nv-generate-mr-brain-finetune

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

NV-Generate-MR-Brain-Finetune

Purpose

用途

Used for finetuning the NV-Generate-CTMR
```
rflow-mr-brain
```
diffusion UNet from user-supplied NIfTI training volumes.
Not for clinical interpretation, regulatory use, or approving synthetic data for production training.
The wrapper stages the config glue locally and delegates execution to existing upstream scripts:
```
scripts.diff_model_create_training_data
```
,
```
scripts.diff_model_train
```
, and optionally
```
scripts.diff_model_infer
```
. It does not execute the notebook.

Manifest I/O: inputs are

datalist

and

data_base_dir

; outputs are

finetuned_checkpoint

, optional

inference_outputs

, and

result_json

The underlying training contract is the upstream config/env JSON (the same one driven from cell
```
[10]
```
of
```
train_diff_unet_tutorial.ipynb
```
). The wrapper stages those JSON files for you and exposes the most-tuned fields as CLI flags; the sections below document the fields, their defaults, and how to monitor/tune a run.

用于基于用户提供的NIfTI训练卷微调NV-Generate-CTMR的
```
rflow-mr-brain
```
扩散UNet模型。
不可用于临床解读、监管用途或批准合成数据用于生产训练。
该包装器会在本地准备配置文件，并将执行任务委托给现有的上游脚本：
```
scripts.diff_model_create_training_data
```
、
```
scripts.diff_model_train
```
，以及可选的
```
scripts.diff_model_infer
```
。它不会运行笔记本文件。

清单输入输出：输入为

datalist

和

data_base_dir

；输出为

finetuned_checkpoint

、可选的

inference_outputs

和

result_json

。

底层训练协议为上游配置/环境JSON（与
```
train_diff_unet_tutorial.ipynb
```
的单元格
```
[10]
```
所使用的相同）。包装器会为您准备这些JSON文件，并将最常用的配置字段作为CLI标志暴露出来；以下部分将记录这些字段、默认值以及如何监控和调优运行过程。

Instructions

操作说明

Read
```
skill_manifest.yaml
```
before changing arguments, side effects, or validation gates.
Run
```
scripts/run_mr_brain_finetune.py
```
from the Medical AI Skills repo root.
If a host agent exposes
```
run_script
```
, use
```
run_script("scripts/run_mr_brain_finetune.py", args=[...])
```
; otherwise run the Bash/Python command below.
Use
```
--preflight
```
first when checking a new datalist; remove
```
--preflight
```
only when the user explicitly wants to launch GPU finetuning.
For a staged preflight input bundle directory, use
```
BUNDLE/preflight_datalist.json
```
as the datalist and
```
BUNDLE/preflight_dataset
```
as
```
--data-base-dir
```
when those files are present.

在修改参数、副作用或验证规则前，请先阅读
```
skill_manifest.yaml
```
。
从Medical AI Skills仓库根目录运行
```
scripts/run_mr_brain_finetune.py
```
。
如果主机Agent暴露了
```
run_script
```
接口，请使用
```
run_script("scripts/run_mr_brain_finetune.py", args=[...])
```
；否则运行下方的Bash/Python命令。
检查新数据列表时，请先使用
```
--preflight
```
参数；仅当用户明确要启动GPU微调时，再移除该参数。
对于已准备好的预校验输入包目录，当
```
BUNDLE/preflight_datalist.json
```
和
```
BUNDLE/preflight_dataset
```
文件存在时，使用前者作为数据列表，后者作为
```
--data-base-dir
```
。

Examples

示例

Validate and stage a preflight finetune check from an input bundle (the recommended first step — no GPU, no training). This is the single canonical command; replace

INPUT_BUNDLE

and

OUT_DIR

with your paths:

bash

export NV_GENERATE_ROOT="${NV_GENERATE_ROOT:-.workbench_data/upstreams/NV-Generate-CTMR}" && \
python skills/nv-generate-mr-brain-finetune/scripts/run_mr_brain_finetune.py \
  INPUT_BUNDLE/preflight_datalist.json \
  --data-base-dir INPUT_BUNDLE/preflight_dataset \
  --output-dir OUT_DIR \
  --modality mri_t1 \
  --preflight

For real GPU finetuning and other variations, see Usage below.

