vrr/unsloth
2
0
Fork 0
mirror of https://github.com/unslothai/unsloth.git synced 2026-10-21 14:00:32 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 45
unsloth/tests/python
History
Exact
Daniel Han 16e32f3820
Studio: Linux torch 2.10 to 2.12 take our causal-conv1d and mamba-ssm wheels (#14338) 
Stacked on #14336. Merge only after the Linux torch 2.10 to 2.12
causal-conv1d / mamba-ssm cells from #14336 are published on
`prebuilt-wheels-cu13`; before that these URLs would 404 in front of the
upstream fallback.

## Why

Linux users on torch 2.10, 2.11 or 2.12 with a CUDA 13 torch download
upstream's causal-conv1d (205 MB) and mamba-ssm (351 MB) wheels, built
for torch 2.10 with `-lineinfo` and every arch. Our builds of the same
packages for torch 2.13 / 2.14 are about 29 / 54 MB on x86_64 and 42 /
79 MB on aarch64.

## Changes

- `_UNSLOTH_PREBUILT_EXTRA_CELLS` gains linux_x86_64 and linux_aarch64
torch 2.10 / 2.11 / 2.12 on cp311 to cp313.
- `_UNSLOTH_PREBUILT_EXTRA_CELL_PREFIXES` limits those Linux cells to
causal_conv1d and mamba_ssm, so flash-attn keeps upstream 2.8.1 there
(125 MB, smaller than ours).
- Unchanged: a CUDA 12 torch, cp310 and cp314 still resolve upstream,
and torch 2.13 / 2.14 and Windows resolve as before.

