restructuring and updated readme.

README.md

@@ -13,14 +13,30 @@ task_categories:
   - image-text-to-text
 pretty_name: "Gomoku VLM Dataset (LoRA finetuning)"
 configs:
-  - config_name: train_basic_visual_strategy_split
+  - config_name: visual
     data_files:
       - split: train
-        path: "train/*.parquet"
-  - config_name: eval
+        path: "visual/train/*.parquet"
+      - split: validation
+        path: "visual/eval/*.parquet"
+
+  - config_name: strategy
     data_files:
+      - split: train
+        path: "strategy/train/*.parquet"
       - split: validation
-        path: "eval/*.parquet"
+        path: "strategy/eval/*.parquet"
+
+  # Curriculum
+  - config_name: visual_curriculum
+    data_files:
+      - split: train
+        path:
+          - "curriculum/step_1.parquet"
+          - "curriculum/step_2.parquet"
+          - "curriculum/step_3.parquet"
+          - "curriculum/step_4.parquet"
+
   - config_name: test
     data_files:
       - split: test
@@ -30,10 +46,16 @@ configs:
 # Gomoku VLM Dataset (LoRA finetuning)
 
 This repository contains a synthetic, image-grounded instruction dataset for training and evaluating **vision-language models (VLMs)** on **Gomoku (15×15)**.  
-The dataset is designed for **LoRA finetuning** of image-text-to-text models (e.g. google/gemma-3-4b-it style) on two main skill families:
+The dataset is designed for **LoRA finetuning** of image-text-to-text vision-language models on **two complementary capabilities**:
+
+- **Visual**  
+  Tasks where the model must read the board image and produce a structured answer about the current position.  
+  This includes purely perceptual objectives (cell classification, counting) and also *visually grounded reasoning* such as run/line detection, matrix reconstruction, end-state recognition, and yes/no tactical assessments that can be decided from the current snapshot (e.g., “immediate win exists”, “opponent threatens immediate win”).  
+  
+- **Curriculum:** Curriculum-learning variant for *visual skills*: the training data is split into four steps that progressively move from simpler to more complex board states and visually grounded objectives (e.g., from basic cell/count tasks toward more advanced structured board understanding)
 
-- **Perception**: read board state from an image and answer factual questions (counts, locations, piece colors, etc.)
-- **Strategy**: answer tactical/strategic questions (win-in-1, recommended moves, etc.)
+- **Strategy / policy (action selection)**  
+  Tasks that require choosing an action (e.g., best move / win-in-1 move selection) and decision-making that approximates a bot’s policy.
 
 Each example includes:
 1) a rendered board image,  
@@ -44,19 +66,68 @@ Each example includes:
 
 ## Dataset structure
 
-This dataset is published as multiple *configs* (subsets) to support different training regimes:
+This dataset is organized into multiple Hugging Face **configs** that mirror the repository folders:
+
+## Configs
+
+- **`visual`**  
+  Perception-focused questions (board reading, counting, localization, etc.).  
+  Splits:
+  - `train` → `visual/train/*.parquet`
+  - `validation` → `visual/eval/*.parquet`
 
-- `train_basic_visual_strategy_split`: main training subset (split into visual + strategy)
-- `train_curriculum_strategy`: curriculum-style strategy training (sharded in steps)
-- `train_curriculum_visual`: curriculum-style perception/visual training (sharded in steps)
-- `eval`: validation set
-- `eval_reduced`: smaller validation set for faster iteration
-- `test`: test set
+- **`strategy`**  
+  Tactical / strategic questions (e.g., win-in-1 style tasks, move selection based on bot-policy).  
+  Splits:
+  - `train` → `strategy/train/*.parquet`
+  - `validation` → `strategy/eval/*.parquet`
 
-You can select a config using Hugging Face `datasets`:
+- **`visual_curriculum`**  
+  Step-wise curriculum training data as **four growing steps**:
+  - `curriculum/step_1.parquet`
+  - `curriculum/step_2.parquet`
+  - `curriculum/step_3.parquet`
+  - `curriculum/step_4.parquet`
+
+- **`test`**  
+  Test Dataset:
+  - `test` → `test/combined.parquet`
+
+## Loading the dataset
 
 ```python
 from datasets import load_dataset
 
-ds = load_dataset("eganscha/gomoku_vlm_ds", "train_basic_visual_strategy_split")
-print(ds)
+# Visual / perception
+ds_visual = load_dataset("eganscha/gomoku_vlm_ds", "visual")
+print(ds_visual)
+
+# Strategy
+ds_strategy = load_dataset("eganscha/gomoku_vlm_ds", "strategy")
+print(ds_strategy)
+
+# Curriculum steps (visual curriculum)
+ds_curr = load_dataset("eganscha/gomoku_vlm_ds", "visual_curriculum")
+print(ds_curr)
+
+# Test
+ds_test = load_dataset("eganscha/gomoku_vlm_ds", "test")
+print(ds_test)
+```
+
+## Downloading the dataset locally
+
+### Make sure the hf CLI is installed
+```bash
+curl -LsSf https://hf.co/cli/install.sh | bash
+```
+
+### Source bashrc
+```bash
+source ~/.bashrc
+```
+
+### Download Dataset
+```bash
+hf download eganscha/gomoku_vlm_ds --repo-type=dataset --local-dir ./gomoku_vlm_ds
+```

main_ds/eval/strategy.parquet → curriculum/step_1.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ee8fa40df2c83402cc472843708a9d797c85af6b138bf2cb1170aa0f05596d32
-size 16528890
+oid sha256:d8a167b6eb22e59e77a9c5605767849d0ecc775680fecd3b44cee5d43b62883b
+size 51472120

main_ds/train/strategy.parquet → curriculum/step_2.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e9ebb5b2b0dc7d1f1efe758fe8a108e1fdf152fe2a8b690dd61abc4b18c08fb8
-size 236815969
+oid sha256:d970880fe81b496457d7e9d4ed7011d39438db352fa83ecccafb913d2ab9a154
+size 160992477

rehearsal_ds/eval/visual.parquet → curriculum/step_3.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:93e0a685796604f77b99d4f03c6052df3ca656fba56e3516c6f3ee2e914a3dd8
-size 64300392
+oid sha256:f9aee8836e8e330a75e014976bd7a20de146e18cfd5023a05c0ba5abc3e0283f
+size 247712785

rehearsal_ds/train/visual.parquet → curriculum/step_4.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a63a8a284657dfa702c92179c9397347b4d669a917a63aa0b05fc82e351c466a
-size 804581284
+oid sha256:453787ae10ec65066952a6631dfc7dcfad038405d322598cd44472251bad7267
+size 401155761