diffai-python 0.3.4__tar.gz → 0.3.5__tar.gz

{diffai_python-0.3.4 → diffai_python-0.3.5}/Cargo.toml

@@ -6,7 +6,7 @@ members = [
 ]
 
 [workspace.package]
-version = "0.3.4"
+version = "0.3.5"
 edition = "2021"
 authors = ["kako-jun"]
 license = "MIT"

{diffai_python-0.3.4 → diffai_python-0.3.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: diffai-python
-Version: 0.3.4
+Version: 0.3.5
 Classifier: Development Status :: 4 - Beta
 Classifier: Environment :: Console
 Classifier: Intended Audience :: Developers
@@ -72,13 +72,13 @@ pip install diffai-python
 After installation, the `diffai` command is available:
 
 ```bash
-# Compare ML models
-diffai model_v1.safetensors model_v2.safetensors --stats
+# Compare ML models (30+ analysis features automatic)
+diffai model_v1.safetensors model_v2.safetensors
 
 # Compare NumPy arrays
-diffai data_v1.npy data_v2.npy --stats
+diffai data_v1.npy data_v2.npy
 
-# JSON output for automation
+# JSON output for automation (all ML features included)
 diffai model_v1.pt model_v2.pt --output json
 ```
 
@@ -93,8 +93,6 @@ print(result.raw_output)
 
 # With options
 options = diffai.DiffOptions(
-    stats=True,
-    architecture_comparison=True,
     output_format=diffai.OutputFormat.JSON
 )
 result = diffai.diff("model_v1.pt", "model_v2.pt", options)
@@ -108,18 +106,22 @@ if result.is_json:
 ### Advanced ML Analysis
 
 ```python
-# Comprehensive ML model analysis
+# Comprehensive ML model analysis (automatic for ML models)
 result = diffai.diff(
     "baseline.safetensors", 
     "improved.safetensors",
-    stats=True,
-    architecture_comparison=True,
-    memory_analysis=True,
-    anomaly_detection=True,
-    convergence_analysis=True
+    stats=True  # Enable statistical analysis
 )
 
 print(result.raw_output)
+
+# ML-specific analysis features (automatic for ML models)
+# - architecture_comparison: Model architecture and structural changes
+# - memory_analysis: Memory usage and optimization opportunities
+# - anomaly_detection: Numerical anomalies and training issues
+# - convergence_analysis: Training convergence patterns
+# - gradient_analysis: Gradient flow health assessment
+# - quantization_analysis: Quantization effect analysis
 ```
 
 ## Supported Formats
@@ -134,21 +136,21 @@ print(result.raw_output)
 - **JSON**: Machine-readable format for automation
 - **YAML**: Human-readable structured format
 
-## ML Analysis Features
+## ML Analysis Features (Automatic)
 
-The package provides 11 specialized ML analysis features:
+The package provides 30+ specialized ML analysis features that run automatically for PyTorch and Safetensors files:
 
-- `--stats`: Detailed tensor statistics
-- `--architecture-comparison`: Model structure comparison
-- `--memory-analysis`: Memory usage analysis  
-- `--anomaly-detection`: Numerical anomaly detection
-- `--convergence-analysis`: Training convergence analysis
-- `--gradient-analysis`: Gradient information analysis
-- `--similarity-matrix`: Layer similarity comparison
-- `--change-summary`: Detailed change summary
-- `--quantization-analysis`: Quantization impact analysis
-- `--sort-by-change-magnitude`: Sort by change magnitude
-- `--show-layer-impact`: Layer-specific impact analysis
+- **Detailed tensor statistics**: Mean, std, min, max, shape, dtype
+- **Model structure comparison**: Architecture and structural changes
+- **Memory usage analysis**: Memory optimization opportunities
+- **Numerical anomaly detection**: Training issues and anomalies
+- **Training convergence analysis**: Convergence patterns
+- **Gradient information analysis**: Gradient flow health
+- **Layer similarity comparison**: Inter-layer analysis
+- **Detailed change summary**: Comprehensive change patterns
+- **Quantization impact analysis**: Quantization effects
+- **Change magnitude sorting**: Priority-sorted differences
+- **Plus 20+ additional specialized features**
 
