng23 in #1047
• [docs][tiny] remove termynal from sphinx conf by @oelachqar in #1041
• Fix a typo in the VS Code environment page. by @taenin in #1049
• Define WSL in our vscode docs. by @taenin in #1052
• [tiny] disable unit tests on safe paths by @oelachqar in #1051
• [docs] Fix contributing and open issue links by @oelachqar in #1050
• [evaluations/generative_benchmark] Broken link by @kaisopos in #1054
• Remove dangling reference to jupyter in Makefile help by @xrdaukar in #1053
• [evaluations/generative_benchmark] Removing notebook link by @kaisopos in #1055
• Support constrained decoding in SGLang inference engine by @xrdaukar in #1032
• [tiny] Update tutorials page by @wizeng23 in #1056
• Minor updates to Launch.md by @taenin in #1059
• [docs] Update docs/user_guides/infer/infer.md by @xrdaukar in #1058
• Nits for common_workflows.md by @mkoukoumidis in #1061
• Nit fixes for acknowledgements.md by @mkoukoumidis in #1057
• Add sample trouble shooting for remote jobs. by @taenin in #1062
• Add a Github Issues selector for questions and have it redirect to Discord. by @taenin in #1064
• Package checking: Adding functionality for checking package versioning and fast failing by @kaisopos in #1031
• Fix various typos in contributing.md by @taenin in #1066
• SGLang inference documentation by @xrdaukar in #1065
• Replace assert in NativeInferenceEngine with RuntimeError by @xrdaukar in #1068
• Update dev set up instructions to use a Fork. by @taenin in #1067
• Define inference configs for more models by @xrdaukar in #1069
• [Evaluation] HF Leaderboards yaml files by @kaisopos in #1071
• Specify engine: NATIVE is inference configs by @xrdaukar in #1075
• Improve handling of image path and URLs by @xrdaukar in #1074
• [Doc > Quickstart] Should we add links to guides for better discoverability? by @kaisopos in #1076
• Add e2e tests for running tutorial notebooks by @oelachqar in #1079
• Ignore all experimental files when running our circular dependency test. by @taenin in #1081
• [Super Nit Doc Update] environments.md by @kaisopos in #1082
• Add an env var for loading user registered values (dataset, models, clouds) when initializing the Oumi Registry by @taenin in #1077
• Update internal model configs to support default tokenizer_pad_token and chat_template by model type by @xrdaukar in #1078
• [Minor] Notebook typo by @kaisopos in #1085
• Upgrade transformers to 4.47 by @wizeng23 in #1033
• [tiny][docs] Update recipes page by @wizeng23 in #1072
• Configure e2e integration test for Llama 3.2 Vision 11B by @xrdaukar in #1086
• Nits for cli_reference.md by @mkoukoumidis in #1063
• [Documentation] Evaluate | Leaderboards Page by @kaisopos in #1084
• [Documentation] Evaluate | Main Page (revision) by @kaisopos in #1089
• [tiny] Fix precommit by @oelachqar in #1092
• Add timeout for unit & integration tests by @oelachqar in #1091
• Add GitHub Actions workflow for doctests by @oelachqar in #1093
• [docs] remove unused page, fix links by @oelachqar in #1094
• [Documentation] Evaluate | Main Page (small refactor) by @kaisopos in #1095
• Rewrite of the main Oumi Launch page. by @taenin in #1087
• Remove pytest.mark.skip() for basic e2e tests by @xrdaukar in #1088
• [tiny] Upgrade minimum numpy version to unblock python3.12 installation by @oelachqar in #1099
• Update our Readme with a new header image. by @taenin in #1098
• [docs] Minor refresh to dataset resource pages by @oelachqar in #1097
• [docs] Add docs guide page by @oelachqar in #1096
• Add a quick unit test to ensure new dependencies are not added to the top-level CLI by @taenin in https://github.com/o...