从输入包验证并准备预校验微调检查（推荐的第一步——无需GPU，不进行训练）。这是标准命令；请将

INPUT_BUNDLE

和

OUT_DIR

替换为您的路径：

bash

export NV_GENERATE_ROOT="${NV_GENERATE_ROOT:-.workbench_data/upstreams/NV-Generate-CTMR}" && \
python skills/nv-generate-mr-brain-finetune/scripts/run_mr_brain_finetune.py \
  INPUT_BUNDLE/preflight_datalist.json \
  --data-base-dir INPUT_BUNDLE/preflight_dataset \
  --output-dir OUT_DIR \
  --modality mri_t1 \
  --preflight

如需进行实际GPU微调及其他变体操作，请查看下方的【2. 使用方式（一键训练）】部分。

Available Scripts

可用脚本

Script Purpose Arguments

Script	Purpose	Arguments
`scripts/run_mr_brain_finetune.py`	Primary entrypoint declared by `skill_manifest.yaml` .	`DATALIST.json --data-base-dir DATA_DIR --output-dir OUT_DIR [--epochs N] [--modality mri_t1] [--num-gpus N] [--no-amp] [--model-config FILE] [--run-inference] [--preflight]`

scripts/run_mr_brain_finetune.py

Primary entrypoint declared by

skill_manifest.yaml

DATALIST.json --data-base-dir DATA_DIR --output-dir OUT_DIR [--epochs N] [--modality mri_t1] [--num-gpus N] [--no-amp] [--model-config FILE] [--run-inference] [--preflight]

脚本用途参数

脚本	用途	参数
`scripts/run_mr_brain_finetune.py`	`skill_manifest.yaml` 声明的主入口脚本	`DATALIST.json --data-base-dir DATA_DIR --output-dir OUT_DIR [--epochs N] [--modality mri_t1] [--num-gpus N] [--no-amp] [--model-config FILE] [--run-inference] [--preflight]`

scripts/run_mr_brain_finetune.py

skill_manifest.yaml

声明的主入口脚本

DATALIST.json --data-base-dir DATA_DIR --output-dir OUT_DIR [--epochs N] [--modality mri_t1] [--num-gpus N] [--no-amp] [--model-config FILE] [--run-inference] [--preflight]

Prerequisites

前置条件

NV_GENERATE_ROOT

may point to a current checkout of

https://github.com/NVIDIA-Medtech/NV-Generate-CTMR

containing

scripts/diff_model_create_training_data.py

scripts/diff_model_train.py

, and

scripts/diff_model_infer.py

NV_GENERATE_ROOT

is unset, the wrapper searches

.workbench_data/upstreams/NV-Generate-CTMR

```
CUDA_VISIBLE_DEVICES
```
is optional and can be used to select the GPU for real training.
Runtime requirements: NVIDIA CUDA GPU for real training, Python packages from the upstream
```
requirements.txt
```
, and downloaded MR-brain weights.
Side effects: writes staged configs, embeddings, checkpoints, optional inference images, and logs under the caller-provided
```
--output-dir
```
; may write model caches under the upstream checkout and
```
~/.cache/huggingface/
```
; may contact
```
https://huggingface.co
```
for model assets and
```
https://github.com
```
for the upstream checkout.
The datalist is a MONAI-style JSON object with
```
training[].image
```
paths relative to
```
--data-base-dir
```
.
```
training[].modality
```
is optional and defaults to
```
mri_t1
```
.

NV_GENERATE_ROOT

需指向

https://github.com/NVIDIA-Medtech/NV-Generate-CTMR

的当前检出版本，其中需包含

scripts/diff_model_create_training_data.py

、

scripts/diff_model_train.py

和

scripts/diff_model_infer.py

。

如果未设置

NV_GENERATE_ROOT

，包装器会搜索

.workbench_data/upstreams/NV-Generate-CTMR

。

```
CUDA_VISIBLE_DEVICES
```
为可选参数，可用于选择实际训练时使用的GPU。
运行环境要求：实际训练需NVIDIA CUDA GPU，需安装上游
```
requirements.txt
```
中的Python包，且需下载脑部MR模型权重。
副作用：会在用户指定的
```
--output-dir
```
下写入准备好的配置文件、嵌入向量、检查点、可选的推理图像和日志；可能会在上游检出目录和
```
~/.cache/huggingface/
```
下写入模型缓存；可能会连接
```
https://huggingface.co
```
获取模型资源，连接
```
https://github.com
```
获取上游检出版本。
数据列表为MONAI风格的JSON对象，其中
```
training[].image
```
路径为相对于
```
--data-base-dir
```
的路径。
```
training[].modality
```
为可选参数，默认值为
```
mri_t1
```
。

1. Config and environment JSON (adapt to your data)

1. 配置与环境JSON（适配您的数据）

This is a thin wrapper around the upstream

train_diff_unet_tutorial.ipynb

flow. Each run performs four steps, delegating the heavy lifting to the model author's scripts:

Stage configs — copy the three config JSONs and rewrite only the run-specific paths and
```
n_epochs
```
(notebook cell 15).

python -m scripts.diff_model_create_training_data

→ latent

*_emb.nii.gz

embeddings (cell 17).

