unsloth/tests
Daniel Han 6dae2f525b
Stop false RoPE 'default' warning and fix rope drift gate on transformers 5 (#6223)
* Handle rope_type 'default' on transformers 5 to stop false RoPE warning

transformers 5 reports rope_type="default" for every plain (unscaled) config
and dropped "default" from ROPE_INIT_FUNCTIONS. _compute_config_rope_inv_freq
then did ROPE_INIT_FUNCTIONS["default"], hit KeyError, returned None and logged
"Could not apply RoPE scaling 'default'; long-context generation may degrade"
on every model load. The inv_freq was still correct (the constructor recomputes
vanilla on None), but the warning is a false alarm for unscaled models.

Compute the unscaled inv_freq directly for rope_type "default"/None instead of
going through ROPE_INIT_FUNCTIONS, so plain configs return the right value with
no warning. Scaled types (llama3/linear/yarn/...) are unchanged.

Also skip test_object_style_rope_scaling_on_config_delegates_correctly when
transformers strict-validates rope_scaling (5.x): it rejects a non-dict object
on config.rope_scaling, so the object-style delegation path cannot be set up
there. The test still runs and asserts on transformers <5.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2026-06-11 20:37:01 -07:00
..
notebooks CI: scope GITHUB_TOKEN permissions, add MLX CI, unblock ~60 skipped tests (#5312) 2026-05-11 03:19:13 -07:00
python feat: support text-only loading of Gemma 3 27B via FastLanguageModel (skip SiglipVisionModel) (#5816) 2026-06-09 22:52:39 -07:00
qlora Formatting: ruff line-length 100, kwarg-spacing passes, drop blank after short local imports (#6079) 2026-06-08 04:24:13 -07:00
saving Reduce and tighten code comments and docstrings repo-wide (#6095) 2026-06-08 23:09:51 -07:00
security Reduce and tighten code comments and docstrings repo-wide (#6095) 2026-06-08 23:09:51 -07:00
sh Installer: GPU detection follow-ups after #6174 (poisoned venv repair, llama.cpp routing, probe bounds) (#6183) 2026-06-11 05:06:02 -07:00
studio fix(rocm): stop overwriting ROCR_VISIBLE_DEVICES in apply_gpu_ids (#6123) 2026-06-11 16:39:36 +01:00
utils Stop false RoPE 'default' warning and fix rope drift gate on transformers 5 (#6223) 2026-06-11 20:37:01 -07:00
version_compat Reduce and tighten code comments and docstrings repo-wide (#6095) 2026-06-08 23:09:51 -07:00
vllm_compat Reduce and tighten code comments and docstrings repo-wide (#6095) 2026-06-08 23:09:51 -07:00
__init__.py Qwen 3, Bug Fixes (#2445) 2025-04-30 22:38:39 -07:00
_zoo_aggressive_cuda_spoof.py Reduce and tighten code comments and docstrings repo-wide (#6095) 2026-06-08 23:09:51 -07:00
conftest.py Reduce and tighten code comments and docstrings repo-wide (#6095) 2026-06-08 23:09:51 -07:00
run_all.sh studio/setup.sh: cope with fresh CUDA toolkits like 13.3 (#5826) 2026-05-29 05:09:20 -07:00
test_attention_implementation.py fix(gpt-oss): prefer flex attention over sdpa (#5701) 2026-05-22 08:38:38 -07:00
test_callback_signature_drift.py Formatting: ruff line-length 100, kwarg-spacing passes, drop blank after short local imports (#6079) 2026-06-08 04:24:13 -07:00
test_cli_export_unpacking.py Reduce and tighten code comments and docstrings repo-wide (#6095) 2026-06-08 23:09:51 -07:00
test_enforce_kwargs_spacing.py Reduce and tighten code comments and docstrings repo-wide (#6095) 2026-06-08 23:09:51 -07:00
test_finetune_last_n_layers.py Reduce and tighten code comments and docstrings repo-wide (#6095) 2026-06-08 23:09:51 -07:00
test_gemma4_chat_template.py Formatting: ruff line-length 100, kwarg-spacing passes, drop blank after short local imports (#6079) 2026-06-08 04:24:13 -07:00
test_get_model_name.py Formatting: ruff line-length 100, kwarg-spacing passes, drop blank after short local imports (#6079) 2026-06-08 04:24:13 -07:00
test_import_fixes_drift.py Fix bitsandbytes ROCm GPU arch and warp size detection on Windows (#6127) 2026-06-11 06:27:39 -07:00
test_loader_glob_skip.py Formatting: ruff line-length 100, kwarg-spacing passes, drop blank after short local imports (#6079) 2026-06-08 04:24:13 -07:00
test_model_registry.py Reduce and tighten code comments and docstrings repo-wide (#6095) 2026-06-08 23:09:51 -07:00
test_moe_lora_targets.py Merge moe-lora-target-fix CI fixes 2026-06-08 20:20:00 +05:30
test_multi_image_grpo_chunking.py Reduce and tighten code comments and docstrings repo-wide (#6095) 2026-06-08 23:09:51 -07:00
test_nvfp4_quant_load.py Merge nvfp4_load CI fixes 2026-06-08 23:23:53 +05:30
test_peft_weight_converter_compat.py Formatting: ruff line-length 100, kwarg-spacing passes, drop blank after short local imports (#6079) 2026-06-08 04:24:13 -07:00
test_public_api_surface.py Reduce and tighten code comments and docstrings repo-wide (#6095) 2026-06-08 23:09:51 -07:00
test_raw_text.py Reduce and tighten code comments and docstrings repo-wide (#6095) 2026-06-08 23:09:51 -07:00
test_resolve_model_class.py Formatting: ruff line-length 100, kwarg-spacing passes, drop blank after short local imports (#6079) 2026-06-08 04:24:13 -07:00
test_studio_install_workspace_guard.py Reduce and tighten code comments and docstrings repo-wide (#6095) 2026-06-08 23:09:51 -07:00
test_studio_root_resilience.py Formatting: ruff line-length 100, kwarg-spacing passes, drop blank after short local imports (#6079) 2026-06-08 04:24:13 -07:00
test_tool_mask_zoo_compat.py Formatting: ruff line-length 100, kwarg-spacing passes, drop blank after short local imports (#6079) 2026-06-08 04:24:13 -07:00
test_video_path_validation.py fix/validate dataset video paths before training (#5136) 2026-06-10 05:32:32 -07:00
test_windows_rocm_bnb_version.py Auto-set BNB_ROCM_VERSION from the installed wheel on Windows + ROCm (#5986) 2026-06-10 08:11:26 -07:00