 ## API Reference
 
@@ -173,12 +175,9 @@ class DiffOptions:
     recursive: bool = False
     verbose: bool = False
     
-    # ML analysis options  
-    stats: bool = False
-    architecture_comparison: bool = False
-    memory_analysis: bool = False
-    anomaly_detection: bool = False
-    # ... and more
+    # For scientific data (NumPy/MATLAB)
+    stats: bool = False  # Only used for NumPy/MATLAB files
+    # Note: ML analysis runs automatically for PyTorch/Safetensors
 ```
 
 ### Results
@@ -204,9 +203,7 @@ class DiffResult:
 before = "model_baseline.safetensors"
 after = "model_finetuned.safetensors"
 
-result = diffai.diff(before, after, 
-                    stats=True, 
-                    convergence_analysis=True)
+result = diffai.diff(before, after)
 ```
 
 ### MLOps Integration
@@ -214,9 +211,7 @@ result = diffai.diff(before, after,
 # Automated model validation in CI/CD
 def validate_model_changes(old_model, new_model):
     result = diffai.diff(old_model, new_model,
-                        output_format=diffai.OutputFormat.JSON,
-                        anomaly_detection=True,
-                        memory_analysis=True)
+                        output_format=diffai.OutputFormat.JSON)
     
     if result.is_json:
         # Check for critical issues
@@ -244,9 +239,7 @@ def log_model_comparison(run_id1, run_id2):
     
     # Compare with diffai
     result = diffai.diff(model1_path, model2_path,
-                        output_format=diffai.OutputFormat.JSON,
-                        stats=True,
-                        architecture_comparison=True)
+                        output_format=diffai.OutputFormat.JSON)
     
     # Log results to MLflow
     with mlflow.start_run():
@@ -269,10 +262,7 @@ def log_model_comparison_wandb(model1_path, model2_path):
     """Log model comparison to Weights & Biases"""
     
     result = diffai.diff(model1_path, model2_path,
-                        output_format=diffai.OutputFormat.JSON,
-                        stats=True,
-                        memory_analysis=True,
-                        convergence_analysis=True)
+                        output_format=diffai.OutputFormat.JSON)
     
     # Log to wandb
     wandb.log({"model_comparison": result.data})
@@ -289,8 +279,7 @@ def log_model_comparison_wandb(model1_path, model2_path):
 ### Jupyter Notebooks
 ```python
 # Interactive analysis in notebooks
-result = diffai.diff("checkpoint_100.pt", "checkpoint_200.pt", 
-                    stats=True, memory_analysis=True)
+result = diffai.diff("checkpoint_100.pt", "checkpoint_200.pt")
 
 # Display results
 if result.is_json:

{diffai_python-0.3.4 → diffai_python-0.3.5}/README.md

@@ -34,13 +34,13 @@ pip install diffai-python
 After installation, the `diffai` command is available:
 
 ```bash
-# Compare ML models
-diffai model_v1.safetensors model_v2.safetensors --stats
+# Compare ML models (30+ analysis features automatic)
+diffai model_v1.safetensors model_v2.safetensors
 
 # Compare NumPy arrays
-diffai data_v1.npy data_v2.npy --stats
+diffai data_v1.npy data_v2.npy
 
-# JSON output for automation
+# JSON output for automation (all ML features included)
 diffai model_v1.pt model_v2.pt --output json
 ```
 
@@ -55,8 +55,6 @@ print(result.raw_output)
 
 # With options
 options = diffai.DiffOptions(
-    stats=True,
-    architecture_comparison=True,
     output_format=diffai.OutputFormat.JSON
 )
 result = diffai.diff("model_v1.pt", "model_v2.pt", options)
@@ -70,18 +68,22 @@ if result.is_json:
 ### Advanced ML Analysis
 