## Tests

Updated the tests that pinned upstream URLs for cu13 Linux torch 2.10 to
2.12. They now assert our wheel for cu13 cp311 to cp313 and upstream for
cu12 / cp310. Added tests that flash-attn stays upstream and that every
Linux extra cell the resolver asks for is one the workflow builds.
Affected suites: 1125 passed.
2026-10-20 22:18:45 -07:00
..
fixtures Keep Mistral Small 3.2 tool definitions in training rows (#13597) 2026-10-13 06:41:04 -07:00
__init__.py Consolidate dual venvs and separate install from update (#4530) 2026-03-25 05:24:21 -07:00
_grpo_ddp_worker.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
conftest.py Fix silent dropping of embed_tokens/lm_head when listed in target_modules (#9331) 2026-08-23 21:07:02 -07:00
test_accelerator_skip_guards.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_amd_extras_contract.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_bitsandbytes_kernel_readiness.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_change_system_message.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_compiler_watch_temp_scan.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_conftest_bitsandbytes_preimport.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_construct_chat_template_validation.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_cpo_processor_text_tokenizer.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_cross_platform_parity.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_custom_studio_home_keeps_uv.py Keep uv, its cache and the uv-managed Python inside a custom UNSLOTH_STUDIO_HOME (#13373) 2026-10-12 08:50:31 -07:00
test_desktop_chosen_studio_home.py Studio Desktop: choose where Unsloth installs on Windows (#13446) 2026-10-15 05:16:12 -07:00
test_docker_build_ref_resolution.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_cpu_fallback.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_credential_probe.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_hub_readme.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_labext_cell_nav.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_llama_cuda_backend.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_llama_prebuilt_probe.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_nb_compat_pre_run_cell.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_nb_install_cell_sig.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_nb_ipython_profile_writable.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_nb_populate_retry.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_nb_refresh_state_durability.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_nb_strip_colab_durability.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_nb_strip_colab_race.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_nb_strip_colab_scope.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_nb_sync_publish_ownership.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_nb_sync_race.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_nb_sync_upstream_removal.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_nb_view_ownership.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_nvidia_toolkit_install.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_pip_shim_training_stack.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_ppa_retry.py Docker: retry the deadsnakes PPA add so a Launchpad 504 does not fail the build 2026-09-05 19:43:34 -07:00
test_docker_publish_ref_freeze.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_publish_tag_scheme.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_rocm.py Docker ROCm Studio image: run GGUF chat on the GPU through Vulkan (#13852) 2026-10-15 09:45:28 -07:00
test_docker_run_env_forward.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_run_kernel_cwd.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_run_local_models.py Docker: mount the host's LM Studio, Ollama and Hermes models into the container (#10936) 2026-09-14 19:08:46 -07:00
test_docker_run_output_publish.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_run_pin_scan.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_run_url_keeps_outputs.py Keep the model a training script saves when it runs in Docker (#12994) 2026-10-08 18:57:41 +13:00
test_docker_single_compile_worker_options.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_studio_home_volume.py Studio: keep skills created in Docker on the unsloth-studio volume (#13912) 2026-10-18 05:50:30 -07:00
test_docker_studio_launch_port_guard.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_studio_launch_view_config.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_studio_password_banner.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_studio_port.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_studio_rocm_jupyter.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_studio_stop_grace.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_tf_sidecar_vllm_floor.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_docker_update_helpers.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_dpo_vision_processor_passthrough.py tests: read checked-in files as UTF-8 instead of the platform default (#7438) 2026-07-26 23:31:56 -07:00
test_e2e_no_torch_sandbox.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_embed_lm_head_target_modules_redirect.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_fast_inference_mask_device.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_fast_language_model_text_only.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_fast_model_config_passthrough.py tests: read checked-in files as UTF-8 instead of the platform default (#7438) 2026-07-26 23:31:56 -07:00
test_fast_sentence_transformer_embedding_parity.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_fast_sentence_transformer_prompts.py Keep EmbeddingGemma and Qwen3-Embedding prompts when fine-tuning (#12795) 2026-10-06 01:36:39 -07:00
test_fast_sentence_transformer_redirect_lifecycle.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_flash_attn_install_python_stack.py Studio: Linux torch 2.10 to 2.12 take our causal-conv1d and mamba-ssm wheels (#14338) 2026-10-20 22:18:45 -07:00
test_gemma4_base_bos_token.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_get_chat_template_escaping.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_get_chat_template_sharegpt_mapping.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_get_lora_parameters_bias_fp8_block_size.py FSDP2: shard QLoRA 4-bit weights with fused LoRA kernels kept on (#13880) 2026-10-16 03:48:23 -07:00
test_get_lora_parameters_fp8_block_size.py FSDP2: shard QLoRA 4-bit weights with fused LoRA kernels kept on (#13880) 2026-10-16 03:48:23 -07:00
test_get_lora_parameters_fsdp2_quant_state.py QLoRA quant-state fallback: gate per weight, not a process-global flag (#14231) 2026-10-19 10:21:11 -07:00
test_gkd_chunked_jsd.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_gkd_liger_objective.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_gpt_oss_final_return_token.py End the last gpt-oss training turn with <|return|>, not <|end|> (#13657) 2026-10-14 14:52:09 +13:00
test_gpu_init_ldconfig_guard.py tests: read checked-in files as UTF-8 instead of the platform default (#7438) 2026-07-26 23:31:56 -07:00
