mirror of https://github.com/kubeflow/trainer.git
* chore(manifests): add manifests for llama3_2_1B and llama3_2_3B. Signed-off-by: Electronic-Waste <2690692950@qq.com> * chore(manifests): Add detailed configuration for llama3_2_1B. Signed-off-by: Electronic-Waste <2690692950@qq.com> * chore(manifests): Add detailed configuration for llama3_2_3B. Signed-off-by: Electronic-Waste <2690692950@qq.com> * chore(manifests): load pvc to trainer node & add initializers. Signed-off-by: Electronic-Waste <2690692950@qq.com> * fix(manifest): fix DependsOn error in CI. Signed-off-by: Electronic-Waste <2690692950@qq.com> * fix(runtime): fix launching command for CTRs. Signed-off-by: Electronic-Waste <2690692950@qq.com> * fix(manifest): remove torchtune temporally in base/runtimes. Signed-off-by: Electronic-Waste <2690692950@qq.com> * fix(manifest): provide torchtune-trainer image tag override in overlays. Signed-off-by: Electronic-Waste <2690692950@qq.com> * fix(manifest): update image for data-initializer and model-initializer. Signed-off-by: Electronic-Waste <2690692950@qq.com> --------- Signed-off-by: Electronic-Waste <2690692950@qq.com> |
||
---|---|---|
.. | ||
base | ||
overlays | ||
third-party/jobset |