 ```python
-# Comprehensive ML model analysis
+# Comprehensive ML model analysis (automatic for ML models)
 result = diffai.diff(
     "baseline.safetensors", 
     "improved.safetensors",
-    stats=True,
-    architecture_comparison=True,
-    memory_analysis=True,
-    anomaly_detection=True,
-    convergence_analysis=True
+    stats=True  # Enable statistical analysis
 )
 
 print(result.raw_output)
+
+# ML-specific analysis features (automatic for ML models)
+# - architecture_comparison: Model architecture and structural changes
+# - memory_analysis: Memory usage and optimization opportunities
+# - anomaly_detection: Numerical anomalies and training issues
+# - convergence_analysis: Training convergence patterns
+# - gradient_analysis: Gradient flow health assessment
+# - quantization_analysis: Quantization effect analysis
 ```
 
 ## Supported Formats
@@ -96,21 +98,21 @@ print(result.raw_output)
 - **JSON**: Machine-readable format for automation
 - **YAML**: Human-readable structured format
 
-## ML Analysis Features
+## ML Analysis Features (Automatic)
 
-The package provides 11 specialized ML analysis features:
+The package provides 30+ specialized ML analysis features that run automatically for PyTorch and Safetensors files:
 
-- `--stats`: Detailed tensor statistics
-- `--architecture-comparison`: Model structure comparison
-- `--memory-analysis`: Memory usage analysis  
-- `--anomaly-detection`: Numerical anomaly detection
-- `--convergence-analysis`: Training convergence analysis
-- `--gradient-analysis`: Gradient information analysis
-- `--similarity-matrix`: Layer similarity comparison
-- `--change-summary`: Detailed change summary
-- `--quantization-analysis`: Quantization impact analysis
-- `--sort-by-change-magnitude`: Sort by change magnitude
-- `--show-layer-impact`: Layer-specific impact analysis
+- **Detailed tensor statistics**: Mean, std, min, max, shape, dtype
+- **Model structure comparison**: Architecture and structural changes
+- **Memory usage analysis**: Memory optimization opportunities
+- **Numerical anomaly detection**: Training issues and anomalies
+- **Training convergence analysis**: Convergence patterns
+- **Gradient information analysis**: Gradient flow health
+- **Layer similarity comparison**: Inter-layer analysis
+- **Detailed change summary**: Comprehensive change patterns
+- **Quantization impact analysis**: Quantization effects
+- **Change magnitude sorting**: Priority-sorted differences
+- **Plus 20+ additional specialized features**
 
 ## API Reference
 
@@ -135,12 +137,9 @@ class DiffOptions:
     recursive: bool = False
     verbose: bool = False
     
-    # ML analysis options  
-    stats: bool = False
-    architecture_comparison: bool = False
-    memory_analysis: bool = False
-    anomaly_detection: bool = False
-    # ... and more
+    # For scientific data (NumPy/MATLAB)
+    stats: bool = False  # Only used for NumPy/MATLAB files
+    # Note: ML analysis runs automatically for PyTorch/Safetensors
 ```
 
 ### Results
@@ -166,9 +165,7 @@ class DiffResult:
 before = "model_baseline.safetensors"
 after = "model_finetuned.safetensors"
 