v0.1.1.0-alpha.1

What's Changed

• Minimal SkyPilot config for blip2 and llava models for GCP with TRL_SFT by @xrdaukar in #573
• Inference Engine async writes by @taenin in #574
• Misc cleanups in JsonlinesDataset by @xrdaukar in #576
• Split out cloud dependencies by @taenin in #575
• Disable sdpa for blip2 by @xrdaukar in #579
• Set accelerate version to fix FSDP model saving by @wizeng23 in #580
• Remove AWS as a default dep by @taenin in #582
• Update ProfilerParams docstrings to follow the new style by @xrdaukar in #583
• Minor updates in scripts/benchmarks/minimal_multimodal_training.py by @xrdaukar in #585
• Add @override annotations to methods of few Dataset subclasses by @xrdaukar in #584
• Add dataset class for dolly dataset by @oelachqar in #586
• Refactor debugging/device utils, and add new GPU stats measurement functions by @xrdaukar in #587
• Add text jsonlines dataset class by @oelachqar in #589
• Define DataCollationParams by @xrdaukar in #581
• Misc updates to Polaris launcher scripts by @xrdaukar in #591
• Set up a new version of the Oumi CLI using Typer by @taenin in #588
• Update handling of GPU fan speed info by @xrdaukar in #595
• Add support for magpie dataset variants by @oelachqar in #594
• Rename GenerationConfig to GenerationParams by @wizeng23 in #592
• Fix cli infer test by @wizeng23 in #598
• Judge Notebook 1: default judge by @kaisopos in #593
• [Tiny] update missing dataset import by @oelachqar in #599
• Update training script to support data collators by @xrdaukar in #590
• Update accelerate version to 1.0.0 by @wizeng23 in #601
• Remove deprecated dataset code paths by @oelachqar in #596
• Refactor Aya & Ultrachat to use oumi dataset sft classes by @oelachqar in #597
• Add Llama train/eval/infer E2E integration test by @wizeng23 in #602
• Set docstring for NVidiaGpuRuntimeInfo struct by @xrdaukar in #603
• Add generation params to inference engines by @oelachqar in #600
• [bug] Fix issue loading jsonl datasets from file by @oelachqar in #604
• Add Llama 3B configs by @wizeng23 in #605
• Align pyright checks with latest Pylance version by @oelachqar in #611
• Fix apply_chat_template issue in VisionLanguageSftDataset by @xrdaukar in #609
• More robust make setup by @oelachqar in #610
• Fix a bug where the new CLI was improperly importing functions from top-level modules. by @taenin in #613
• Add support for the Launch command suite in the new CLI by @taenin in #612
• Support HuggingFaceH4/llava-instruct-mix-vsft dataset by @xrdaukar in #608
• [tiny] Fix .gitignore by @wizeng23 in #616
• [tiny] add gpt2 chat template, and update tests to use it by @oelachqar in #617
• Turn off pretty-printing exceptions in our CLI by @taenin in #618
• Cleanup dependencies by @oelachqar in #615
• Upgrade oumi dependencies by @oelachqar in #606
• Update makefile to use uv, add Jupyter target by @oelachqar in #614
• Add miniconda installation target, cleanup unused make commands by @oelachqar in #620
• Update several notebooks with the new EvaluationConfig format. by @taenin in #621
• Make sure conda env is registered by @oelachqar in #622
• Add Llama 3b sft/lora/qlora configs for Polaris by @wizeng23 in #626
• Add check if installation is successful by @oelachqar in #625
• Initial Cambrian integration by @xrdaukar in #557
• [tiny] alpaca - minor reproducibility boost by @optas in #619
• explicitly specify the model's dtype in LMH by @optas in #607
• [tiny] Add flops for T4 GPU by @wizeng23 in #628
• Use a timestamp for job directories on Polaris by @taenin in #627
• [tiny] Fix bug with Polaris job num by @wizeng23 in #629
• Update two VLLM configs. by @xrdaukar in #624
• Add pip install -U uv; to make setup for existing envs by @xrdaukar in #630
• Disable MFU logging for non-packed datasets by @wizeng23 in #632
• Add config example for long context fine-tuning by @oelachqar in #631
• Add distribution mode flag to llama_tune by @wizeng23 in #635
• Judge Notebook 2: Custom Judge by @kaisopos in #623
• Bugfixes for LLAVA by @xrdaukar in #634
• Update sphinx config and docs to fix misc errors and warnings by @oelachqar in #639
• Factor out OUMI_TOTAL_NUM_GPUS env var by @wizeng23 in #636
• Remove bitsandbytes from train dependencies by @oelachqar in #643
• Enable intershinx to allow linking to external documentation pages by @oelachqar in #640
