Clean up fix/osd-decoder-improvements branch (resolves #61)

.gitignore

@@ -52,3 +52,5 @@ docs/*.typ
 # Dataset-specific
 *.npz.tmp
 *.stim.bak
+
+/.claude

DECODER_CONFIG.md

@@ -0,0 +1,70 @@
+# BP+OSD Decoder Configuration for p=0.001 Target
+
+## Changes Made
+
+### 1. Dataset Update
+- **Changed from**: `sc_d3_r3_p0010_z` (p=0.010)
+- **Changed to**: `sc_d3_r3_p0001_z` (p=0.001)
+- Dataset contains 10,000 syndrome samples
+
+### 2. BP Iterations Increased
+- **Changed from**: `BP_MAX_ITER = 20`
+- **Changed to**: `BP_MAX_ITER = 50`
+- Rationale: More iterations needed for convergence at lower error rates
+
+### 3. OSD Order
+- **Maintained**: `OSD_ORDER = 15` (searches 2^15 = 32,768 candidates)
+- This provides good balance between accuracy and runtime
+
+### 4. Test Configuration
+- **Samples**: 1000 (for reasonable runtime ~1.6 hours)
+- **Batch size**: 50
+- **Device**: CUDA (GPU acceleration enabled)
+
+## Target Performance
+
+According to paper http://arxiv.org/abs/2005.07016:
+- **Target logical error rate**: 0.0001 (0.01%)
+- **Code**: d=3 rotated surface code
+- **Physical error rate**: p=0.001
+
+## Current Status
+
+The decoder is running with the updated configuration:
+- File: `run_batch_demo.py`
+- Output log: `decoder_output_1000.log`
+- Monitor script: `monitor_decoder.sh`
+
+### Preliminary Results (100 samples)
+- BP-only: 0.0000 (0 errors)
+- BP+OSD-15: 0.0100 (1 error)
+
+### Expected Runtime
+- ~6 seconds per sample
+- 1000 samples: ~1.6 hours
+- Full 10,000 samples: ~16 hours
+
+## Monitoring Progress
+
+To monitor the decoder progress:
+```bash
+./monitor_decoder.sh
+```
+
+Or check the log directly:
+```bash
+tail -f decoder_output_1000.log
+```
+
+## Next Steps
+
+1. Wait for 1000-sample run to complete
+2. If logical error rate is close to target (0.0001), consider running with more samples
+3. If error rate is higher, may need to:
+   - Increase OSD order (e.g., OSD-20 or OSD-25)
+   - Further increase BP iterations
+   - Adjust damping parameter
+
+## Files Modified
+
+- `run_batch_demo.py`: Updated dataset path, BP_MAX_ITER, and NUM_SAMPLES

README.md

@@ -171,6 +171,24 @@ python examples/evidence_example.py
 python examples/minimal_example.py
 ```
 
+### Tanner Graph Decoding Tutorial
+
+For a comprehensive walkthrough of using Belief Propagation on Tanner graphs for surface code decoding, see the [Tanner Graph Walkthrough](https://giggleliu.github.io/BPDecoderPlus/tanner_graph_walkthrough/) documentation.
+
+The walkthrough covers:
+
+- **Tanner graph theory** - Bipartite graph representation of parity check codes
+- **Complete decoding pipeline** - From circuit generation to BP decoding and evaluation
+- **Visualization** - Interactive graph structures and convergence analysis
+- **Parameter tuning** - Damping, tolerance, and iteration optimization
+- **Hands-on examples** - Runnable code with d=3 surface code datasets
+
+**Run the companion script:**
+
+```bash
+uv run python examples/tanner_graph_walkthrough.py
+```
+
 ## Project Structure
 
 ```