Improve cref_sref dataset usage README

README.md

@@ -1,205 +1,258 @@
-# 0426 Lora Triplet Dataset
+---
+pretty_name: "FreeStyle Dataset: cref/sref LoRA Triplets"
+task_categories:
+- image-to-image
+tags:
+- image-to-image
+- style-reference
+- content-reference
+- lora
+- triplet
+- prompt-provenance
+size_categories:
+- 100K<n<1M
+---
 
-This directory contains the normalized export of the lora-triplet portion of:
+# FreeStyle Dataset: cref/sref LoRA Triplets
 
-- `/data/vgo/xingpeng/new_vgo/Sref_Cref_MiniVGO/configs/data/0426_cref_sref_full_diffusion.yaml`
+This repository contains a file-based image triplet dataset for content-reference (`cref`) and style-reference (`sref`) training/evaluation. The exported payload is under [`cref_sref/`](./cref_sref/).
 
-It covers three nonzero-weight lora-triplet sources:
+Each row in a `triplets.csv` file represents one training sequence with three images:
 
-- `cref_sref_qwen_lora_part1`
-- `cref_sref_flux_lora_part1`
-- `cref_sref_illustrious_lora_part1`
+- **content**: image used as the content reference (`cref_0`)
+- **style**: image used as the style reference (`sref_0`)
+- **target**: image for the combined content+style result
 
-## Directory Layout
+The images are deduplicated within each source/role. A single image file can be reused by multiple triplet rows.
 
-For the Hugging Face release, the dataset card stays at repository root and the exported data is placed under `cref_sref/`:
+## Quick Start
 
-```text
-<repo-root>/
-  README.md
-  cref_sref/
-    HF_UPLOAD_CHECKLIST.md
-    README.md
-    qwen/
-    flux/
-    illustrious/
+Install lightweight dependencies:
+
+```bash
+pip install huggingface_hub pandas pillow
 ```
 
-The working export root may also contain `logs/`. That directory is internal and can be excluded from the Hugging Face upload.
+Download and inspect the metadata for one source without downloading all images:
+
+```python
+from huggingface_hub import hf_hub_download
+import pandas as pd
+
+repo_id = "Blue2Giant/FreeStyle_Dataset"
+source = "qwen"  # one of: qwen, flux, illustrious
+
+triplets_csv = hf_hub_download(
+    repo_id=repo_id,
+    repo_type="dataset",
+    filename=f"cref_sref/{source}/triplets.csv",
+)
+triplets = pd.read_csv(triplets_csv)
+
+print(len(triplets))
+print(triplets[[
+    "sequence_id",
+    "content_image_path",
+    "style_image_path",
+    "target_image_path",
+    "content_generation_prompt",
+    "style_generation_prompt",
+    "target_generation_prompt",
+]].head())
+```
 
-Each source subdirectory under `cref_sref/` has the same structure:
+Download the three images for one triplet:
 
-```text
-<source-name>/
-  README.md
-  summary.json
-  triplets.csv
-  content_images.csv
-  style_images.csv
-  target_images.csv
-  images/
-    content/...
-    style/...
-    target/...
-  _state/
-    manifest.json
-    triplets.jsonl
-    content_images.jsonl
-    style_images.jsonl
-    target_images.jsonl
+```python
+from huggingface_hub import hf_hub_download
+from PIL import Image
+import pandas as pd
+
+repo_id = "Blue2Giant/FreeStyle_Dataset"
+source = "qwen"
+
+triplets_csv = hf_hub_download(repo_id=repo_id, repo_type="dataset", filename=f"cref_sref/{source}/triplets.csv")
+triplets = pd.read_csv(triplets_csv)
+row = triplets.iloc[0]
+
+paths = {}
+for role in ["content", "style", "target"]:
+    rel = row[f"{role}_image_path"]  # relative to cref_sref/<source>/
+    paths[role] = hf_hub_download(
+        repo_id=repo_id,
+        repo_type="dataset",
+        filename=f"cref_sref/{source}/{rel}",
+    )
+
+images = {role: Image.open(path).convert("RGB") for role, path in paths.items()}
+print(row["sequence_id"], images["content"].size, images["style"].size, images["target"].size)
 ```
 
-## What A Triplet Means
+## Recommended Download Patterns
 
-Each triplet row corresponds to one vault training sequence and three training images:
+This is a large file-tree dataset. Prefer selective downloads rather than cloning/downloading the entire repository at once.
 