-result = diffai.diff(before, after, 
-                    stats=True, 
-                    convergence_analysis=True)
+result = diffai.diff(before, after)
 ```
 
 ### MLOps Integration
@@ -176,9 +173,7 @@ result = diffai.diff(before, after,
 # Automated model validation in CI/CD
 def validate_model_changes(old_model, new_model):
     result = diffai.diff(old_model, new_model,
-                        output_format=diffai.OutputFormat.JSON,
-                        anomaly_detection=True,
-                        memory_analysis=True)
+                        output_format=diffai.OutputFormat.JSON)
     
     if result.is_json:
         # Check for critical issues
@@ -206,9 +201,7 @@ def log_model_comparison(run_id1, run_id2):
     
     # Compare with diffai
     result = diffai.diff(model1_path, model2_path,
-                        output_format=diffai.OutputFormat.JSON,
-                        stats=True,
-                        architecture_comparison=True)
+                        output_format=diffai.OutputFormat.JSON)
     
     # Log results to MLflow
     with mlflow.start_run():
@@ -231,10 +224,7 @@ def log_model_comparison_wandb(model1_path, model2_path):
     """Log model comparison to Weights & Biases"""
     
     result = diffai.diff(model1_path, model2_path,
-                        output_format=diffai.OutputFormat.JSON,
-                        stats=True,
-                        memory_analysis=True,
-                        convergence_analysis=True)
+                        output_format=diffai.OutputFormat.JSON)
     
     # Log to wandb
     wandb.log({"model_comparison": result.data})
@@ -251,8 +241,7 @@ def log_model_comparison_wandb(model1_path, model2_path):
 ### Jupyter Notebooks
 ```python
 # Interactive analysis in notebooks
-result = diffai.diff("checkpoint_100.pt", "checkpoint_200.pt", 
-                    stats=True, memory_analysis=True)
+result = diffai.diff("checkpoint_100.pt", "checkpoint_200.pt")
 
 # Display results
 if result.is_json:

{diffai_python-0.3.4 → diffai_python-0.3.5}/diffai-core/README.md

@@ -14,19 +14,25 @@ A next-generation diff tool specialized for **AI/ML and scientific computing wor
 $ diff model_v1.safetensors model_v2.safetensors
 Binary files model_v1.safetensors and model_v2.safetensors differ
 
-# diffai shows meaningful model changes
-$ diffai model_v1.safetensors model_v2.safetensors --stats
+# diffai shows meaningful model changes with full analysis
+$ diffai model_v1.safetensors model_v2.safetensors
   ~ fc1.bias: mean=0.0018->0.0017, std=0.0518->0.0647
   ~ fc1.weight: mean=-0.0002->-0.0001, std=0.0514->0.0716
   ~ fc2.weight: mean=-0.0008->-0.0018, std=0.0719->0.0883
   gradient_analysis: flow_health=healthy, norm=0.015000, ratio=1.0500
+  deployment_readiness: readiness=0.92, strategy=blue_green, risk=low
+  quantization_analysis: compression=0.0%, speedup=1.8x, precision_loss=1.5%
+
+[WARNING]
+• Memory usage increased moderately (+250MB). Monitor resource consumption.
+• Inference speed moderately affected (1.3x slower). Consider optimization opportunities.
 ```
 
 ## Key Features
 
 - **AI/ML Native**: Direct support for PyTorch (.pt/.pth), Safetensors (.safetensors), NumPy (.npy/.npz), and MATLAB (.mat) files
 - **Tensor Analysis**: Automatic calculation of tensor statistics (mean, std, min, max, shape, memory usage)
-- **ML Analysis Functions**: Statistical analysis, quantization analysis, architecture comparison, and more
+- **Comprehensive ML Analysis**: 30+ analysis functions including quantization, architecture, memory, convergence, anomaly detection, and deployment readiness - all enabled by default
 - **Scientific Data Support**: NumPy arrays and MATLAB matrices with complex number support
 - **Pure Rust Implementation**: No system dependencies, works on Windows/Linux/macOS without additional installations
 - **Multiple Output Formats**: Colored CLI, JSON for MLOps integration, YAML for human-readable reports
@@ -76,38 +82,43 @@ cargo build --release
 ### Basic Model Comparison
 
 ```bash
-# Compare PyTorch models
-diffai model_old.pt model_new.pt --stats
+# Compare PyTorch models with full analysis (default)
+diffai model_old.pt model_new.pt
 
-# Compare Safetensors with statistical analysis
-diffai checkpoint_v1.safetensors checkpoint_v2.safetensors --stats
+# Compare Safetensors with complete ML analysis
+diffai checkpoint_v1.safetensors checkpoint_v2.safetensors
 
