v0.26.1

What's Changed

• fix: properly gate usage of torch.cuda.device_count (backport #3358) by @mergify in #3359

Full Changelog: v0.26.0...v0.26.1

v0.26.0

What's Changed

• feat: add model family in cfg by @jaideepr97 in #3255
• pick up model id defaults from config yaml by @jaideepr97 in #3286
• ci: temporarily disable small job on PRs by @jaideepr97 in #3290
• ci: add new Llama e2e large sized job by @jaideepr97 in #3288
• provides link to CUDA install via Nvidia: fixes issue 3144 by @JoeAldinger in #3272
• chore: remove transformers cap and raise training floor by @jaideepr97 in #3264
• feat(deps): Bump torch upper bound (<2.7.0) + set allowable range for vllm versions (>=0.8.0,<0.9.0) by @courtneypacheco in #3283
• chore: Remove python 3.10 support by @booxter in #3292
• feat: Handle lists of pydantic models when dumping to yaml by @booxter in #3295
• build(deps): bump step-security/harden-runner from 2.11.0 to 2.11.1 by @dependabot in #3251
• ci: Add e2e-nvidia-l40s-x4 job variant for python 3.12 by @booxter in #3296
• chore: Uncap requires-python in package metadata by @booxter in #3293
• fix: edit issue by @reidliu41 in #3268
• fix: update GPU acceleration guide by @emmanuel-ferdman in #3305
• fix: Adding explanation of CLI behavior and expanding the error message by @jpodivin in #3306
• fix: respect global teacher id for generation by @jaideepr97 in #3309
• build(deps): bump step-security/harden-runner from 2.11.1 to 2.12.0 by @dependabot in #3308
• Fix floating ip add for ibm cloud by @danmcp in #3310
• ci: add verbose logging to PyPI publishing by @nathan-weinberg in #3314
• build(deps): bump actions/setup-python from 5.5.0 to 5.6.0 by @dependabot in #3316
• build(deps): bump actions/download-artifact from 4.2.1 to 4.3.0 by @dependabot in #3315
• use constraints-dev.txt in e2e tests by @ktdreyer in #3320
• fix: Removing torchscript from autoconfig by @jpodivin in #3307
• mypy: Enable pydantic plugin by @booxter in #3012
• build(deps): bump docker/build-push-action from 6.15.0 to 6.16.0 by @dependabot in #3317
• Preserve storage after e2e tests by @danmcp in #3303
• feat: Add fallback logic to the small, medium, and large E2E tests to select a new AZ when AWS has insufficient capacity by @courtneypacheco in #3161
• fix: Ignore unraisable exceptions during unitcov testing by @bbrowning in #3326
• test: enable serving output in e2e by @cdoern in #3327
• ci: Always collect df -h stats regardless of failure by @booxter in #3325
• chore(deps): Bump minimum instructlab-training version by @courtneypacheco in #3312
• cap upper versions for instructlab 0.26 release series by @ktdreyer in #3335

New Contributors

• @emmanuel-ferdman made their first contribution in #3305

Full Changelog: v0.25.1...v0.26.0

What's Changed

• feat: add model family in cfg by @jaideepr97 in #3255
• pick up model id defaults from config yaml by @jaideepr97 in #3286
• ci: temporarily disable small job on PRs by @jaideepr97 in #3290
• ci: add new Llama e2e large sized job by @jaideepr97 in #3288
• provides link to CUDA install via Nvidia: fixes issue 3144 by @JoeAldinger in #3272
• chore: remove transformers cap and raise training floor by @jaideepr97 in #3264
• feat(deps): Bump torch upper bound (<2.7.0) + set allowable range for vllm versions (>=0.8.0,<0.9.0) by @courtneypacheco in #3283

