# ggormm1/qiren architecture

> Architecture diagram of the GitHub repository ggormm1/qiren, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-11-21). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/ggormm1/qiren
- Repository: https://github.com/ggormm1/qiren

## Overview

QIREN provides image and sound implicit-representation training in qinr, alongside an INR-GAN face-generation stack. The principal workflows are task data/configuration into model training, and GAN training from prepared image datasets through evaluation or generation. The graph also preserves GAN projection, style mixing, and custom training operators. Several GAN capabilities are represented from the file tree and README, but their implementation details were not sampled; unsupported connections are therefore omitted.

Main entry points: [`inr-gan/metrics/metric_main.py`](https://github.com/ggormm1/qiren/blob/main/inr-gan/metrics/metric_main.py) (Metric dispatcher), [`qinr/train.py`](https://github.com/ggormm1/qiren/blob/main/qinr/train.py) (Task training), [`inr-gan/training/training_loop.py`](https://github.com/ggormm1/qiren/blob/main/inr-gan/training/training_loop.py) (Training loop), [`inr-gan/training/networks.py`](https://github.com/ggormm1/qiren/blob/main/inr-gan/training/networks.py) (Generator and discriminator)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_qinr["QIREN representations"]
  node_qtrain["Task training<br/>[train.py]"]
  node_qconfig["Configuration<br/>[config_parser.py]"]
  node_qimage["Image data<br/>[data_process.py]"]
  node_qsound["Sound data<br/>[data_process.py]"]
  node_qmodel["Implicit models<br/>[modules.py]"]
  node_qdiff["Differential operators<br/>[diff_operators.py]"]
  node_qlog["Training logs<br/>[Log.py]"]
end

subgraph group_gan_train["GAN training"]
  node_gtrain["GAN trainer<br/>[train.py]"]
  node_gdata["Dataset loader<br/>[dataset.py]"]
  node_gprep["Dataset preparation<br/>[dataset_tool.py]"]
  node_gloop["Training loop<br/>[training_loop.py]"]
  node_gloss["GAN loss<br/>[loss.py]"]
  node_gstats["Training statistics<br/>[training_stats.py]"]
end

subgraph group_gan_tools["GAN workflows"]
  node_gcalc["Metric command<br/>[calc_metrics.py]"]
  node_ggenerate["Image generation<br/>[generate.py]"]
  node_gstyle["Style mixing<br/>[style_mixing.py]"]
  node_gproject["Latent projection<br/>[projector.py]"]
  node_glegacy["Legacy model loading<br/>[legacy.py]"]
end

subgraph group_gan_metrics["GAN evaluation"]
  node_gmetrics["Metric dispatcher<br/>[metric_main.py]"]
  node_gmetricutil["Metric utilities<br/>[metric_utils.py]"]
  node_gfid["FID metric"]
  node_gkid["KID metric"]
  node_gis["Inception score<br/>[inception_score.py]"]
  node_gpr["Precision recall"]
  node_gppl["Path length"]
end

subgraph group_gan_core["GAN model core"]
  node_gnetwork["Generator and discriminator<br/>[networks.py]"]
  node_glayers["Network layers<br/>[layers.py]"]
  node_gops["Image operations<br/>[upfirdn2d.py]"]
  node_gconv["Convolution gradients<br/>[conv2d_gradfix.py]"]
  node_gpersist["Model persistence<br/>[persistence.py]"]
  node_gcustom["Custom operators<br/>[custom_ops.py]"]
end

node_user(("Researcher"))