test_grpo_ddp_gradient_sync.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_grpo_ddp_model_config.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_grpo_hidden_states_pre_head.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_grpo_logit_transform_fallback.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_grpo_logit_transform_families.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_grpo_update_sampling_params.py Stop ignoring the min_p and stop settings the GRPO notebooks pass to vLLM (#11595) 2026-09-23 05:56:53 -07:00
test_import_without_bitsandbytes.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_install_python_stack.py Studio: download prebuilt kernel wheels once (#14051) 2026-10-20 08:32:40 -07:00
test_install_uv_override_space.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_installer_conda_path_guard.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_installer_differential_comparer.py Tests: make the --help, media scan, installer comparer and harness resolver checks hold on every host (#14116) 2026-10-18 04:05:42 -07:00
test_installer_embedded_nvidia_probe.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_installer_helper_declaration_order.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_installer_master_root.py Studio: honor UNSLOTH_HOME in installers (#14293) 2026-10-20 22:13:20 +08:00
test_loader_resolution_reporting.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_mistral_tools_chat_template.py Keep Mistral Small 3.2 tool definitions in training rows (#13597) 2026-10-13 06:41:04 -07:00
test_mlx_public_trainer_api.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_module_entry_point.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_no_torch_filtering.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_orpo_processor_text_tokenizer.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_pad_row_kept_when_emitted.py Keep the pad embedding row when the tokenizer emits the pad token, never zero an EOS row (#13628) 2026-10-14 04:00:56 -07:00
test_pad_token_fix.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_patch_trl_rl_trainers_defensive.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_posix_only_marks_import_on_windows.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_prebuilt_cuda_wheels.py Studio: Linux torch 2.10 to 2.12 take our causal-conv1d and mamba-ssm wheels (#14338) 2026-10-20 22:18:45 -07:00
test_prebuilt_cuda_wheels_arm.py Prebuilt CUDA wheels: pin NVIDIA's re-published CUDA 13.4.1 manifest (win_arm64) (#13954) 2026-10-15 20:21:19 -07:00
test_prebuilt_cuda_wheels_pure.py Prebuilt wheels: drop -lineinfo, unreachable SASS and flash-attn PTX; add extension-free mamba-ssm (#14142) 2026-10-19 04:40:02 -07:00
test_prebuilt_cuda_wheels_targets.py Prebuilt wheels: drop -lineinfo, unreachable SASS and flash-attn PTX; add extension-free mamba-ssm (#14142) 2026-10-19 04:40:02 -07:00
test_prebuilt_cuda_wheels_windows.py Prebuilt wheels: build causal-conv1d and mamba-ssm for Linux torch 2.10 to 2.12 (#14336) 2026-10-20 06:11:17 -07:00
test_pwsh_runner_encoding.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_remote_code_requirements.py Studio: load and run DeepSeek-OCR and DeepSeek-OCR-2 (#13310) 2026-10-11 20:37:18 -07:00
test_remove_special_tokens_no_bos.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_revision_forwarding.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_rl_config_pickling.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_rocm_aotriton_gate.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_rocm_bf16_capability.py Core: decide bfloat16, compile and TMA options from every visible NVIDIA GPU on mixed hosts (#13923) 2026-10-15 21:58:52 -07:00
test_sentence_transformer_config_kwargs.py Fix notebook failures on Kaggle T4x2 and duplicate import warnings (#13024) 2026-10-08 04:13:01 -07:00
test_sentence_transformer_gemma_masks.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_sentence_transformer_max_seq_length.py Use the requested max_seq_length for encoder embedding models (#12915) 2026-10-07 18:16:33 +13:00
test_sentence_transformer_merged_pair_loss.py SentenceTransformer: add opt-in FP32 merged-pair ranking loss (#11867) 2026-09-26 09:50:32 -07:00
test_sentence_transformer_unpadding.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_sign_wheel_binaries.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_ssm_kernels_install_python_stack.py Studio: restore Qwen causal-conv1d on demand, narrow fatal mamba-ssm, provision SSM kernels in export (#13971) 2026-10-19 04:32:13 -07:00
test_studio_import_no_torch.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_studio_runtime_gate.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_text_only_remote_composite.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_to_sharegpt_optional_none.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_tokenizers_and_torch_constraint.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_torchcodec_torch_compat.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_unsloth_nb_pip_magic.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_unsloth_pip_shim.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_unsloth_run_tool_policy_resolver.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_uv_cache_selector_agreement.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_uv_requirements_path_space.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_v100_fullft_precision.py Tests: run the precision template in one namespace, keep the block-compile packing test off leaked env (#14425) 2026-10-20 19:14:17 -07:00
test_virustotal_scan.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_vision_lora_targeting.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_whisper_source_build_root.py Studio: update in the background and restart when ready (#9890) 2026-08-29 20:11:35 +12:00
test_windows_arm64_requirement_markers.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_windows_arm64_wheelhouse_tags.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_windows_git_gate.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_windows_installer_concurrency_guard.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_windows_installer_resolver_fallback.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_windows_no_torch_setup.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_windows_on_arm_python_arch.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_windows_python_venv_hardening.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_windows_setup_download_progress.py Tests: give the setup.ps1 download progress pwsh its own startup cache, and catch argv0-less pwsh spawns (#12882) 2026-10-06 11:37:23 -07:00
test_windows_setup_output_encoding.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_windows_studio_update_launcher.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_windows_uv_cache_selection.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_windows_uv_managed_python.py Windows installer: always build Studio on a uv-managed Python, never one already installed (#13399) 2026-10-12 06:27:51 -07:00
test_windows_vcredist_download_tls.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_windows_xformers_installer.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00
test_windows_xformers_wheel_match.py Trim comments in tests (#13844) 2026-10-15 05:30:27 -07:00