Full Changelog: v0.25.0...v0.26.0a1

What's Changed

• fix: Logger can't format exception string by @courtneypacheco in #3135
• feat: Add new in-house GitHub action to detect secrets exposure by @courtneypacheco in #3112
• build(deps): bump aws-actions/configure-aws-credentials from 4.0.3 to 4.1.0 by @dependabot in #3138
• ci: disable spell check by @booxter in #3133
• feat: add --state option to filter process list output by @reidliu41 in #3125
• add a sanity check for tensor-parallel-size by @reidliu41 in #2927
• build(deps): bump step-security/harden-runner from 2.10.4 to 2.11.0 by @dependabot in #3158
• Pause the e2e-nvidia-l40s-x8.yml workflow by @kami619 in #3166
• fix: Temporarily cap trl to avoid breaking changes from being consumed by @courtneypacheco in #3167
• build(deps): bump docker/build-push-action from 6.13.0 to 6.14.0 by @dependabot in #3174
• Fix: Add missing torch and psutil dependencies to CUDA installation instructions by @jjaegii in #3162
• tests: fix failure due to a slight difference in HF exception message by @booxter in #3177
• chore: Update vLLM to 0.7.3 by @fabiendupont in #3178
• build(deps): bump actions/cache from 4.2.0 to 4.2.1 by @dependabot in #3175
• build(deps): bump actions/download-artifact from 4.1.8 to 4.1.9 by @dependabot in #3189
• infra: fix cloud-instance ec2 wait loop by @ktdreyer in #3186
• ci: label scripts/infra files as CI/CD by @ktdreyer in #3187
• mergify: remove old ruff.sh reference by @ktdreyer in #3188
• docs: update a feat change by @reidliu41 in #3146
• Make SDG batch size configurable via system profile by @npalaska in #3157
• build(deps): bump docker/build-push-action from 6.14.0 to 6.15.0 by @dependabot in #3196
• build(deps): bump docker/metadata-action from 5.6.1 to 5.7.0 by @dependabot in #3195
• build(deps): bump docker/setup-buildx-action from 3.9.0 to 3.10.0 by @dependabot in #3194
• fix(infra): fix nvidia drivers installation script by @booxter in #3190
• feat(infra): add cloud-instance list command; ignore shutting-down instances by @booxter in #3198
• build(deps): bump actions/cache from 4.2.1 to 4.2.2 by @dependabot in #3203
• Add option to search multiple subnets with cloud-instance launch by @danmcp in #3201
• feat(infra): add -i option for instance id to cloud-instance.sh by @booxter in #3199
• Handle INSTANCE_ID and details support for IBM Cloud and add rename m… by @danmcp in #3205
• docs: corrected broken links by @nilesho in #3206
• infra: cloud-instance.sh improvements by @ktdreyer in #3217
• ci: trigger docs job when docs/ are touched by @booxter in #3170
• requirements: Add missing filelock dependency by @booxter in #3049
• Fix import of read_config in debug-llama by @radeksm in #3212
• ci: install git-core package instead of git by @ktdreyer in #3140
• fix: ilab taxonomy diff provides full error details when it fails by @jwm4 in #3220
• docs: Add DK-Bench docs by @alimaredia in #3124
• fix cannot change context in cli issue by @reidliu41 in #2731
• ui: Humanize help message describing available command groups. by @booxter in #3053
• fix(infra): reset nvidia-driver module before enabling -open one by @booxter in #3222
• infra: do not reinstall ibmcloud CLI by @ktdreyer in #3224
• fix: Adding metadata to document chunks by @dmartinol in #3184
• Adds support for simple global/general model configuration by @RobotSail in #3197
• infra: remove os version by @ktdreyer in #3225
• ci: Remove spelling checks by @booxter in #3130
• build(deps): bump actions/download-artifact from 4.1.9 to 4.2.1 by @dependabot in #3240
• build(deps): bump actions/cache from 4.2.2 to 4.2.3 by @dependabot in #3239
• Bump minimum version of SDG to 0.7.2 by @ktdreyer in #3238
• docs: Add MOTD to warn about updating the kernel by @derekhiggins in #3234
• chore: Update lint.yml to reference in-house security lint GH action by @courtneypacheco in #3228
• build(deps): bump actions/setup-python from 5.4.0 to 5.5.0 by @dependabot in #3244
• chore: cap instructlab-training version to maintain stability while breaking changes are introduced by @RobotSail in #3247
• Document removing any CUDA packages before Upgrade by @derekhiggins in #3242
• docs: pin Sphinx to <8.2.0 by @booxter in #3169
• refactor: use CUDA_HOME var when applicable by @nathan-weinberg in #3249
• infra: default cloud-instance.sh to latest CentOS 9 AMI by @ktdreyer in #3248
• chore: update docling api references by @jaideepr97 in #3254
• chore: bump the training library to 0.8.0 and refactor tokenizer usage by @RobotSail in #3250
• Revert "chore: update docling api references" by @courtneypacheco in #3266
• ci: temporarily disable macos workflow by @jaideepr97 in #3270
• re-introduce the docling fix; fix DISABLE_GPU_ACCELERATION setting via GITHUB_ENV file by @booxter in #3271
• chore: Bump SDG version to ensure the latest docling changes are consumed by @courtneypacheco in #3281

New Contributors

• @kami619 made their first contribution in #3166
• @npalaska made their first contribution in #3157
• @nilesho made their first contribution in #3206

Full Changelog: v0.24.3...v0.25

Fixes

• instructlab-sdg recently published a new v0.7.2 release that contains a bugfix related chunking knowledge documents, PDF or Markdown documents containing a table. See the linked tag for more information. This release of instructlab pulls in those fixes.

What's Changed

• deps: Bump lower bound for instructlab-sdg from 0.7.0 to 0.7.2 by @courtneypacheco in #3230

Full Changelog: v0.23.4...v0.23.5

What's Changed

• requirements: Add missing filelock dependency (backport #3049) by @mergify in #3219
• fix: ilab taxonomy diff provides full error details when it fails (backport #3220) by @mergify in #3223

Full Changelog: v0.24.2...v0.24.3

What's Changed

• requirements: Add missing filelock dependency (backport #3049) by @mergify in #3218
• fix: ilab taxonomy diff provides full error details when it fails (backport #3220) by @mergify in #3221

Full Changelog: v0.23.3...v0.23.4

What's Changed

• mergify: remove old ruff.sh reference (backport #3188) by @mergify in #3193
• infra: fix cloud-instance ec2 wait loop (backport #3186) by @mergify in #3204
• Make SDG batch size configurable via system profile (backport #3157) by @mergify in #3207

Full Changelog: v0.24.1...v0.24.2

What's Changed

• tests: fix failure due to a slight difference in HF exception message (backport #3177) by @mergify in #3210
• Make SDG batch size configurable via system profile (backport #3157) by @mergify in #3208

Full Changelog: v0.23.2...v0.23.3

What's Changed

• mergify: remove old ruff.sh reference (backport #3188) by @mergify in #3193

Full Changelog: v0.24.1...v0.24.2a0