docs: consolidate training/fine-tuning guides under training_guides/ - #256
Merged
Merged
Conversation
Move every training and fine-tuning page (mdx, md, ipynb) plus their assets out of kubeflow/how_to/ and workbench/how_to/ into a new top-level docs/en/training_guides/ section. Add an index page that points to each guide based on the user's workflow. Tighten each moved guide: drop verbose background, collapse duplicate parameter / data sections, and keep the prerequisites, YAMLs, and notebooks intact. Standardize file names to kebab-case. Update internal links (training_guides → workbench/how_to / model_inference / kueue references) and external references from the MLOps coding-agents page. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Quote `<your-namespace>` and `<vendor-fine-tuning-image>` so the embedded YAML is copy-pasteable without YAML parsing surprises. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
WalkthroughThis PR consolidates training and fine-tuning documentation from deprecated ChangesTraining Guides Reorganization
Estimated code review effort🎯 2 (Simple) | ⏱️ ~10 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
There was a problem hiding this comment.
Actionable comments posted: 8
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@docs/en/training_guides/fine-tune-and-pretrain-llms-on-ascend-npu.mdx`:
- Around line 11-13: The three notebook links (`qwen3_finetune_verify.ipynb`,
`qwen25_pretrain_verify.ipynb`, `qwen3_0.6b_finetune_verify.ipynb`) are
root-absolute (`/...ipynb`) and will break after the reorg; update each link in
the table to a relative path (e.g. remove the leading slash or use
`./qwen3_finetune_verify.ipynb`) so the markdown in this file
(`fine-tune-and-pretrain-llms-on-ascend-npu.mdx`) points to the notebooks now
located under the same `training_guides` directory.
In `@docs/en/training_guides/fine-tune-with-trainer-v2.mdx`:
- Around line 30-33: The links to the notebooks use GitHub "/tree/" paths which
are for directories and can break navigation; update both URLs in the text that
reference fine-tune-with-trainer-v2.ipynb and
fine-tune-with-trainer-v2-mindspeed-npu.ipynb to point to the actual file blob
URLs (replace "/tree/" with "/blob/" and include the branch/path as in the repo)
so the links open the notebook file directly; ensure you update the two
occurrences that mention the notebook filenames in the paragraph describing
TrainingRuntime/TrainJob and the Huawei Ascend NPU note.
- Line 46: Update the YAML example where the line reads namespace:
<your-namespace> to quote the placeholder so parsers don't misinterpret angle
brackets; locate the YAML block containing the symbol "namespace:
<your-namespace>" in the fine-tune-with-trainer-v2.mdx doc and change it to use
quoted placeholder (e.g., namespace: "<your-namespace>" or namespace:
'<your-namespace>').
In `@docs/en/training_guides/fine-tuning-using-notebooks.mdx`:
- Around line 163-165: Remove the unsafe git configuration flag `-c
http.sslVerify=false` from the documented git commands (the `git -c
http.sslVerify=false -c lfs.activitytimeout=36000 clone
"https://${gitauth}@${BASE_MODEL_URL#https://}"` and the `(cd ${BASE_MODEL_NAME}
&& git -c http.sslVerify=false -c lfs.activitytimeout=36000 lfs pull)`) and the
other occurrences referenced (around the other snippets at 169-170 and 270),
leaving only `-c lfs.activitytimeout=36000` (or no `-c` flags) and update
surrounding text to advise configuring proper CA trust instead of disabling TLS
verification.
- Around line 406-408: The non-NVIDIA vendor template still uses hard-coded
NVIDIA resource keys (nvidia.com/gpualloc, nvidia.com/gpucores,
nvidia.com/gpumem); update that resource block in the fine-tuning notebook
content to use explicit vendor placeholders (e.g., <vendor>.com/gpualloc,
<vendor>.com/gpucores, <vendor>.com/gpumem or a descriptive placeholder like
ACME.com/*) so the example is generic and schedulable on non-NVIDIA
accelerators; ensure any accompanying text mentions replacing <vendor> with the
actual accelerator vendor name.