-- `content`: the image used for `cref_0`
-- `style`: the image used for `sref_0`
-- `target`: the image used for the combined content+style target
+Download all CSV/JSON/README metadata, but no images:
 
-So the key relationship is:
+```python
+from huggingface_hub import snapshot_download
 
-- `triplets.csv` = one row per training sequence
-- `content_images.csv` = one row per unique content image
-- `style_images.csv` = one row per unique style image
-- `target_images.csv` = one row per unique target image
+snapshot_dir = snapshot_download(
+    repo_id="Blue2Giant/FreeStyle_Dataset",
+    repo_type="dataset",
+    allow_patterns=[
+        "README.md",
+        "cref_sref/README.md",
+        "cref_sref/*/README.md",
+        "cref_sref/*/summary.json",
+        "cref_sref/*/*.csv",
+    ],
+)
+print(snapshot_dir)
+```
 
-The images are deduplicated. The same exported image path can appear in many triplet rows.
+Download one complete source, including images:
 
-## How To Read The Files
+```python
+from huggingface_hub import snapshot_download
 
-### 1. `triplets.csv`
+source = "qwen"  # qwen is the smallest; flux and illustrious are much larger
+snapshot_dir = snapshot_download(
+    repo_id="Blue2Giant/FreeStyle_Dataset",
+    repo_type="dataset",
+    allow_patterns=["README.md", "cref_sref/README.md", f"cref_sref/{source}/**"],
+)
+print(snapshot_dir)
+```
 
-Use this file when you want to understand the training example itself.
+If you need the full dataset, use `snapshot_download(repo_type="dataset")`, but expect many files.
 
-Important columns:
+## Repository Layout
 
-- `sequence_id`: unique id of the vault sequence
-- `base_model`: one of `qwen`, `flux`, `illustrious`
-- `pair_key`: pair identifier
-- `content_model_id`
-- `style_model_id`
-- `content_image_path`
-- `style_image_path`
-- `target_image_path`
-- `content_original_path`
-- `style_original_path`
-- `target_original_path`
-- `content_match_status`
-- `style_match_status`
-- `target_match_status`
-- `content_prompt_status`
-- `style_prompt_status`
-- `target_prompt_status`
-- `content_generation_prompt`
-- `style_generation_prompt`
-- `target_generation_prompt`
-- `vault_texts_json`
+```text
+<repo-root>/
+  README.md                         # this guide
+  cref_sref/
+    README.md                       # payload-level guide
+    HF_UPLOAD_CHECKLIST.md
+    qwen/
+      README.md
+      summary.json
+      triplets.csv
+      content_images.csv
+      style_images.csv
+      target_images.csv
+      images/
+        content/...
+        style/...
+        target/...
+    flux/
+      ... same layout ...
+    illustrious/
+      ... same layout ...
+```
 
-### 2. `content_images.csv` / `style_images.csv` / `target_images.csv`
+Public upload notes:
 
-Use these files when you want image-level metadata.
+- `_state/` export-resume folders are omitted.
+- `logs/` are omitted.
+- Paths stored in `triplets.csv` are relative to the source directory, e.g. relative to `cref_sref/qwen/`.
 
-Important columns:
+## Sources and Counts
 
-- `exported_image_path`: relative path under the source directory
-- `original_path`: recovered original generation image path when matched
-- `match_status`: whether original-path matching succeeded
-- `prompt_status`: whether the original generation prompt was recovered
-- `generation_prompt`
-- `base_prompt`
-- `sequence_count`: how many triplets reuse this exported image
-- `sequence_ids_json`: which triplets reuse this image
+| Source | Triplet rows | Unique content images | Unique style images | Unique target images |
+|---|---:|---:|---:|---:|
+| `qwen` | 33,582 | 90 | 424 | 5,162 |
+| `flux` | 273,682 | 482 | 13,037 | 129,237 |
+| `illustrious` | 172,589 | 8,045 | 1,581 | 65,952 |
+| **Total** | **479,853** | **8,617** | **15,042** | **200,351** |
 
-## How To View One Triplet
+Total exported image files across source/role directories: **224,010**.
 