 # Compare NumPy arrays
-diffai data_v1.npy data_v2.npy --stats
+diffai data_v1.npy data_v2.npy
 
 # Compare MATLAB files
-diffai experiment_v1.mat experiment_v2.mat --stats
+diffai experiment_v1.mat experiment_v2.mat
 ```
 
-### Advanced ML Analysis
+### ML Analysis Features
 
 ```bash
-# Current available analysis
-diffai baseline.safetensors finetuned.safetensors --stats --quantization-analysis
-
-# Combined analysis with sorting
-diffai original.pt optimized.pt --stats --quantization-analysis --sort-by-change-magnitude
+# Full ML analysis runs automatically for PyTorch/Safetensors
+diffai baseline.safetensors finetuned.safetensors
+# Outputs: 30+ analysis types including quantization, architecture, memory, etc.
 
 # JSON output for automation
-diffai model_v1.safetensors model_v2.safetensors --stats --output json
+diffai model_v1.safetensors model_v2.safetensors --output json
 
 # Detailed diagnostic information with verbose mode
-diffai model_v1.safetensors model_v2.safetensors --verbose --stats --architecture-comparison
+diffai model_v1.safetensors model_v2.safetensors --verbose
 
-# Future Phase 3 features (coming soon)
-diffai model_v1.safetensors model_v2.safetensors --architecture-comparison --memory-analysis
+# YAML output for human-readable reports
+diffai model_v1.safetensors model_v2.safetensors --output yaml
 ```
 
+## 📚 Documentation
+
+- **[Working Examples & Demonstrations](docs/examples/)** - See diffai in action with real outputs
+- **[API Documentation](https://docs.rs/diffai-core)** - Rust library documentation
+- **[User Guide](docs/user-guide.md)** - Comprehensive usage guide
+- **[ML Analysis Guide](docs/ml-analysis-guide.md)** - Deep dive into ML-specific features
+
 ## Supported File Formats
 
 ### ML Model Formats
@@ -128,26 +139,28 @@ diffai model_v1.safetensors model_v2.safetensors --architecture-comparison --mem
 
 ## ML Analysis Functions
 
-### Currently Available (v0.2.7)
-- `--stats` - Detailed tensor statistics (mean, std, min, max, shape, memory)
-- `--quantization-analysis` - Analyze quantization effects and efficiency
-- `--sort-by-change-magnitude` - Sort differences by magnitude for prioritization
-- `--show-layer-impact` - Layer-by-layer impact analysis
-- `--architecture-comparison` - Compare model architectures and structural changes
-- `--memory-analysis` - Analyze memory usage and optimization opportunities
-- `--anomaly-detection` - Detect numerical anomalies in model parameters
-- `--change-summary` - Generate detailed change summaries
-- `--convergence-analysis` - Analyze convergence patterns in model parameters
-- `--gradient-analysis` - Analyze gradient information when available
-- `--similarity-matrix` - Generate similarity matrix for model comparison
-
-### Coming in Phase 4 (ML Framework Expansion)
+### Automatic Comprehensive Analysis (v0.3.4)
+When comparing PyTorch or Safetensors files, diffai automatically runs 30+ ML analysis features:
+
+**Automatic Features Include:**
+- **Statistical Analysis**: Detailed tensor statistics (mean, std, min, max, shape, memory)
+- **Quantization Analysis**: Analyze quantization effects and efficiency
+- **Architecture Comparison**: Compare model architectures and structural changes
+- **Memory Analysis**: Analyze memory usage and optimization opportunities
+- **Anomaly Detection**: Detect numerical anomalies in model parameters
+- **Convergence Analysis**: Analyze convergence patterns in model parameters
+- **Gradient Analysis**: Analyze gradient information when available
+- **Deployment Readiness**: Assess production deployment readiness
+- **Regression Testing**: Automatic performance degradation detection
+- **Plus 20+ additional specialized features**
+
+### Future Enhancements
 - TensorFlow format support (.pb, .h5, SavedModel)
 - ONNX format support
 - Advanced visualization and charting features
 
 ### Design Philosophy
-diffai follows UNIX philosophy: simple, composable tools that do one thing well. Features are orthogonal and can be combined for powerful analysis workflows.
+diffai provides comprehensive analysis by default for ML models, eliminating choice paralysis. Users get all relevant insights without needing to remember or specify dozens of analysis flags.
 
 ## Debugging and Diagnostics
 
@@ -158,12 +171,12 @@ Get comprehensive diagnostic information for debugging and performance analysis:
 # Basic verbose output
 diffai model1.safetensors model2.safetensors --verbose
 
-# Verbose with ML analysis features
-diffai data1.json data2.json --verbose --stats --epsilon 0.001 --ignore-keys-regex "^id$"
+# Verbose with structured data filtering
+diffai data1.json data2.json --verbose --epsilon 0.001 --ignore-keys-regex "^id$"
 ```
 
