v0.22.0

What's Changed

• Make dask_cuda import lazy by @sidnarayanan in #220
• Explicit pad token configuration by @sidnarayanan in #221

Full Changelog: v0.21.0...v0.22.0

v0.21.0

What's Changed

• Added a CITATION.cff by @jamesbraza in #218
• Local LLM call arguments by @sidnarayanan in #219

Full Changelog: v0.20.0...v0.21.0

v0.20.0

What's Changed

• Fixing evaluate_consensus's extract_answer_fn typing by @jamesbraza in #213
• Consensus sampling decomposition and per-group standard error by @jamesbraza in #211
• Fixing compute_pass_at_k return type by @jamesbraza in #215
• Created NoToolsSimpleAgent by @jamesbraza in #214
• Basic local LLM support by @sidnarayanan in #216
• Local LLM Op and Agent by @sidnarayanan in #217

Full Changelog: v0.19.0...v0.20.0

v0.19.0

What's Changed

• remove_nonterminal_branches by @sidnarayanan in #208
• Add num_epochs to OfflineTrainer by @sidnarayanan in #209
• LLM response validation by @sidnarayanan in #212

Full Changelog: v0.18.0...v0.19.0

v0.18.0

What's Changed

• Removed duplicated num_rollouts_per_env by @jamesbraza in #192
• Suppress env frame errors; serialize logprob by @sidnarayanan in #197
• Got SimpleAgent example in the README to work by @jamesbraza in #200
• Better HTTP Error Handling by @ludomitch in #199
• Removed unnecessary before_transition from TrajectoryFileCallback by @jamesbraza in #196
• Removed incorrect SimpleAgent description in docstring by @jamesbraza in #193
• Added forgotten DEFAULT_LLM_COMPLETION_TIMEOUT to agent.__all__ by @jamesbraza in #194
• Lock file maintenance by @renovate in #201
• Added 'pass @ k' to alg by @jamesbraza in #202
• Documenting exact figures/equations for 'pass @ k' alg by @jamesbraza in #203
• TrajectoryFileCallback having async file I/O and controlling Environment dump by @jamesbraza in #205
• Pinning llmclient at 0.0.9 by @sidnarayanan in #206

New Contributors

• @ludomitch made their first contribution in #199

Full Changelog: v0.17.0...v0.18.0

v0.17.0

What's Changed

• Pinned llm client by @maykcaldas in #179
• Renovate lockFileMaintenance respecting schedule, 2-week stability period, no openai pinning by @jamesbraza in #180
• Lock file maintenance by @renovate in #178
• Revert "Exposing Dataset in Callback.before_eval_loop (#177)" by @jamesbraza in #185
• Save env files on callback too by @whitead in #186
• Fix callback file opening by @sidnarayanan in #188
• Lock file maintenance by @renovate in #190
• Improved README by @whitead in #187
• Consensus sampling algorithm by @jamesbraza in #189
• Guaranteed Message type in APEOpt by @maykcaldas in #191

Full Changelog: v0.16.0...v0.17.0

v0.16.0

What's Changed

• Added llmclient dependency by @maykcaldas in #175

New Contributors

• @maykcaldas made their first contribution in #175

Full Changelog: v0.15.0...v0.16.0

v0.15.0

What's Changed

• Lock file maintenance by @renovate in #163
• after_train_step bugfix by @sidnarayanan in #164
• Supporting structured outputs in LLMModel by @jamesbraza in #165
• Fixed codeflash's formatter-cmds by @jamesbraza in #166
• Replay buffers by @sidnarayanan in #167
• Providing a variable for tool_choice by @jamesbraza in #168
• Update astral-sh/setup-uv action to v4 by @renovate in #170
• Update all non-major dependencies by @renovate in #169
• Evaluator num_rollouts_per_env by @sidnarayanan in #171
• Lock file maintenance by @renovate in #173
• Made UIndexMemoryModel test class and combined two tests by @jamesbraza in #172
• Populated LICENSE by @jamesbraza in #174
• Exposing Dataset in Callback.before_eval_loop by @jamesbraza in #177

Full Changelog: v0.14.5...v0.15.0

v0.14.5

What's Changed

• Replay buffer bugfix; pin litellm by @sidnarayanan in #159
• Opting into ruff preview rules and validate-pyproject-schema-store hook swap by @jamesbraza in #160
• Allow q function to return a tensor by @sidnarayanan in #161
• Catch closing exceptions by @sidnarayanan in #162

Full Changelog: v0.14.4...v0.14.5

v0.14.4

What's Changed

• Lock file maintenance by @renovate in #154
• Created new hook Callback.before_rollout by @jamesbraza in #155
• Prioritized replay buffer by @sidnarayanan in #157
• Lock file maintenance by @renovate in #156

Full Changelog: v0.14.3...v0.14.4