-### Method 1: inspect one row from `triplets.csv`
+## File Semantics
 
-```bash
-python3 - <<'PY'
-import csv
-path = '/path/to/repo/cref_sref/qwen/triplets.csv'
-with open(path, 'r', encoding='utf-8', newline='') as fh:
-    row = next(csv.DictReader(fh))
-for key in [
-    'sequence_id',
-    'pair_key',
-    'content_image_path',
-    'style_image_path',
-    'target_image_path',
-    'content_match_status',
-    'style_match_status',
-    'target_match_status',
-    'content_generation_prompt',
-    'style_generation_prompt',
-    'target_generation_prompt',
-]:
-    print(f'{key}: {row[key]}')
-PY
-```
+### `triplets.csv`
 
-### Method 2: load the three images for a given sequence
+Use `triplets.csv` as the main training/evaluation index. Important columns include:
 
-```bash
-python3 - <<'PY'
-import csv
-from pathlib import Path
-
-base = Path('/path/to/repo/cref_sref/qwen')
-with open(base / 'triplets.csv', 'r', encoding='utf-8', newline='') as fh:
-    row = next(csv.DictReader(fh))
-
-print('sequence_id:', row['sequence_id'])
-print('content image:', base / row['content_image_path'])
-print('style image:', base / row['style_image_path'])
-print('target image:', base / row['target_image_path'])
-PY
-```
+- `sequence_id`: unique sequence id
+- `base_model`: one of `qwen`, `flux`, `illustrious`
+- `pair_key`: pair identifier
+- `content_model_id`, `style_model_id`
+- `content_image_path`, `style_image_path`, `target_image_path`
+- `content_original_path`, `style_original_path`, `target_original_path`
+- `content_match_status`, `style_match_status`, `target_match_status`
+- `content_prompt_status`, `style_prompt_status`, `target_prompt_status`
+- `content_generation_prompt`, `style_generation_prompt`, `target_generation_prompt`
+- `vault_texts_json`
+
+For model training, the most common fields are the three `*_image_path` columns and, when available, the three `*_generation_prompt` columns.
+
+### `content_images.csv`, `style_images.csv`, `target_images.csv`
 
-### Method 3: join a triplet row to image-level metadata
+These files are deduplicated image-level metadata tables. Important columns include:
+
+- `exported_image_path`: image path relative to the source directory
+- `original_path`: best-effort recovered original generation image path
+- `match_status`: original-path matching status
+- `prompt_status`: prompt recovery status
+- `generation_prompt`
+- `base_prompt`
+- `sequence_count`: number of triplet rows that reuse this image
+- `sequence_ids_json`: triplet ids that reuse this image
 
-Join:
+Join keys:
 
 - `triplets.csv.content_image_path` -> `content_images.csv.exported_image_path`
 - `triplets.csv.style_image_path` -> `style_images.csv.exported_image_path`
 - `triplets.csv.target_image_path` -> `target_images.csv.exported_image_path`
 
-This lets you answer:
+Example join:
 
-- Which triplets reuse the same image?
-- What is the recovered original path?
-- Was the original prompt recovered?
+```python
+from huggingface_hub import hf_hub_download
+import pandas as pd
 
-## How To Interpret Match And Prompt Status
+repo_id = "Blue2Giant/FreeStyle_Dataset"
+source = "qwen"
 
-### `match_status`
+triplets = pd.read_csv(hf_hub_download(repo_id=repo_id, repo_type="dataset", filename=f"cref_sref/{source}/triplets.csv"))
+content_meta = pd.read_csv(hf_hub_download(repo_id=repo_id, repo_type="dataset", filename=f"cref_sref/{source}/content_images.csv"))
+content_meta = content_meta.set_index("exported_image_path")
 
-- `matched`: exact visual-key match found in the original candidate pool
-- `unmatched`: candidate pool exists, but no exact unique match was found
+row = triplets.iloc[0]
+print(content_meta.loc[row["content_image_path"]])
+```
+
+## Match and Prompt Status Values
+
+`match_status` describes whether the exporter could map an exported vault image back to an original candidate image:
+
+- `matched`: exact visual-key match found in the candidate pool
+- `unmatched`: candidate pool existed, but no exact unique match was found
 - `ambiguous`: more than one candidate matched the same visual key
 - `no_candidates`: no candidate pool was available for that lookup
 
-### `prompt_status`
+`prompt_status` describes whether generation prompt metadata was recovered:
 
-- `resolved`: generation prompt metadata was recovered
+- `resolved`: prompt metadata was recovered
 - `unmatched_original`: original image path was not matched
-- `missing_prompt_payload`: prompt sidecar json was missing
+- `missing_prompt_payload`: prompt sidecar JSON was missing
 - `missing_prompt_entry`: prompt file existed, but the specific image entry was missing
 - `missing_prompt_index`: image filename could not be mapped to a prompt index
 
-## Important Semantics
-
-- Exported images are vault training images, not raw copies of original one-lora or dual-lora PNG files.
-- `original_path` and prompt recovery fields are best-effort provenance fields.
-- Some rows intentionally remain unmatched rather than risk incorrect prompt assignment.
-- `_state/` is internal resume state used during export; it is not required for ordinary dataset consumption.
+## Important Notes
 
-## Source-Level Summary
+- Exported images are vault training images, not guaranteed to be raw copies of the original one-LoRA or dual-LoRA generation files.
+- `original_path` and prompt recovery fields are best-effort provenance fields. Do not assume every row has a resolved prompt.
+- Rows with unresolved provenance are intentionally kept rather than forcing incorrect prompt assignments.
+- The repository is organized as ordinary files. Use `huggingface_hub`, `pandas`, and `Pillow` for flexible access.
 