In `@docs/en/training_guides/kubeflow-trainer-quick-start.md`:
- Line 93: The notebook link for "kubeflow-trainer-mnist.ipynb" currently uses a
/tree/ URL which points to a directory view; update the URL in the markdown line
that contains the link text kubeflow-trainer-mnist.ipynb to use the GitHub
/blob/ path (or the raw file URL) instead of /tree/ so the notebook can be
opened/downloaded directly.
In `@docs/en/training_guides/training-hub-fine-tuning.mdx`:
- Around line 48-49: Replace the GitHub "tree" URLs for
sft-comprehensive-tutorial.ipynb and osft-comprehensive-tutorial.ipynb with
direct raw-file or repo-relative documentation links so the table's "Download"
action opens/serves the notebook file (for example use the
raw.githubusercontent.com URL for each .ipynb or a docs permalink that serves
the file) and update the table entries for "sft-comprehensive-tutorial.ipynb"
and "osft-comprehensive-tutorial.ipynb" accordingly.
In `@docs/en/training_guides/training-runtimes.mdx`:
- Around line 52-58: The smoke-test instructions are incomplete: before applying
trainjob-smoke.yaml, download the remote file (trainjob-smoke.yaml), edit it to
set metadata.namespace to your namespace and spec.runtimeRef.name to the runtime
you want, then kubectl apply the edited file; alternatively show a patch/kubectl
-n <your-namespace> apply -f /tmp/trainjob-smoke.yaml example so the subsequent
kubectl -n <your-namespace> get trainjobs and the trainjob=$(kubectl -n
<your-namespace> ...) log command target the same namespace and runtime.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 83ff1be6-fbf5-42d4-ab29-d4f0df007f90
⛔ Files ignored due to path filters (1)
docs/en/training_guides/assets/build-train-image/uv.lockis excluded by!**/*.lock
📒 Files selected for processing (31)
docs/en/kubeflow/how_to/fine-tune-with-trainer-v2.mdxdocs/en/kubeflow/how_to/training-runtimes.mdxdocs/en/model_inference/inference_service/how_to/mlops_with_coding_agents.mdxdocs/en/training_guides/assets/build-train-image/fine_tune_with_llamafactory.Containerfiledocs/en/training_guides/assets/build-train-image/fine_tune_with_llamafactory_npu.Containerfiledocs/en/training_guides/assets/build-train-image/pyproject.tomldocs/en/training_guides/assets/kf-trainingruntime.yamldocs/en/training_guides/assets/kf-trainjob.yamldocs/en/training_guides/assets/training-runtimes/llamafactory0.9-cann8.5-arm64-trainingruntime.yamldocs/en/training_guides/assets/training-runtimes/llamafactory0.9-cu126-amd64-trainingruntime.yamldocs/en/training_guides/assets/training-runtimes/mindspeed-llm-cann8.5-arm64-trainingruntime.yamldocs/en/training_guides/assets/training-runtimes/torch2.6-cann8.5-arm64-trainingruntime.yamldocs/en/training_guides/assets/training-runtimes/torch2.6-cu126-amd64-trainingruntime.yamldocs/en/training_guides/assets/training-runtimes/traininghub0.1-cu126-amd64-trainingruntime.yamldocs/en/training_guides/assets/training-runtimes/trainjob-smoke.yamldocs/