fix: raise error when prefill exceeds sliding window KV cache size (#2215)

chore(ruff): bump target-version to py310 and update type annotations (#2221)

Added correct rope_base to all Mistral configs (#2218)

fix: handle empty prompt list in R1Base.apply() (#2216)

Add YaRN rope adjustment + DeepSeekV3 rope_interleave (#2202)

chore(workflows): Sync GPU CI workflow image with lightning-thunder repo (#2208)

fix: flaky `test_litgpt_chat_endtoend` (#2209)

fix: default val_split_fraction to 0.05 for single-file JSON data (#2204)

fix: correct LoRA index computation uses wrong feature size (#2201)

feat(logger) : Add LitLogger Support for Lightning.ai Experiment Tracking (#2198)

build(deps): update huggingface-hub requirement from <1.3,>=0.30 to >=0.30,<1.4 (#2200)

build(deps-dev): update bitsandbytes requirement from <0.43,>=0.42 to >=0.42,<0.50 (#2199)

Allow train.max_steps for profiling runs (#2193)

fix: gemma-3 checkpoint conversion from litgpt to hf (#2195)

feat: add generate_strategy option to litgpt serve (#2188)