node_user -->|"starts training"| node_qtrain
node_qtrain -->|"loads settings"| node_qconfig
node_qtrain -->|"loads image data"| node_qimage
node_qtrain -->|"loads sound data"| node_qsound
node_qtrain -->|"trains models"| node_qmodel
node_qtrain -->|"uses operators"| node_qdiff
node_qtrain -->|"writes logs"| node_qlog
node_user -->|"starts training"| node_gtrain
node_gtrain -->|"runs loop"| node_gloop
node_gtrain -->|"selects metrics"| node_gmetrics
node_gtrain -->|"initializes statistics"| node_gstats
node_gloop -->|"loads dataset"| node_gdata
node_gloop -.->|"trains networks"| node_gnetwork
node_gloop -.->|"uses loss"| node_gloss
node_gloop -->|"reports statistics"| node_gstats
node_gnetwork -->|"uses layers"| node_glayers
node_gnetwork -->|"uses resampling"| node_gops
node_glayers -->|"uses resampling"| node_gops
node_gops -->|"loads kernels"| node_gcustom
node_gloop -->|"calculates metrics"| node_gmetrics
node_gmetrics -->|"uses utilities"| node_gmetricutil
node_gmetrics -->|"dispatches FID"| node_gfid
node_gmetrics -->|"dispatches KID"| node_gkid
node_gmetrics -->|"dispatches score"| node_gis
node_gmetrics -->|"dispatches precision"| node_gpr
node_gmetrics -->|"dispatches path length"| node_gppl
node_gcalc -->|"calculates metrics"| node_gmetrics
node_user -->|"requests evaluation"| node_gcalc
node_user -.->|"requests images"| node_ggenerate
node_user -.->|"requests style mixing"| node_gstyle
node_user -->|"requests projection"| node_gproject
node_gproject -->|"loads network"| node_glegacy
node_glegacy -->|"reconstructs networks"| node_gnetwork
node_gnetwork -->|"decorates models"| node_gpersist
node_gloop -->|"enables gradients"| node_gconv