en/training_guides/fine-tune-and-pretrain-llms-on-ascend-npu.mdxdocs/en/training_guides/fine-tune-with-trainer-v2-mindspeed-npu.ipynbdocs/en/training_guides/fine-tune-with-trainer-v2.ipynbdocs/en/training_guides/fine-tune-with-trainer-v2.mdxdocs/en/training_guides/fine-tuning-using-notebooks.mdxdocs/en/training_guides/index.mdxdocs/en/training_guides/kubeflow-trainer-mnist.ipynbdocs/en/training_guides/kubeflow-trainer-quick-start.mddocs/en/training_guides/osft-comprehensive-tutorial.ipynbdocs/en/training_guides/sft-comprehensive-tutorial.ipynbdocs/en/training_guides/training-hub-fine-tuning.mdxdocs/en/training_guides/training-runtimes.mdxdocs/en/workbench/how_to/fine-tune-and-pretrain-llms-on-ascend-npu.mdxdocs/en/workbench/how_to/fine_tunning_using_notebooks.mdxdocs/en/workbench/how_to/kubeflow_trainer_quick_start.mddocs/en/workbench/how_to/training_hub_fine_tuning.mdx
💤 Files with no reviewable changes (6)
- docs/en/kubeflow/how_to/training-runtimes.mdx
- docs/en/workbench/how_to/fine-tune-and-pretrain-llms-on-ascend-npu.mdx
- docs/en/workbench/how_to/training_hub_fine_tuning.mdx
- docs/en/workbench/how_to/fine_tunning_using_notebooks.mdx
- docs/en/kubeflow/how_to/fine-tune-with-trainer-v2.mdx
- docs/en/workbench/how_to/kubeflow_trainer_quick_start.md
Comment on lines
+11
to
+13
| | [`qwen3_finetune_verify.ipynb`](/qwen3_finetune_verify.ipynb) | `PyTorch CANN` | Full-parameter SFT of `Qwen3-8B` (MindSpeed-LLM) | | ||
| | [`qwen25_pretrain_verify.ipynb`](/qwen25_pretrain_verify.ipynb) | `PyTorch CANN` | Pretraining of `Qwen2.5-7B` (MindSpeed-LLM) | | ||
| | [`qwen3_0.6b_finetune_verify.ipynb`](/qwen3_0.6b_finetune_verify.ipynb) | `MindSpore CANN` | Full-parameter SFT of `Qwen3-0.6B` (bundled MindSpeed-Core-MS + MindSpeed-LLM) | |
There was a problem hiding this comment.
Use relative notebook links to avoid broken navigation after the move.
On Line 11, Line 12, and Line 13, the notebook URLs are root-absolute (/...ipynb). In this reorg, these assets were moved under training_guides, so root-absolute links can 404 depending on site routing.
Suggested fix
-| [`qwen3_finetune_verify.ipynb`](/qwen3_finetune_verify.ipynb) | `PyTorch CANN` | Full-parameter SFT of `Qwen3-8B` (MindSpeed-LLM) |
-| [`qwen25_pretrain_verify.ipynb`](/qwen25_pretrain_verify.ipynb) | `PyTorch CANN` | Pretraining of `Qwen2.5-7B` (MindSpeed-LLM) |
-| [`qwen3_0.6b_finetune_verify.ipynb`](/qwen3_0.6b_finetune_verify.ipynb) | `MindSpore CANN` | Full-parameter SFT of `Qwen3-0.6B` (bundled MindSpeed-Core-MS + MindSpeed-LLM) |
+| [`qwen3_finetune_verify.ipynb`](./qwen3_finetune_verify.ipynb) | `PyTorch CANN` | Full-parameter SFT of `Qwen3-8B` (MindSpeed-LLM) |
+| [`qwen25_pretrain_verify.ipynb`](./qwen25_pretrain_verify.ipynb) | `PyTorch CANN` | Pretraining of `Qwen2.5-7B` (MindSpeed-LLM) |