Write embedding sidecars — a
```
<emb>.nii.gz.json
```
per embedding with
```
spacing
```
/
```
modality
```
(and body-region indices when the model uses them). This is the one piece of glue that lives in the notebook (cell 19), not in upstream
```
scripts/
```
, and
```
diff_model_train
```
requires it; the skill owns it.

python -m scripts.diff_model_train

(cell 21), optionally

python -m scripts.diff_model_infer

Tune by editing the config JSON, not by adding flags. All training/inference hyperparameters (

lr

batch_size

cache_rate

, inference

dim

spacing

num_inference_steps

cfg_guidance_scale

, …) live in

config_maisi_diff_model_rflow-mr-brain.json

. Edit the upstream copy, or pass your own with

--model-config FILE

(and

--env-config

--model-def

for the other two). The wrapper only ever rewrites the fields below.

Environment JSON (

environment_maisi_diff_model_rflow-mr-brain.json

) — fields the wrapper rewrites per run:

Field	Set from	Notes
`data_base_dir`	`--data-base-dir`	Root for relative `training[].image` paths.
`json_data_list`	your datalist	Staged copy with per-entry `modality` filled in.
`embedding_base_dir` , `model_dir` , `output_dir`	`--output-dir`	Latent embeddings, checkpoints, inference images.
`modality_mapping_path`	upstream	Maps modality name → integer code.
`model_filename`	`--model-filename`	Output checkpoint name (default `diff_unet_3d_rflow-mr-brain_v0.pt` ).
`existing_ckpt_filepath`	upstream weights / `--existing-ckpt-filepath`	Starting checkpoint; cleared by `--train-from-scratch` .
`trained_autoencoder_path`	upstream weights / `--trained-autoencoder-path`	VAE used to encode/decode latents.

Model config (

config_maisi_diff_model_rflow-mr-brain.json

) — the only fields the wrapper touches:

Field	Set from	Default	Notes
`diffusion_unet_train.n_epochs`	`--epochs`	`2` (upstream config ships `1000` )	Convenience override (cell 15 does the same); wrapper default is small for verification.
`diffusion_unet_inference.modality`	`--modality`	from `modality_mapping.json`	Kept consistent with the training modality for optional `--run-inference` .

Everything else in that file (

lr

batch_size

cache_rate

, the rest of

diffusion_unet_inference

) is left exactly as written — edit the JSON to change it.

Runtime flags (not config fields):

--num-gpus N

(

>1

launches

torch.distributed.run

--no-amp

(disable mixed precision, passed through to

diff_model_train

--modality

selects the integer code from

configs/modality_mapping.json

. Supported brain values:

mri

(8),

mri_t1

(9, default),

mri_t2

(10),

mri_flair

(11),

mri_swi

(20), and their

*_skull_stripped

variants (29/30/31/32). Per-case

training[].modality

overrides

--modality

. The modality also feeds the step-3 embedding sidecars.

For an end-to-end reference including example data download and checkpoint loading, see the upstream tutorial

train_diff_unet_tutorial.ipynb

这是上游

train_diff_unet_tutorial.ipynb

流程的轻量包装器。每次运行会执行四个步骤，将核心任务委托给模型作者的脚本：

准备配置文件 —— 复制三个配置JSON文件，仅重写与运行相关的路径和
```
n_epochs
```
（对应笔记本单元格15）。

python -m scripts.diff_model_create_training_data

→ 生成潜在

*_emb.nii.gz

嵌入向量（对应单元格17）。

写入嵌入向量附属文件 —— 每个嵌入向量对应一个
```
<emb>.nii.gz.json
```
文件，包含
```
spacing
```
/
```
modality
```
（当模型使用体区域索引时还包含该索引）。这是笔记本中（单元格19）而非上游
```
scripts/
```
中的唯一衔接逻辑，且
```
diff_model_train
```
需要该文件；本Skill负责处理此步骤。

