Make differentiable Spearman a default dep (diffsort); remove sorting extra

- pyproject: add diffsort>=0.1.0 to dependencies, remove [sorting] optional
- README/Justfile/MAKING_IT_GOOD: no longer mention --extra sorting

Author: Henry Wallace

README.md

@@ -8,7 +8,7 @@ ICF is normalized to \([0, 1]\): **0.0 = very common**, **1.0 = very rare**.
 
 ```bash
 uv sync --extra dev
-# Recommended for multi-task training: uv sync --extra sorting  (torchsort or diffsort for differentiable Spearman; backend is logged at train start)
+# Differentiable Spearman (diffsort) is a default dependency; backend logged at train start
 
 # Train
 uv run tiny-icf-train --help

docs/guides/MAKING_IT_GOOD_MINIMAL_HEURISTICS.md

@@ -20,7 +20,7 @@ Research-backed, low-heuristic improvements. No hand-picked anchor words or ad-h
 
 **Fix:** Use **differentiable Spearman** via soft sorting (Blondel et al., "Fast Differentiable Sorting and Ranking", ICML 2020; [arxiv 2002.08871](https://arxiv.org/abs/2002.08871)). Loss = \( \frac{1}{2}\|r - r_\Psi(\theta)\|^2 \) where \( r_\Psi \) are soft ranks. Implementations: **torchsort** (O(n log n), recommended), **diffsort** (O(n²(log n)²)).
 
-**Implemented:** `loss_unified.spearman_loss_tensor` with `spearman_method="auto"` (default): use **torchsort** if available, else **diffsort**, else rank_relax or built-in soft_rank. All paths are differentiable. Install `uv sync --extra sorting` for torchsort and/or diffsort. At training start we log `Spearman loss backend: <torchsort|diffsort|rank_relax|built-in>`. CLI: `--spearman-reg-strength 0.1`, `--spearman-method auto|torchsort|diffsort|sigmoid`.
+**Implemented:** `loss_unified.spearman_loss_tensor` with `spearman_method="auto"` (default): use **torchsort** if available, else **diffsort** (default dependency), else rank_relax or built-in soft_rank. All paths are differentiable. At training start we log `Spearman loss backend: <torchsort|diffsort|rank_relax|built-in>`. CLI: `--spearman-reg-strength 0.1`, `--spearman-method auto|torchsort|diffsort|sigmoid`.
 
 ---
 

justfile

@@ -64,7 +64,7 @@ sync-s3:
     aws s3 sync models/ s3://arclabs-backups/tiny-icf/models/ --exclude "*" --include "multitask_*.pt" --include "v3_base*.pt" --include "*.pt.cal.json"
 
 # English-only training (better "the"/"and", no lang prefix); uses frequency sampling + spearman-method auto
-# For differentiable Spearman: uv sync --extra sorting (torchsort or diffsort; backend logged at start)
+# Differentiable Spearman (diffsort by default; torchsort if installed); backend logged at start
 # For custom EPOCHS/SAMPLES run: uv run python scripts/train_all_fronts.py ... --epochs N --train-max-samples M
 train-en DATA="data/word_frequency.csv" EPOCHS="30" SAMPLES="200000":
     mkdir -p models/all_fronts_en

pyproject.toml

@@ -10,11 +10,12 @@ dependencies = [
     "numpy>=1.24.0",
     "pandas>=2.0.0",
     "tqdm>=4.65.0",
-    "scipy>=1.10.0", # For correlation metrics
-    "lightning>=2.0.0", # PyTorch Lightning for non-interactive training
-    "aim>=3.29.0", # Experiment tracking
-    "requests>=2.31.0", # For dataset downloading
+    "scipy>=1.10.0",  # For correlation metrics
+    "lightning>=2.0.0",  # PyTorch Lightning for non-interactive training
+    "aim>=3.29.0",  # Experiment tracking
+    "requests>=2.31.0",  # For dataset downloading
     "wordfreq>=3.1.1",
+    "diffsort>=0.1.0",  # Differentiable Spearman (sorting networks); torchsort used if installed
 ]
 
 [project.optional-dependencies]
@@ -31,10 +32,6 @@ dependencies = [
         "sentence-transformers>=2.2.0",  # Lightweight teacher models
         "transformers>=4.30.0",  # For BERT/RoBERTa teacher models (optional)
     ]
-    sorting = [
-        "diffsort>=0.1.0",  # Differentiable sorting networks (ICLR 2022)
-        "torchsort>=0.1.6",  # Fast differentiable sorting (O(n log n), recommended)
-    ]
 
 [project.scripts]
 tiny-icf-train = "tiny_icf.train:main"