+| [`qwen3_0.6b_finetune_verify.ipynb`](./qwen3_0.6b_finetune_verify.ipynb) | `MindSpore CANN` | Full-parameter SFT of `Qwen3-0.6B` (bundled MindSpeed-Core-MS + MindSpeed-LLM) |📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| | [`qwen3_finetune_verify.ipynb`](/qwen3_finetune_verify.ipynb) | `PyTorch CANN` | Full-parameter SFT of `Qwen3-8B` (MindSpeed-LLM) | | |
| | [`qwen25_pretrain_verify.ipynb`](/qwen25_pretrain_verify.ipynb) | `PyTorch CANN` | Pretraining of `Qwen2.5-7B` (MindSpeed-LLM) | | |
| | [`qwen3_0.6b_finetune_verify.ipynb`](/qwen3_0.6b_finetune_verify.ipynb) | `MindSpore CANN` | Full-parameter SFT of `Qwen3-0.6B` (bundled MindSpeed-Core-MS + MindSpeed-LLM) | | |
| | [`qwen3_finetune_verify.ipynb`](./qwen3_finetune_verify.ipynb) | `PyTorch CANN` | Full-parameter SFT of `Qwen3-8B` (MindSpeed-LLM) | | |
| | [`qwen25_pretrain_verify.ipynb`](./qwen25_pretrain_verify.ipynb) | `PyTorch CANN` | Pretraining of `Qwen2.5-7B` (MindSpeed-LLM) | | |
| | [`qwen3_0.6b_finetune_verify.ipynb`](./qwen3_0.6b_finetune_verify.ipynb) | `MindSpore CANN` | Full-parameter SFT of `Qwen3-0.6B` (bundled MindSpeed-Core-MS + MindSpeed-LLM) | |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@docs/en/training_guides/fine-tune-and-pretrain-llms-on-ascend-npu.mdx` around
lines 11 - 13, The three notebook links (`qwen3_finetune_verify.ipynb`,
`qwen25_pretrain_verify.ipynb`, `qwen3_0.6b_finetune_verify.ipynb`) are
root-absolute (`/...ipynb`) and will break after the reorg; update each link in
the table to a relative path (e.g. remove the leading slash or use
`./qwen3_finetune_verify.ipynb`) so the markdown in this file
(`fine-tune-and-pretrain-llms-on-ascend-npu.mdx`) points to the notebooks now
located under the same `training_guides` directory.
Comment on lines
+30
to
+33
| Download [`fine-tune-with-trainer-v2.ipynb`](https://github.com/alauda/aml-docs/tree/master/docs/en/training_guides/fine-tune-with-trainer-v2.ipynb) into your workbench and follow the cells. The notebook creates a `TrainingRuntime`, then submits a `TrainJob` that mounts the shared PVC and uses the `aml-image-builder-secret`. | ||
|
|
||
| For Huawei Ascend NPUs, use [`fine-tune-with-trainer-v2-mindspeed-npu.ipynb`](https://github.com/alauda/aml-docs/tree/master/docs/en/training_guides/fine-tune-with-trainer-v2-mindspeed-npu.ipynb) instead — it runs the MindSpeed-LLM SFT pipeline (HF → MCore checkpoint, preprocess, train) on `huawei.com/Ascend910B4` resources with `runtimeClassName: ascend`. | ||
|
|
There was a problem hiding this comment.
Use file URLs for notebook links instead of /tree/ paths.
Both notebook links target files via /tree/, which is intended for directories and can break user navigation.
Suggested doc fix
-Download [`fine-tune-with-trainer-v2.ipynb`](https://github.com/alauda/aml-docs/tree/master/docs/en/training_guides/fine-tune-with-trainer-v2.ipynb) into your workbench and follow the cells.
+Download [`fine-tune-with-trainer-v2.ipynb`](https://github.com/alauda/aml-docs/blob/master/docs/en/training_guides/fine-tune-with-trainer-v2.ipynb) into your workbench and follow the cells.