 **Verbose output includes:**
-- **Configuration diagnostics**: Active ML features, format settings, filters
+- **Configuration diagnostics**: Format settings, filters, analysis modes
 - **File analysis**: Paths, sizes, detected formats, processing context
 - **Performance metrics**: Processing time, difference counts, optimization status
 - **Directory statistics**: File counts, comparison summaries (with `--recursive`)
@@ -172,9 +185,9 @@ diffai data1.json data2.json --verbose --stats --epsilon 0.001 --ignore-keys-reg
 ```
 === diffai verbose mode enabled ===
 Configuration:
-  Input format: None
+  Input format: Safetensors
   Output format: Cli
-  ML analysis features: statistics, architecture_comparison
+  ML analysis: Full analysis enabled (all 30 features)
   Epsilon tolerance: 0.001
 
 File analysis:
@@ -216,36 +229,36 @@ diffai model1.safetensors model2.safetensors --output yaml
 
 ### Research & Development
 ```bash
-# Compare model before and after fine-tuning
-diffai pretrained_model.safetensors finetuned_model.safetensors \
-  --learning-progress --convergence-analysis --stats
+# Compare model before and after fine-tuning (full analysis automatic)
+diffai pretrained_model.safetensors finetuned_model.safetensors
+# Outputs: learning_progress, convergence_analysis, parameter stats, and 27 more analyses
 
 # Analyze architectural changes during development
-diffai baseline_architecture.pt improved_architecture.pt \
-  --architecture-comparison --param-efficiency-analysis
+diffai baseline_architecture.pt improved_architecture.pt
+# Outputs: architecture_comparison, param_efficiency_analysis, and full ML analysis
 ```
 
 ### MLOps & CI/CD
 ```bash
-# Automated model validation in CI/CD
-diffai production_model.safetensors candidate_model.safetensors \
-  --deployment-readiness --regression-test --risk-assessment
+# Automated model validation in CI/CD (comprehensive analysis)
+diffai production_model.safetensors candidate_model.safetensors
+# Outputs: deployment_readiness, regression_test, risk_assessment, and 27 more analyses
 
-# Performance impact assessment
-diffai original_model.pt optimized_model.pt \
-  --quantization-analysis --memory-analysis --performance-impact-estimate
+# Performance impact assessment with JSON output for automation
+diffai original_model.pt optimized_model.pt --output json
+# Outputs: quantization_analysis, memory_analysis, performance_impact_estimate, etc.
 ```
 
 ### Scientific Computing
 ```bash
 # Compare NumPy experiment results
-diffai baseline_results.npy new_results.npy --stats
+diffai baseline_results.npy new_results.npy
 
 # Analyze MATLAB simulation data
-diffai simulation_v1.mat simulation_v2.mat --stats
+diffai simulation_v1.mat simulation_v2.mat
 