click node_qtrain "https://github.com/ggormm1/qiren/blob/main/qinr/train.py"
click node_qconfig "https://github.com/ggormm1/qiren/blob/main/qinr/config_parser.py"
click node_qimage "https://github.com/ggormm1/qiren/blob/main/qinr/image/data_process.py"
click node_qsound "https://github.com/ggormm1/qiren/blob/main/qinr/sound/data_process.py"
click node_qmodel "https://github.com/ggormm1/qiren/blob/main/qinr/modules.py"
click node_qdiff "https://github.com/ggormm1/qiren/blob/main/qinr/diff_operators.py"
click node_qlog "https://github.com/ggormm1/qiren/blob/main/qinr/Log.py"
click node_gtrain "https://github.com/ggormm1/qiren/blob/main/inr-gan/train.py"
click node_gdata "https://github.com/ggormm1/qiren/blob/main/inr-gan/training/dataset.py"
click node_gprep "https://github.com/ggormm1/qiren/blob/main/inr-gan/dataset_tool.py"
click node_gloop "https://github.com/ggormm1/qiren/blob/main/inr-gan/training/training_loop.py"
click node_gloss "https://github.com/ggormm1/qiren/blob/main/inr-gan/training/loss.py"
click node_gstats "https://github.com/ggormm1/qiren/blob/main/inr-gan/torch_utils/training_stats.py"
click node_gnetwork "https://github.com/ggormm1/qiren/blob/main/inr-gan/training/networks.py"
click node_glayers "https://github.com/ggormm1/qiren/blob/main/inr-gan/training/layers.py"
click node_gops "https://github.com/ggormm1/qiren/blob/main/inr-gan/torch_utils/ops/upfirdn2d.py"
click node_gconv "https://github.com/ggormm1/qiren/blob/main/inr-gan/torch_utils/ops/conv2d_gradfix.py"
click node_gmetrics "https://github.com/ggormm1/qiren/blob/main/inr-gan/metrics/metric_main.py"
click node_gmetricutil "https://github.com/ggormm1/qiren/blob/main/inr-gan/metrics/metric_utils.py"
click node_gfid "https://github.com/ggormm1/qiren/blob/main/inr-gan/metrics/frechet_inception_distance.py"
click node_gkid "https://github.com/ggormm1/qiren/blob/main/inr-gan/metrics/kernel_inception_distance.py"
click node_gis "https://github.com/ggormm1/qiren/blob/main/inr-gan/metrics/inception_score.py"
click node_gpr "https://github.com/ggormm1/qiren/blob/main/inr-gan/metrics/precision_recall.py"
click node_gppl "https://github.com/ggormm1/qiren/blob/main/inr-gan/metrics/perceptual_path_length.py"
click node_gcalc "https://github.com/ggormm1/qiren/blob/main/inr-gan/calc_metrics.py"
click node_ggenerate "https://github.com/ggormm1/qiren/blob/main/inr-gan/generate.py"
click node_gstyle "https://github.com/ggormm1/qiren/blob/main/inr-gan/style_mixing.py"
click node_gproject "https://github.com/ggormm1/qiren/blob/main/inr-gan/projector.py"
click node_glegacy "https://github.com/ggormm1/qiren/blob/main/inr-gan/legacy.py"
click node_gpersist "https://github.com/ggormm1/qiren/blob/main/inr-gan/torch_utils/persistence.py"
click node_gcustom "https://github.com/ggormm1/qiren/blob/main/inr-gan/torch_utils/custom_ops.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_qtrain,node_qconfig,node_qimage,node_qsound,node_qmodel,node_qdiff,node_qlog toneBlue
class node_gtrain,node_gdata,node_gprep,node_gloop,node_gloss,node_gstats toneAmber
class node_gcalc,node_ggenerate,node_gstyle,node_gproject,node_glegacy toneMint
class node_gmetrics,node_gmetricutil,node_gfid,node_gkid,node_gis,node_gpr,node_gppl toneRose
class node_gnetwork,node_glayers,node_gops,node_gconv,node_gpersist,node_gcustom,node_user toneIndigo
```

## Components

### QIREN representations

- [Task training](https://github.com/ggormm1/qiren/blob/main/qinr/train.py): `qinr/train.py`
- [Configuration](https://github.com/ggormm1/qiren/blob/main/qinr/config_parser.py): `qinr/config_parser.py`
- [Image data](https://github.com/ggormm1/qiren/blob/main/qinr/image/data_process.py): `qinr/image/data_process.py`
- [Sound data](https://github.com/ggormm1/qiren/blob/main/qinr/sound/data_process.py): `qinr/sound/data_process.py`
- [Implicit models](https://github.com/ggormm1/qiren/blob/main/qinr/modules.py): `qinr/modules.py`
- [Differential operators](https://github.com/ggormm1/qiren/blob/main/qinr/diff_operators.py): `qinr/diff_operators.py`
- [Training logs](https://github.com/ggormm1/qiren/blob/main/qinr/Log.py): `qinr/Log.py`

### GAN training

- [GAN trainer](https://github.com/ggormm1/qiren/blob/main/inr-gan/train.py): `inr-gan/train.py`
- [Dataset loader](https://github.com/ggormm1/qiren/blob/main/inr-gan/training/dataset.py): `inr-gan/training/dataset.py`
- [Dataset preparation](https://github.com/ggormm1/qiren/blob/main/inr-gan/dataset_tool.py): `inr-gan/dataset_tool.py`
- [Training loop](https://github.com/ggormm1/qiren/blob/main/inr-gan/training/training_loop.py): `inr-gan/training/training_loop.py`
- [GAN loss](https://github.com/ggormm1/qiren/blob/main/inr-gan/training/loss.py): `inr-gan/training/loss.py`
- [Training statistics](https://github.com/ggormm1/qiren/blob/main/inr-gan/torch_utils/training_stats.py): `inr-gan/torch_utils/training_stats.py`

### GAN workflows

- [Metric command](https://github.com/ggormm1/qiren/blob/main/inr-gan/calc_metrics.py): `inr-gan/calc_metrics.py`
- [Image generation](https://github.com/ggormm1/qiren/blob/main/inr-gan/generate.py): `inr-gan/generate.py`
- [Style mixing](https://github.com/ggormm1/qiren/blob/main/inr-gan/style_mixing.py): `inr-gan/style_mixing.py`