-For Huawei Ascend NPUs, use [`fine-tune-with-trainer-v2-mindspeed-npu.ipynb`](https://github.com/alauda/aml-docs/tree/master/docs/en/training_guides/fine-tune-with-trainer-v2-mindspeed-npu.ipynb) instead
+For Huawei Ascend NPUs, use [`fine-tune-with-trainer-v2-mindspeed-npu.ipynb`](https://github.com/alauda/aml-docs/blob/master/docs/en/training_guides/fine-tune-with-trainer-v2-mindspeed-npu.ipynb) instead📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| Download [`fine-tune-with-trainer-v2.ipynb`](https://github.com/alauda/aml-docs/tree/master/docs/en/training_guides/fine-tune-with-trainer-v2.ipynb) into your workbench and follow the cells. The notebook creates a `TrainingRuntime`, then submits a `TrainJob` that mounts the shared PVC and uses the `aml-image-builder-secret`. | |
| For Huawei Ascend NPUs, use [`fine-tune-with-trainer-v2-mindspeed-npu.ipynb`](https://github.com/alauda/aml-docs/tree/master/docs/en/training_guides/fine-tune-with-trainer-v2-mindspeed-npu.ipynb) instead — it runs the MindSpeed-LLM SFT pipeline (HF → MCore checkpoint, preprocess, train) on `huawei.com/Ascend910B4` resources with `runtimeClassName: ascend`. | |
| Download [`fine-tune-with-trainer-v2.ipynb`](https://github.com/alauda/aml-docs/blob/master/docs/en/training_guides/fine-tune-with-trainer-v2.ipynb) into your workbench and follow the cells. The notebook creates a `TrainingRuntime`, then submits a `TrainJob` that mounts the shared PVC and uses the `aml-image-builder-secret`. | |
| For Huawei Ascend NPUs, use [`fine-tune-with-trainer-v2-mindspeed-npu.ipynb`](https://github.com/alauda/aml-docs/blob/master/docs/en/training_guides/fine-tune-with-trainer-v2-mindspeed-npu.ipynb) instead — it runs the MindSpeed-LLM SFT pipeline (HF → MCore checkpoint, preprocess, train) on `huawei.com/Ascend910B4` resources with `runtimeClassName: ascend`. |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@docs/en/training_guides/fine-tune-with-trainer-v2.mdx` around lines 30 - 33,
The links to the notebooks use GitHub "/tree/" paths which are for directories
and can break navigation; update both URLs in the text that reference
fine-tune-with-trainer-v2.ipynb and
fine-tune-with-trainer-v2-mindspeed-npu.ipynb to point to the actual file blob
URLs (replace "/tree/" with "/blob/" and include the branch/path as in the repo)
so the links open the notebook file directly; ensure you update the two
occurrences that mention the notebook filenames in the paragraph describing
TrainingRuntime/TrainJob and the Huawei Ascend NPU note.
| kind: LocalQueue | ||
| metadata: | ||
| name: local-queue | ||
| namespace: <your-namespace> |
There was a problem hiding this comment.
Quote angle-bracket placeholder in YAML example for copy-paste safety.
<your-namespace> should be quoted in YAML examples to avoid parse issues in some tools.
Suggested doc fix
- namespace: <your-namespace>
+ namespace: "<your-namespace>"📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| namespace: <your-namespace> | |
| namespace: "<your-namespace>" |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@docs/en/training_guides/fine-tune-with-trainer-v2.mdx` at line 46, Update the
YAML example where the line reads namespace: <your-namespace> to quote the
placeholder so parsers don't misinterpret angle brackets; locate the YAML block
containing the symbol "namespace: <your-namespace>" in the
fine-tune-with-trainer-v2.mdx doc and change it to use quoted placeholder (e.g.,
namespace: "<your-namespace>" or namespace: '<your-namespace>').
Comment on lines
+163
to
+165
| GIT_LFS_SKIP_SMUDGE=1 git -c http.sslVerify=false -c lfs.activitytimeout=36000 \ | ||
| clone "https://${gitauth}@${BASE_MODEL_URL#https://}" | ||
| (cd ${BASE_MODEL_NAME} && git -c http.sslVerify=false -c lfs.activitytimeout=36000 lfs pull) |
There was a problem hiding this comment.
Remove http.sslVerify=false from documented Git commands.
These commands teach users to bypass TLS certificate validation for clone/pull/push, which weakens transport security and can expose credentials in MITM scenarios. Keep TLS verification on; document proper CA trust configuration instead.