python -m scripts.diff_model_train

（对应单元格21），可选执行

python -m scripts.diff_model_infer

。

通过编辑配置JSON进行调优，而非添加标志。 所有训练/推理超参数（

lr

、

batch_size

、

cache_rate

、推理

dim

spacing

num_inference_steps

cfg_guidance_scale

等）均存储在

config_maisi_diff_model_rflow-mr-brain.json

中。编辑上游副本，或通过

--model-config FILE

（以及

--env-config

--model-def

用于另外两个配置文件）传入您自己的配置。包装器仅会重写以下字段。

环境JSON（

environment_maisi_diff_model_rflow-mr-brain.json

）——包装器会针对每次运行重写以下字段：

字段	来源	说明
`data_base_dir`	`--data-base-dir`	`training[].image` 相对路径的根目录。
`json_data_list`	您的数据列表	已准备好的副本，其中填充了每个条目的 `modality` 。
`embedding_base_dir` , `model_dir` , `output_dir`	`--output-dir`	潜在嵌入向量、检查点、推理图像的存储目录。
`modality_mapping_path`	上游文件	映射模态名称→整数编码。
`model_filename`	`--model-filename`	输出检查点名称（默认值为 `diff_unet_3d_rflow-mr-brain_v0.pt` ）。
`existing_ckpt_filepath`	上游权重 / `--existing-ckpt-filepath`	初始检查点；使用 `--train-from-scratch` 会清空该字段。
`trained_autoencoder_path`	上游权重 / `--trained-autoencoder-path`	用于编码/解码潜在向量的VAE模型。

模型配置（

config_maisi_diff_model_rflow-mr-brain.json

）——包装器仅会修改以下字段：

字段	来源	默认值	说明
`diffusion_unet_train.n_epochs`	`--epochs`	`2` （上游配置默认值为 `1000` ）	便捷覆盖参数（单元格15也会执行相同操作）；包装器默认值较小，用于验证。
`diffusion_unet_inference.modality`	`--modality`	来自 `modality_mapping.json`	与训练模态保持一致，用于可选的 `--run-inference` 操作。

该文件中的其他所有内容（

lr

、

batch_size

、

cache_rate

、

diffusion_unet_inference

的其余参数）均保持原样——如需修改请编辑JSON文件。

运行时标志（非配置字段）：

--num-gpus N

（

>1

时会启动

torch.distributed.run

）、

--no-amp

（禁用混合精度，会传递给

diff_model_train

）。

--modality

会从

configs/modality_mapping.json

中选择整数编码。支持的脑部模态值包括：

mri

（8）、

mri_t1

（9，默认值）、

mri_t2

（10）、

mri_flair

（11）、

mri_swi

（20），以及它们的

*_skull_stripped

变体（29/30/31/32）。每个案例的

training[].modality

会覆盖

--modality

参数。模态信息还会用于第三步的嵌入向量附属文件。

如需包含示例数据下载和检查点加载的端到端参考，请查看上游教程

train_diff_unet_tutorial.ipynb

。

2. Usage (one-line training)

2. 使用方式（一键训练）

Preflight only:

bash

export NV_GENERATE_ROOT="${NV_GENERATE_ROOT:-.workbench_data/upstreams/NV-Generate-CTMR}" && \
python skills/nv-generate-mr-brain-finetune/scripts/run_mr_brain_finetune.py \
  PATH_TO_DATALIST.json \
  --data-base-dir PATH_TO_DATA_ROOT \
  --output-dir runs/nv_generate_mr_brain_finetune_preflight \
  --preflight

Preflight bundle input:

bash

export NV_GENERATE_ROOT="${NV_GENERATE_ROOT:-.workbench_data/upstreams/NV-Generate-CTMR}" && \
python skills/nv-generate-mr-brain-finetune/scripts/run_mr_brain_finetune.py \
  PATH_TO_INPUT_BUNDLE/preflight_datalist.json \
  --data-base-dir PATH_TO_INPUT_BUNDLE/preflight_dataset \
  --output-dir runs/nv_generate_mr_brain_finetune_preflight \
  --preflight

GPU finetuning:

bash

export NV_GENERATE_ROOT="${NV_GENERATE_ROOT:-.workbench_data/upstreams/NV-Generate-CTMR}" && \
python -m pip install -r "$NV_GENERATE_ROOT/requirements.txt" && \
python skills/nv-generate-mr-brain-finetune/scripts/run_mr_brain_finetune.py \
  PATH_TO_DATALIST.json \
  --data-base-dir PATH_TO_DATA_ROOT \
  --output-dir runs/nv_generate_mr_brain_finetune \
  --epochs 2 \
  --modality mri_t1 \
  --run-inference

Replace

PATH_TO_DATALIST.json

and

PATH_TO_DATA_ROOT

with the user's actual paths. Do not use the fixture datalist for real training; it is a preflight-only placeholder.