 # Compare compressed NumPy archives
-diffai dataset_v1.npz dataset_v2.npz --stats
+diffai dataset_v1.npz dataset_v2.npz
 ```
 
 ### Experiment Tracking
@@ -265,7 +278,8 @@ diffai model_a.safetensors model_b.safetensors \
 - `-f, --format <FORMAT>` - Specify input file format
 - `-o, --output <OUTPUT>` - Choose output format (cli, json, yaml)
 - `-r, --recursive` - Compare directories recursively
-- `--stats` - Show detailed statistics for ML models
+
+**Note:** For ML models (PyTorch/Safetensors), comprehensive analysis including statistics runs automatically
 
 ### Advanced Options
 - `--path <PATH>` - Filter differences by specific path
@@ -276,9 +290,16 @@ diffai model_a.safetensors model_b.safetensors \
 
 ## Examples
 
-### Basic Tensor Comparison
+### Basic Tensor Comparison (Automatic)
 ```bash
-$ diffai simple_model_v1.safetensors simple_model_v2.safetensors --stats
+$ diffai simple_model_v1.safetensors simple_model_v2.safetensors
+anomaly_detection: type=none, severity=none, action="continue_training"
+architecture_comparison: type1=feedforward, type2=feedforward, deployment_readiness=ready
+convergence_analysis: status=converging, stability=0.92
+gradient_analysis: flow_health=healthy, norm=0.021069
+memory_analysis: delta=+0.0MB, efficiency=1.000000
+quantization_analysis: compression=0.0%, speedup=1.8x, precision_loss=1.5%
+regression_test: passed=true, degradation=-2.5%, severity=low
   ~ fc1.bias: mean=0.0018->0.0017, std=0.0518->0.0647
   ~ fc1.weight: mean=-0.0002->-0.0001, std=0.0514->0.0716
   ~ fc2.bias: mean=-0.0076->-0.0257, std=0.0661->0.0973
@@ -287,24 +308,28 @@ $ diffai simple_model_v1.safetensors simple_model_v2.safetensors --stats
   ~ fc3.weight: mean=-0.0035->-0.0010, std=0.0990->0.1113
 ```
 
-### Advanced Analysis
+### JSON Output for Automation
 ```bash
-$ diffai baseline.safetensors improved.safetensors --deployment-readiness --architecture-comparison
-deployment_readiness: readiness=0.92, strategy=blue_green, risk=low, timeline=ready_for_immediate_deployment
-architecture_comparison: type1=feedforward, type2=feedforward, depth=3->3, differences=0
-  ~ fc1.bias: mean=0.0018->0.0017, std=0.0518->0.0647
-  ~ fc1.weight: mean=-0.0002->-0.0001, std=0.0514->0.0716
+$ diffai baseline.safetensors improved.safetensors --output json
+{
+  "anomaly_detection": {"type": "none", "severity": "none"},
+  "architecture_comparison": {"type1": "feedforward", "type2": "feedforward"},
+  "deployment_readiness": {"readiness": 0.92, "strategy": "blue_green"},
+  "quantization_analysis": {"compression": "0.0%", "speedup": "1.8x"},
+  "regression_test": {"passed": true, "degradation": "-2.5%"}
+  // ... plus 25+ additional analysis features
+}
 ```
 
 ### Scientific Data Analysis
 ```bash
-$ diffai experiment_data_v1.npy experiment_data_v2.npy --stats
+$ diffai experiment_data_v1.npy experiment_data_v2.npy
   ~ data: shape=[1000, 256], mean=0.1234->0.1456, std=0.9876->0.9654, dtype=float64
 ```
 
 ### MATLAB File Comparison
 ```bash
-$ diffai simulation_v1.mat simulation_v2.mat --stats
+$ diffai simulation_v1.mat simulation_v2.mat
   ~ results: var=results, shape=[500, 100], mean=2.3456->2.4567, std=1.2345->1.3456, dtype=double
   + new_variable: var=new_variable, shape=[100], dtype=single, elements=100, size=0.39KB
 ```

{diffai_python-0.3.4 → diffai_python-0.3.5}/diffai-python/Cargo.lock

@@ -432,7 +432,7 @@ dependencies = [
 
 [[package]]
 name = "diffai-core"
-version = "0.3.4"
+version = "0.3.5"
 dependencies = [
  "anyhow",
  "bytemuck",
@@ -455,7 +455,7 @@ dependencies = [
 
 [[package]]
 name = "diffai-python"
-version = "0.3.4"
+version = "0.3.5"
 dependencies = [
  "anyhow",
  "clap",

{diffai_python-0.3.4 → diffai_python-0.3.5}/diffai-python/Cargo.toml

@@ -3,7 +3,7 @@
 
 [package]
 name = "diffai-python"
-version = "0.3.4"
+version = "0.3.5"
 edition = "2021"
 authors = ["kako-jun"]
 license = "MIT"
@@ -17,7 +17,7 @@ path = "src/main.rs"
 
 [dependencies]
 # Reference to the actual diffai dependencies
-diffai-core = { version = "0.3.4", path = "../diffai-core" }
+diffai-core = { version = "0.3.5", path = "../diffai-core" }
 clap = { version = "4.0", features = ["derive"] }
 colored = "2.0"
 serde = { version = "1.0", features = ["derive"] }

{diffai_python-0.3.4 → diffai_python-0.3.5}/diffai-python/README.md

@@ -34,13 +34,13 @@ pip install diffai-python
 After installation, the `diffai` command is available:
 