Suggested doc patch
- GIT_LFS_SKIP_SMUDGE=1 git -c http.sslVerify=false -c lfs.activitytimeout=36000 \
+ GIT_LFS_SKIP_SMUDGE=1 git -c lfs.activitytimeout=36000 \
clone "https://${gitauth}@${BASE_MODEL_URL#https://}"
- (cd ${BASE_MODEL_NAME} && git -c http.sslVerify=false -c lfs.activitytimeout=36000 lfs pull)
+ (cd ${BASE_MODEL_NAME} && git -c lfs.activitytimeout=36000 lfs pull)
- git -c http.sslVerify=false -c lfs.activitytimeout=36000 \
+ git -c lfs.activitytimeout=36000 \
clone "https://${gitauth}@${DATASET_URL#https://}"
- git -c http.sslVerify=false -c lfs.activitytimeout=36000 push -u ${PUSH_URL} sft-${push_branch}
+ git -c lfs.activitytimeout=36000 push -u ${PUSH_URL} sft-${push_branch}Also applies to: 169-170, 270-270
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@docs/en/training_guides/fine-tuning-using-notebooks.mdx` around lines 163 -
165, Remove the unsafe git configuration flag `-c http.sslVerify=false` from the
documented git commands (the `git -c http.sslVerify=false -c
lfs.activitytimeout=36000 clone "https://${gitauth}@${BASE_MODEL_URL#https://}"`
and the `(cd ${BASE_MODEL_NAME} && git -c http.sslVerify=false -c
lfs.activitytimeout=36000 lfs pull)`) and the other occurrences referenced
(around the other snippets at 169-170 and 270), leaving only `-c
lfs.activitytimeout=36000` (or no `-c` flags) and update surrounding text to
advise configuring proper CA trust instead of disabling TLS verification.
Comment on lines
+406
to
+408
| nvidia.com/gpualloc: "1" | ||
| nvidia.com/gpucores: "50" | ||
| nvidia.com/gpumem: "8192" |
There was a problem hiding this comment.
Vendor template still hard-codes NVIDIA resource keys in the non-NVIDIA section.
In a “non-NVIDIA” template, nvidia.com/* limits are misleading and can make the example unschedulable on other accelerators. Replace these with explicit vendor placeholders.
Suggested doc patch
- nvidia.com/gpualloc: "1"
- nvidia.com/gpucores: "50"
- nvidia.com/gpumem: "8192"
+ <vendor-device-resource-key>: "1"
+ # Optional vendor-specific subresources:
+ # <vendor-device-cores-key>: "50"
+ # <vendor-device-memory-key>: "8192"📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| nvidia.com/gpualloc: "1" | |
| nvidia.com/gpucores: "50" | |
| nvidia.com/gpumem: "8192" | |
| <vendor-device-resource-key>: "1" | |
| # Optional vendor-specific subresources: | |
| # <vendor-device-cores-key>: "50" | |
| # <vendor-device-memory-key>: "8192" |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@docs/en/training_guides/fine-tuning-using-notebooks.mdx` around lines 406 -
408, The non-NVIDIA vendor template still uses hard-coded NVIDIA resource keys
(nvidia.com/gpualloc, nvidia.com/gpucores, nvidia.com/gpumem); update that
resource block in the fine-tuning notebook content to use explicit vendor
placeholders (e.g., <vendor>.com/gpualloc, <vendor>.com/gpucores,
<vendor>.com/gpumem or a descriptive placeholder like ACME.com/*) so the example
is generic and schedulable on non-NVIDIA accelerators; ensure any accompanying
text mentions replacing <vendor> with the actual accelerator vendor name.
|
|
||
| The notebook installs Python packages and downloads MNIST, so the workbench needs outbound network access. | ||
|
|
||
| Download [`kubeflow-trainer-mnist.ipynb`](https://github.com/alauda/aml-docs/tree/master/docs/en/training_guides/kubeflow-trainer-mnist.ipynb) and upload it to your workbench, then follow it to submit the `TrainJob`. |
There was a problem hiding this comment.
Fix the notebook URL path (/tree/ → /blob/ or raw).
The current notebook link points to a file via /tree/, which is a directory route and can break direct navigation/download for users.