- [Latent projection](https://github.com/ggormm1/qiren/blob/main/inr-gan/projector.py): `inr-gan/projector.py`
- [Legacy model loading](https://github.com/ggormm1/qiren/blob/main/inr-gan/legacy.py): `inr-gan/legacy.py`

### GAN evaluation

- [Metric dispatcher](https://github.com/ggormm1/qiren/blob/main/inr-gan/metrics/metric_main.py): `inr-gan/metrics/metric_main.py`
- [Metric utilities](https://github.com/ggormm1/qiren/blob/main/inr-gan/metrics/metric_utils.py): `inr-gan/metrics/metric_utils.py`
- [FID metric](https://github.com/ggormm1/qiren/blob/main/inr-gan/metrics/frechet_inception_distance.py): `inr-gan/metrics/frechet_inception_distance.py`
- [KID metric](https://github.com/ggormm1/qiren/blob/main/inr-gan/metrics/kernel_inception_distance.py): `inr-gan/metrics/kernel_inception_distance.py`
- [Inception score](https://github.com/ggormm1/qiren/blob/main/inr-gan/metrics/inception_score.py): `inr-gan/metrics/inception_score.py`
- [Precision recall](https://github.com/ggormm1/qiren/blob/main/inr-gan/metrics/precision_recall.py): `inr-gan/metrics/precision_recall.py`
- [Path length](https://github.com/ggormm1/qiren/blob/main/inr-gan/metrics/perceptual_path_length.py): `inr-gan/metrics/perceptual_path_length.py`

### GAN model core

- [Generator and discriminator](https://github.com/ggormm1/qiren/blob/main/inr-gan/training/networks.py): `inr-gan/training/networks.py`
- [Network layers](https://github.com/ggormm1/qiren/blob/main/inr-gan/training/layers.py): `inr-gan/training/layers.py`
- [Image operations](https://github.com/ggormm1/qiren/blob/main/inr-gan/torch_utils/ops/upfirdn2d.py): `inr-gan/torch_utils/ops/upfirdn2d.py`
- [Convolution gradients](https://github.com/ggormm1/qiren/blob/main/inr-gan/torch_utils/ops/conv2d_gradfix.py): `inr-gan/torch_utils/ops/conv2d_gradfix.py`
- [Model persistence](https://github.com/ggormm1/qiren/blob/main/inr-gan/torch_utils/persistence.py): `inr-gan/torch_utils/persistence.py`
- [Custom operators](https://github.com/ggormm1/qiren/blob/main/inr-gan/torch_utils/custom_ops.py): `inr-gan/torch_utils/custom_ops.py`

### Other components

- **Researcher**

## Connections

- Researcher → Task training: starts training (evidence: `qinr/train.py`)
- Task training → Configuration: loads settings (evidence: `qinr/train.py`)
- Task training → Image data: loads image data (evidence: `qinr/train.py`)
- Task training → Sound data: loads sound data (evidence: `qinr/train.py`)
- Task training → Implicit models: trains models (evidence: `qinr/train.py`)
- Task training → Differential operators: uses operators (evidence: `qinr/train.py`)
- Task training → Training logs: writes logs (evidence: `qinr/train.py`)
- Researcher → GAN trainer: starts training (evidence: `inr-gan/train.py`)
- GAN trainer → Training loop: runs loop (evidence: `inr-gan/train.py`)
- GAN trainer → Metric dispatcher: selects metrics (evidence: `inr-gan/train.py`)
- GAN trainer → Training statistics: initializes statistics (evidence: `inr-gan/train.py`)
- Training loop → Dataset loader: loads dataset (evidence: `inr-gan/training/training_loop.py`)
- Training loop → Generator and discriminator: trains networks
- Training loop → GAN loss: uses loss
- Training loop → Training statistics: reports statistics (evidence: `inr-gan/training/training_loop.py`)
- Generator and discriminator → Network layers: uses layers (evidence: `inr-gan/training/networks.py`)
- Generator and discriminator → Image operations: uses resampling (evidence: `inr-gan/training/networks.py`)
- Network layers → Image operations: uses resampling (evidence: `inr-gan/training/layers.py`)
- Image operations → Custom operators: loads kernels (evidence: `inr-gan/torch_utils/ops/upfirdn2d.py`)
- Training loop → Metric dispatcher: calculates metrics (evidence: `inr-gan/training/training_loop.py`)
- Metric dispatcher → Metric utilities: uses utilities (evidence: `inr-gan/metrics/metric_main.py`)
- Metric dispatcher → FID metric: dispatches FID (evidence: `inr-gan/metrics/metric_main.py`)
- Metric dispatcher → KID metric: dispatches KID (evidence: `inr-gan/metrics/metric_main.py`)
- Metric dispatcher → Inception score: dispatches score (evidence: `inr-gan/metrics/metric_main.py`)
- Metric dispatcher → Precision recall: dispatches precision (evidence: `inr-gan/metrics/metric_main.py`)
- Metric dispatcher → Path length: dispatches path length (evidence: `inr-gan/metrics/metric_main.py`)
- Metric command → Metric dispatcher: calculates metrics (evidence: `inr-gan/calc_metrics.py`)
- Researcher → Metric command: requests evaluation (evidence: `inr-gan/calc_metrics.py`)
- Researcher → Image generation: requests images
- Researcher → Style mixing: requests style mixing
- Researcher → Latent projection: requests projection (evidence: `inr-gan/projector.py`)
- Latent projection → Legacy model loading: loads network (evidence: `inr-gan/projector.py`)
- Legacy model loading → Generator and discriminator: reconstructs networks (evidence: `inr-gan/legacy.py`)
- Generator and discriminator → Model persistence: decorates models (evidence: `inr-gan/training/networks.py`)
- Training loop → Convolution gradients: enables gradients (evidence: `inr-gan/training/training_loop.py`)