• Tune few training params for LLAVA and blip2 models by @xrdaukar in #642
• Added support for specifying the inference engine via the InferenceConfig by @taenin in #638
• Add popular pre-training dataset classes by @oelachqar in #641
• Remove openai dependency by @oelachqar in #644
• Update our documentation to point to the new CLI. by @taenin in #645
• Enable dataloaders for VLLM-s (llava and blip2) by @xrdaukar in #646
• Allow gradient clipping to be optional by @optas in #649
• Add support for add_generation_prompt in LLAVA chat template by @xrdaukar in #648
• Add a description to the Launch CLI by @taenin in #651
• Add all Llama FSDP GCP configs by @wizeng23 in #637
• Coerce model params to correct dtype for QLoRA FSDP by @wizeng23 in #652
• Use uv for pip install commands by @wizeng23 in #653
• Update sphinx docs by @oelachqar in #654
• [Docs] Refactor docs pipeline by @oelachqar in #655
• [docs] swap and configure sphinx theme by @oelachqar in #656
• [Docs] Add documentation placeholders by @oelachqar in #658
• [Docs] Add sphinx-bibtex by @oelachqar in #659
• [Docs] fix rendering issues by @oelachqar in #660
• [docs] fix broken links by @oelachqar in #661
• Fix broken link in readme (dev_setup) by @kaisopos in #662
• [docs][tiny] fix minor doc typos by @oelachqar in #666
• [docs] add autodoc2 template by @oelachqar in #665
• [docs] Add content links and references by @oelachqar in #668
• [docs] switch to myst-nb for rendering notebooks by @oelachqar in #669
• [docs] Add script to generate module summaries by @oelachqar in #670
• [docs] Include cli reference by @oelachqar in #671
• Add dataset submodules by @oelachqar in #667
• Update notebooks to include a descriptive title by @oelachqar in #664
• Update tests/utils/test_device_utils.py by @xrdaukar in #672
• [Inference] Bug in generation config stop tokens by @kaisopos in #663
• Support rewriting special label values to -100 (ignore_index) to exclude from loss by @xrdaukar in #657
• Rename emails and website url to Oumi by @wizeng23 in #675
• Update scri...

Initial release

What's Changed

• Add python project configs by @oelachqar in #1
• Add repo skeleton by @oelachqar in #2
• Export lema entrypoint scripts by @oelachqar in #3
• Update static type checking config by @oelachqar in #5
• Add example jupyter / colab notebook by @oelachqar in #4
• Refactor config parsing to use omegaconf by @oelachqar in #6
• Updating documentation (Dev Environment Setup) by @kaisopos in #7
• Add tests and vscode config by @oelachqar in #8
• Added DPOTrainer example to repo, as well as cuda device cleanup to training loop by @jgreer013 in #9
• Adding torch as top-level module dependency by @optas in #10
• Add configs for specific hardware requirements by @jgreer013 in #11
• Sort pre-commit hooks lexicographically by @xrdaukar in #12
• Add logging config by @oelachqar in #13
• Lema inference by @xrdaukar in #14
• Panos dev by @optas in #16
• Add job launcher by @oelachqar in #15
• Making split of data a flexible variable by @optas in #17
• Configure max file size in precommit hooks by @xrdaukar in #18
• Minor bugfix and documentation update by @oelachqar in #19
• adding pynvml to train env by @kaisopos in #20
• Panos dev by @optas in #22
• Augmenting Types for training hyperparams by @optas in #23
• Train refactoring (config file visibility) + a few minor changes by @kaisopos in #21
• Minimal test for train function by @xrdaukar in #25
• Fix leftover '_torch_dtype' in 'ModelParams' by @xrdaukar in #26
• Update GPU types list in the default SkyPilot config by @xrdaukar in #27
• Add a missing lema-infer command under [project.scripts] by @xrdaukar in #28
• add basic pytests for evaluate and infer by @xrdaukar in #29
• Update README and pyproject.toml by @wizeng23 in #30
• A helper function to print info about available CUDA devices by @xrdaukar in #31
• Update SkyPilot cconfig to start using torchrun by @xrdaukar in #32
• Support basic single-node, multi-gpu training by @xrdaukar in #33
• Run all precommit hooks on the repo by @xrdaukar in #35
• Add experimental code for llama cpp inference by @jgreer013 in #37
• Create skeleton of STYLE_GUIDE.md by @xrdaukar in #36
• Adding support for training custom models (for now just a dummy model). by @kaisopos in #38