仅预校验：

bash

export NV_GENERATE_ROOT="${NV_GENERATE_ROOT:-.workbench_data/upstreams/NV-Generate-CTMR}" && \
python skills/nv-generate-mr-brain-finetune/scripts/run_mr_brain_finetune.py \
  PATH_TO_DATALIST.json \
  --data-base-dir PATH_TO_DATA_ROOT \
  --output-dir runs/nv_generate_mr_brain_finetune_preflight \
  --preflight

预校验包输入：

bash

export NV_GENERATE_ROOT="${NV_GENERATE_ROOT:-.workbench_data/upstreams/NV-Generate-CTMR}" && \
python skills/nv-generate-mr-brain-finetune/scripts/run_mr_brain_finetune.py \
  PATH_TO_INPUT_BUNDLE/preflight_datalist.json \
  --data-base-dir PATH_TO_INPUT_BUNDLE/preflight_dataset \
  --output-dir runs/nv_generate_mr_brain_finetune_preflight \
  --preflight

GPU微调：

bash

export NV_GENERATE_ROOT="${NV_GENERATE_ROOT:-.workbench_data/upstreams/NV-Generate-CTMR}" && \
python -m pip install -r "$NV_GENERATE_ROOT/requirements.txt" && \
python skills/nv-generate-mr-brain-finetune/scripts/run_mr_brain_finetune.py \
  PATH_TO_DATALIST.json \
  --data-base-dir PATH_TO_DATA_ROOT \
  --output-dir runs/nv_generate_mr_brain_finetune \
  --epochs 2 \
  --modality mri_t1 \
  --run-inference

请将

PATH_TO_DATALIST.json

和

PATH_TO_DATA_ROOT

替换为用户的实际路径。请勿使用测试数据列表进行实际训练；它仅作为预校验占位符。

3. Monitor training (TensorBoard)

3. 监控训练（TensorBoard）

scripts.diff_model_train

writes TensorBoard event files under the staged

model_dir

(

OUT_DIR/artifacts/models

). Launch TensorBoard against the output directory and watch the loss curve:

bash

python -m pip install tensorboard && \
tensorboard --logdir runs/nv_generate_mr_brain_finetune/artifacts

The run summary is written to

OUT_DIR/artifacts/workflow_summary.json

(checkpoint path, embedding sidecars, inference outputs); the JSON the wrapper prints to stdout mirrors the same paths plus

exit_code

and a

stderr_tail

for quick triage.

scripts.diff_model_train

会在准备好的

model_dir

（

OUT_DIR/artifacts/models

）下写入TensorBoard事件文件。针对输出目录启动TensorBoard并查看损失曲线：

bash

python -m pip install tensorboard && \
tensorboard --logdir runs/nv_generate_mr_brain_finetune/artifacts

运行摘要会写入

OUT_DIR/artifacts/workflow_summary.json

（包含检查点路径、嵌入向量附属文件、推理输出）；包装器打印到标准输出的JSON会镜像这些路径，同时包含

exit_code

和

stderr_tail

用于快速排查问题。

4. Hyperparameter tuning and common pitfalls

4. 超参数调优与常见问题

Loss not decreasing / unstable — lower
```
diffusion_unet_train.lr
```
(default
```
1e-5
```
) in the model-config JSON, or keep AMP on (default);
```
--no-amp
```
is slower but more numerically stable on older GPUs.
Out-of-memory — keep
```
diffusion_unet_train.batch_size
```
at
```
1
```
and
```
cache_rate
```
at
```
0
```
in the config JSON, and confirm the autoencoder/UNet fit your GPU before scaling. Multi-GPU (
```
--num-gpus N
```
) shards the batch via
```
torch.distributed.run
```
.
Few cases / quick check — keep
```
--epochs
```
small (the wrapper default
```
2
```
is for verification, not convergence; the upstream config ships
```
1000
```
).
Wrong modality conditioning — set
```
--modality
```
or per-case
```
training[].modality
```
to a value present in
```
configs/modality_mapping.json
```
; a mismatch produces a clear error rather than silently mislabeling latents.
Slow startup on first run —
```
diff_model_create_training_data
```
precomputes latent embeddings once; reuse the same
```
--output-dir
```
to avoid recomputing them.