-Final exported sequence counts:
+## Where to Start
 
-- `qwen`: `33,582`
-- `flux`: `273,682`
-- `illustrious`: `172,589`
+1. Choose a source: start with `qwen` for a smaller first pass.
+2. Read `cref_sref/<source>/README.md` and `summary.json`.
+3. Load `cref_sref/<source>/triplets.csv`.
+4. For each row, resolve image paths relative to `cref_sref/<source>/`.
+5. Optionally join to the deduplicated `*_images.csv` metadata tables for provenance and reuse statistics.
 
-For detailed per-source match counts, see each source's `summary.json`.

cref_sref/README.md

@@ -1,25 +1,75 @@
-# cref_sref
+# `cref_sref` Payload Guide
 
-This directory is the Hugging Face payload for the 0426 lora-triplet release.
+This folder contains the content-reference/style-reference LoRA triplet payload for the FreeStyle dataset.
 
-Contents:
-- `qwen/`
-- `flux/`
-- `illustrious/`
+Start from the root [`README.md`](../README.md) for full usage examples. This file summarizes the payload layout.
 
-Each source directory includes:
-- `README.md`
-- `summary.json`
-- `triplets.csv`
-- `content_images.csv`
-- `style_images.csv`
-- `target_images.csv`
-- `images/`
+## Sources
+
+| Source | Triplet rows | Unique content images | Unique style images | Unique target images |
+|---|---:|---:|---:|---:|
+| `qwen` | 33,582 | 90 | 424 | 5,162 |
+| `flux` | 273,682 | 482 | 13,037 | 129,237 |
+| `illustrious` | 172,589 | 8,045 | 1,581 | 65,952 |
+
+## Layout
+
+Each source directory has the same structure:
+
+```text
+<source>/
+  README.md
+  summary.json
+  triplets.csv
+  content_images.csv
+  style_images.csv
+  target_images.csv
+  images/
+    content/...
+    style/...
+    target/...
+```
+
+`_state/` and `logs/` are intentionally not included in the public upload.
+
+## How to Use
+
+For a source such as `qwen`:
+
+1. Read `qwen/triplets.csv`; it is the sequence-level index.
+2. For each row, use:
+   - `content_image_path`
+   - `style_image_path`
+   - `target_image_path`
+3. Resolve those paths relative to `cref_sref/qwen/`.
+4. Use `content_images.csv`, `style_images.csv`, and `target_images.csv` only when you need image-level metadata, reuse counts, original-path provenance, or prompt recovery details.
+
+Minimal Python example:
+
+```python
+from huggingface_hub import hf_hub_download
+from PIL import Image
+import pandas as pd
+
+repo_id = "Blue2Giant/FreeStyle_Dataset"
+source = "qwen"
+
+triplets = pd.read_csv(hf_hub_download(repo_id=repo_id, repo_type="dataset", filename=f"cref_sref/{source}/triplets.csv"))
+row = triplets.iloc[0]
+
+content_path = hf_hub_download(repo_id=repo_id, repo_type="dataset", filename=f"cref_sref/{source}/{row['content_image_path']}")
+style_path = hf_hub_download(repo_id=repo_id, repo_type="dataset", filename=f"cref_sref/{source}/{row['style_image_path']}")
+target_path = hf_hub_download(repo_id=repo_id, repo_type="dataset", filename=f"cref_sref/{source}/{row['target_image_path']}")
+
+content = Image.open(content_path).convert("RGB")
+style = Image.open(style_path).convert("RGB")
+target = Image.open(target_path).convert("RGB")
+```
+
+## Notes
 
-Notes:
 - `triplets.csv` is sequence-level.
 - `*_images.csv` files are deduplicated image-level metadata.
-- `_state/` is omitted from the public upload.
-- `logs/` are omitted from the public upload.
+- The same image can be reused by many triplets; check `sequence_count` in the image metadata tables.
+- Prompt/provenance fields are best effort. Rows with unresolved provenance are intentionally preserved.
 
-Start with the source-level `README.md` and `triplets.csv`.