• Fix custom model name in test_train.py by @xrdaukar in #39
• Configure pyright (static type checker) and resolve existing type errors to make it pass by @xrdaukar in #41
• fix trailing whitespace warning in STYLE_GUIDE.md by @xrdaukar in #43
• Configure initial GitHub Actions workflow to run pre-commits and tests by @xrdaukar in #44
• A variety of proposed extensions to finetune a chat-based model (starting with Zephyr) by @optas in #34
• Fix syntax error in ultrachat by @xrdaukar in #48
• Create initial version of CONTRIBUTING.md by @xrdaukar in #46
• Reduce the number of training steps from 5 to 3 to make test_train.py faster by @xrdaukar in #49
• Adding registry for custom models. by @kaisopos in #42
• Add config and streaming args to DataParams by @wizeng23 in #47
• Update Pre-review Tests to only run on pull_request by @xrdaukar in #50
• Add training flags to computes tokens-based stats by @xrdaukar in #51
• reduce test training steps in another test which I missed before by @xrdaukar in #53
• Rename var names of *Params classes by @wizeng23 in #52
• Make some NVIDIA-specific dependencies optional by @xrdaukar in #54
• fix trl version as 0.8.6 by @xrdaukar in #56
• Remove reference to torch.cuda.clock_rate by @xrdaukar in #57
• Update inference to support non-interactive batch mode. by @kaisopos in #58
• Update README.md to include Linux/WSL specific instructions by @xrdaukar in #59
• Minor formatting improvements in README.md by @xrdaukar in #60
• Minor: Updating Lora Params by @optas in #55
• Support dataset packing by @wizeng23 in #63
• Disallow relative imports in LeMa by @xrdaukar in #65
• Add text_col param that's required for SFTTrainer by @wizeng23 in #66
• Refactor common config parsing logic (YAML, arg_list) into a common util by @xrdaukar in #68
• Standardize test naming convention by @wizeng23 in #69
• Adding support for a hardcoded evaluation with MMLU. by @kaisopos in #67
• Minor changes to the default configs/skypilot/sky.yaml config by @xrdaukar in #71
• Prototype to pass config.model.model_max_length to Trainers by @xrdaukar in #70
• [Inference] Remove the prepended prompts from model responses. by @kaisopos in #73
• Add a util to print versioning info by @xrdaukar in #74
• Switch to tempfile.TemporaryDirectory() in test_train.py by @xrdaukar in #75
• Update docstring verbs to descriptive form by @wizeng23 in #76
• Add sample accelerate and fsdp configs by @xrdaukar in #77
• Refactor code to get device rank and world size into a helper function by @xrdaukar in #79
• Add a simple util to print model summary e.g., layer names, architecture summary by @xrdaukar in #80
• Freeze numpy to pre 2.0 version by @xrdaukar in #81
• Adding inference support for next logit probability. by @kaisopos in #78
• Create FSDP configs for Phi3 by @xrdaukar in #82
• Auto-format pyproject.toml with "Even Better TOML" by @xrdaukar in #83
• Minor cleanup updates to SkyPilot configs by @xrdaukar in #84
• Mixed Precision Training, Flash-Attention-2, Print-trainable-params by @optas in #85
• Update README.md to include basic instructions for multi-GPU training (DDP, FSDP) by @xrdaukar in #86
• Start using $SKYPILOT_NUM_GPUS_PER_NODE in SkyPilot config by @xrdaukar in #90
• Add configs for FineWeb Llama2 pretraining by @wizeng23 in #89
• Quantization by @optas in #87
• Update the default SkyPilot config to print more debug/context info by @xrdaukar in #92
• Add license by @oelachqar in #93
• Initial version of SkyPilot config for multi-node training (num_nodes: N) by @xrdaukar in #94
• MMLU eval refactor. by @kaisopos in #88
• Remove comparison between LOCAL_RANK and RANK by @xrdaukar in #96
• Handling the loading of peft adapters and other minor issues (e.g., adding more logging parameters) by @optas in #91
• Update configs/skypilot/sky_llama2b.yaml to start using sky_init.sh by @xrdaukar in #97
• Add bool param to resume training from the last known checkpoint (if exists) by @xrdaukar in #99
• Inference: save/restore probabilities to/from file. by @kaisopos in #98
• Add support for dataset mixtures during training by @taenin in #95
• Add train, test, and validation splits to the LeMa config. by @taenin in #101
• nanoGPT (GPT2) pretraining recipe by @wizeng23 in #103
• Minor: Updates on Zephyr-Config by @optas in https://githu...