损失未下降/不稳定 —— 在模型配置JSON中降低
```
diffusion_unet_train.lr
```
（默认值
```
1e-5
```
），或保持AMP启用（默认设置）；
```
--no-amp
```
速度较慢，但在旧GPU上数值稳定性更高。
内存不足 —— 在配置JSON中保持
```
diffusion_unet_train.batch_size
```
为
```
1
```
、
```
cache_rate
```
为
```
0
```
，并在扩展前确认自动编码器/UNet适配您的GPU。多GPU（
```
--num-gpus N
```
）会通过
```
torch.distributed.run
```
拆分批次。
案例数量少/快速检查 —— 保持
```
--epochs
```
较小（包装器默认值
```
2
```
用于验证，而非收敛；上游配置默认值为
```
1000
```
）。
模态条件错误 —— 将
```
--modality
```
或每个案例的
```
training[].modality
```
设置为
```
configs/modality_mapping.json
```
中存在的值；不匹配会产生明确错误，而非静默标记错误的潜在向量。
首次启动速度慢 ——
```
diff_model_create_training_data
```
会预先计算一次潜在嵌入向量；重复使用相同的
```
--output-dir
```
可避免重新计算。

5. Evaluate the finetuned model

5. 评估微调后的模型

Use the staged checkpoint (

OUT_DIR/artifacts/models/<model_filename>

) as the diffusion UNet for generation, then inspect the synthesized volumes:

Pass
```
--run-inference
```
here for a quick built-in sanity render, or
Point the
```
nv-generate-mr-brain
```
inference skill at the finetuned checkpoint to generate fresh brain MRI volumes for qualitative review.

This skill gates file accounting and command provenance only — anatomical realism and downstream utility must be judged by a domain expert on the generated images.

使用准备好的检查点（

OUT_DIR/artifacts/models/<model_filename>

）作为扩散UNet进行生成，然后检查合成的卷数据：

在此处使用
```
--run-inference
```
进行快速内置的合理性渲染，或
将
```
nv-generate-mr-brain
```
推理Skill指向微调后的检查点，生成新的脑部MRI卷用于定性评估。

本Skill仅负责文件管理和命令溯源——生成图像的解剖学真实性和下游效用必须由领域专家判断。

Limitations

局限性

Requires a current upstream
```
NV-Generate-CTMR
```
checkout with the existing diffusion training scripts. The skill itself stages the required config and datalist glue locally and does not depend on the notebook or PR #33.
Full training can be expensive and is not deterministic across hardware, CUDA, and package versions.
The wrapper gates file accounting and command provenance, not anatomical realism or downstream model utility.
Not for clinical deployment, clinical interpretation, autonomous diagnosis, regulatory submission, or production training-data approval.

需要当前上游
```
NV-Generate-CTMR
```
检出版本及现有的扩散训练脚本。本Skill本身会在本地准备所需的配置和数据列表衔接逻辑，不依赖笔记本或PR #33。
完整训练成本较高，且在不同硬件、CUDA和包版本下结果不具有确定性。
包装器仅负责文件管理和命令溯源，不保证解剖学真实性或下游模型效用。
不可用于临床部署、临床解读、自主诊断、监管提交或生产训练数据审批。

Troubleshooting

故障排除

Error	Cause	Fix
`diffusion training scripts were not found`	`NV_GENERATE_ROOT` does not point at a current NV-Generate-CTMR checkout.	Clone or update `https://github.com/NVIDIA-Medtech/NV-Generate-CTMR` and set `NV_GENERATE_ROOT` .
`missing datalist image`	`training[].image` paths are not relative to `--data-base-dir` or files are absent.	Fix the datalist or pass the correct data root.
CUDA or MONAI import failure	Runtime environment lacks upstream dependencies.	Install `"$NV_GENERATE_ROOT/requirements.txt"` in the selected environment.

错误	原因	解决方法
`diffusion training scripts were not found`	`NV_GENERATE_ROOT` 未指向当前NV-Generate-CTMR检出版本。	克隆或更新 `https://github.com/NVIDIA-Medtech/NV-Generate-CTMR` 并设置 `NV_GENERATE_ROOT` 。
`missing datalist image`	`training[].image` 路径不相对于 `--data-base-dir` 或文件不存在。	修正数据列表或传入正确的数据根目录。
CUDA或MONAI导入失败	运行环境缺少上游依赖。	在所选环境中安装 `"$NV_GENERATE_ROOT/requirements.txt"` 中的依赖。