Suggested doc fix
-Download [`kubeflow-trainer-mnist.ipynb`](https://github.com/alauda/aml-docs/tree/master/docs/en/training_guides/kubeflow-trainer-mnist.ipynb) and upload it to your workbench, then follow it to submit the `TrainJob`.
+Download [`kubeflow-trainer-mnist.ipynb`](https://github.com/alauda/aml-docs/blob/master/docs/en/training_guides/kubeflow-trainer-mnist.ipynb) and upload it to your workbench, then follow it to submit the `TrainJob`.📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| Download [`kubeflow-trainer-mnist.ipynb`](https://github.com/alauda/aml-docs/tree/master/docs/en/training_guides/kubeflow-trainer-mnist.ipynb) and upload it to your workbench, then follow it to submit the `TrainJob`. | |
| Download [`kubeflow-trainer-mnist.ipynb`](https://github.com/alauda/aml-docs/blob/master/docs/en/training_guides/kubeflow-trainer-mnist.ipynb) and upload it to your workbench, then follow it to submit the `TrainJob`. |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@docs/en/training_guides/kubeflow-trainer-quick-start.md` at line 93, The
notebook link for "kubeflow-trainer-mnist.ipynb" currently uses a /tree/ URL
which points to a directory view; update the URL in the markdown line that
contains the link text kubeflow-trainer-mnist.ipynb to use the GitHub /blob/
path (or the raw file URL) instead of /tree/ so the notebook can be
opened/downloaded directly.
Comment on lines
+48
to
+49
| | SFT comprehensive tutorial | SFT | [`sft-comprehensive-tutorial.ipynb`](https://github.com/alauda/aml-docs/tree/master/docs/en/training_guides/sft-comprehensive-tutorial.ipynb) | | ||
| | OSFT comprehensive tutorial | OSFT | [`osft-comprehensive-tutorial.ipynb`](https://github.com/alauda/aml-docs/tree/master/docs/en/training_guides/osft-comprehensive-tutorial.ipynb) | |
There was a problem hiding this comment.
Fix notebook “Download” links to actual downloadable targets.
The table labels these as downloads, but https://github.com/.../tree/master/...ipynb opens repo pages instead of directly downloading/opening the notebook content. Use raw file URLs (or repo-relative docs links) so the action matches user intent.
Suggested patch
-| SFT comprehensive tutorial | SFT | [`sft-comprehensive-tutorial.ipynb`](https://github.com/alauda/aml-docs/tree/master/docs/en/training_guides/sft-comprehensive-tutorial.ipynb) |
-| OSFT comprehensive tutorial | OSFT | [`osft-comprehensive-tutorial.ipynb`](https://github.com/alauda/aml-docs/tree/master/docs/en/training_guides/osft-comprehensive-tutorial.ipynb) |
+| SFT comprehensive tutorial | SFT | [`sft-comprehensive-tutorial.ipynb`](https://raw.githubusercontent.com/alauda/aml-docs/master/docs/en/training_guides/sft-comprehensive-tutorial.ipynb) |
+| OSFT comprehensive tutorial | OSFT | [`osft-comprehensive-tutorial.ipynb`](https://raw.githubusercontent.com/alauda/aml-docs/master/docs/en/training_guides/osft-comprehensive-tutorial.ipynb) |🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@docs/en/training_guides/training-hub-fine-tuning.mdx` around lines 48 - 49,
Replace the GitHub "tree" URLs for sft-comprehensive-tutorial.ipynb and
osft-comprehensive-tutorial.ipynb with direct raw-file or repo-relative
documentation links so the table's "Download" action opens/serves the notebook
file (for example use the raw.githubusercontent.com URL for each .ipynb or a
docs permalink that serves the file) and update the table entries for
"sft-comprehensive-tutorial.ipynb" and "osft-comprehensive-tutorial.ipynb"
accordingly.
Comment on lines
+52
to
+58
| A shared smoke template applies to any runtime — set `spec.runtimeRef.name` to the runtime you want to exercise: | ||
|
|
||
| ```bash | ||
| kubectl apply -f $base/trainjob-smoke.yaml | ||
| kubectl -n <your-namespace> get trainjobs | ||
| trainjob=$(kubectl -n <your-namespace> get trainjobs -o name | tail -1) | ||
| kubectl -n <your-namespace> logs -f -l jobset.sigs.k8s.io/jobset-name=${trainjob##*/}-node |
There was a problem hiding this comment.
TrainJob smoke-test commands are incomplete and can fail as written
Line 52 says to set spec.runtimeRef.name, but Line 55 applies a remote YAML directly, so users never actually change runtimeRef.name. Also, the smoke YAML carries an explicit metadata.namespace, while Lines 56-58 query <your-namespace>, which can return nothing if they differ.
Please add an explicit mutation step (download/edit/patch) before apply, or provide per-runtime/per-namespace apply examples.
Suggested doc patch
A shared smoke template applies to any runtime — set `spec.runtimeRef.name` to the runtime you want to exercise:
```bash
-kubectl apply -f $base/trainjob-smoke.yaml
+curl -L "$base/trainjob-smoke.yaml" -o /tmp/trainjob-smoke.yaml
+# edit /tmp/trainjob-smoke.yaml:
+# - metadata.namespace: <your-namespace>
+# - spec.runtimeRef.name: <your-runtime-name>
+kubectl apply -f /tmp/trainjob-smoke.yaml
kubectl -n <your-namespace> get trainjobs
trainjob=$(kubectl -n <your-namespace> get trainjobs -o name | tail -1)
kubectl -n <your-namespace> logs -f -l jobset.sigs.k8s.io/jobset-name=${trainjob##*/}-node</details>
<details>
<summary>🤖 Prompt for AI Agents</summary>
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In @docs/en/training_guides/training-runtimes.mdx around lines 52 - 58, The
smoke-test instructions are incomplete: before applying trainjob-smoke.yaml,
download the remote file (trainjob-smoke.yaml), edit it to set
metadata.namespace to your namespace and spec.runtimeRef.name to the runtime you
want, then kubectl apply the edited file; alternatively show a patch/kubectl -n
apply -f /tmp/trainjob-smoke.yaml example so the subsequent
kubectl -n get trainjobs and the trainjob=$(kubectl -n
...) log command target the same namespace and runtime.
</details>
<!-- fingerprinting:phantom:poseidon:hawk -->
<!-- This is an auto-generated comment by CodeRabbit -->
rspress build failed with `ShikiError: Language `Dockerfile` is not included in this bundle`. Match the previous convention of bare code fences for Containerfile blocks. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Deploying alauda-ai with
|
| Latest commit: |
507cc40
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://194a7a4b.alauda-ai.pages.dev |
| Branch Preview URL: | https://codex-training-guides-reorg.alauda-ai.pages.dev |
Move `mlops_with_coding_agents.mdx` and `coding_agents_with_inference_service.mdx` out of docs/en/model_inference/inference_service/how_to/ into a new top-level docs/en/agentic_mlops/ section, standardizing names to kebab-case. Rewrite every relative link to resolve from the new location: the two pages still reference each other and link out to inference-service how-tos, Envoy AI Gateway, hardware profiles, training guides, LLM compressor, and the MLflow page. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This was referenced Jun 9, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
docs/en/kubeflow/how_to/anddocs/en/workbench/how_to/into a single top-leveldocs/en/training_guides/section, with a landingindex.mdxthat routes users by workflow.mlops_with_coding_agents.mdx,coding_agents_with_inference_service.mdx) out ofdocs/en/model_inference/inference_service/how_to/into a new top-leveldocs/en/agentic_mlops/section.Notes
training-runtimesREADME links to asset paths underdocs/en/kubeflow/how_to/...; those will need a follow-up pointing atdocs/en/training_guides/....llms.txtandllmstxt-state.jsonstill reference the old paths; they regenerate fromllmstxt-config.yamlon the next docs build.Test plan
yarn buildagainst this branch — site builds, thetraining_guidesandagentic_mlopssections appear in the sidebar, every moved page renders.llms.txt/llmstxt-state.json.