 ```bash
-# Compare ML models
-diffai model_v1.safetensors model_v2.safetensors --stats
+# Compare ML models (30+ analysis features automatic)
+diffai model_v1.safetensors model_v2.safetensors
 
 # Compare NumPy arrays
-diffai data_v1.npy data_v2.npy --stats
+diffai data_v1.npy data_v2.npy
 
-# JSON output for automation
+# JSON output for automation (all ML features included)
 diffai model_v1.pt model_v2.pt --output json
 ```
 
@@ -55,8 +55,6 @@ print(result.raw_output)
 
 # With options
 options = diffai.DiffOptions(
-    stats=True,
-    architecture_comparison=True,
     output_format=diffai.OutputFormat.JSON
 )
 result = diffai.diff("model_v1.pt", "model_v2.pt", options)
@@ -70,18 +68,22 @@ if result.is_json:
 ### Advanced ML Analysis
 
 ```python
-# Comprehensive ML model analysis
+# Comprehensive ML model analysis (automatic for ML models)
 result = diffai.diff(
     "baseline.safetensors", 
     "improved.safetensors",
-    stats=True,
-    architecture_comparison=True,
-    memory_analysis=True,
-    anomaly_detection=True,
-    convergence_analysis=True
+    stats=True  # Enable statistical analysis
 )
 
 print(result.raw_output)
+
+# ML-specific analysis features (automatic for ML models)
+# - architecture_comparison: Model architecture and structural changes
+# - memory_analysis: Memory usage and optimization opportunities
+# - anomaly_detection: Numerical anomalies and training issues
+# - convergence_analysis: Training convergence patterns
+# - gradient_analysis: Gradient flow health assessment
+# - quantization_analysis: Quantization effect analysis
 ```
 
 ## Supported Formats
@@ -96,21 +98,21 @@ print(result.raw_output)
 - **JSON**: Machine-readable format for automation
 - **YAML**: Human-readable structured format
 
-## ML Analysis Features
+## ML Analysis Features (Automatic)
 
-The package provides 11 specialized ML analysis features:
+The package provides 30+ specialized ML analysis features that run automatically for PyTorch and Safetensors files:
 
-- `--stats`: Detailed tensor statistics
-- `--architecture-comparison`: Model structure comparison
-- `--memory-analysis`: Memory usage analysis  
-- `--anomaly-detection`: Numerical anomaly detection
-- `--convergence-analysis`: Training convergence analysis
-- `--gradient-analysis`: Gradient information analysis
-- `--similarity-matrix`: Layer similarity comparison
-- `--change-summary`: Detailed change summary
-- `--quantization-analysis`: Quantization impact analysis
-- `--sort-by-change-magnitude`: Sort by change magnitude
-- `--show-layer-impact`: Layer-specific impact analysis
+- **Detailed tensor statistics**: Mean, std, min, max, shape, dtype
+- **Model structure comparison**: Architecture and structural changes
+- **Memory usage analysis**: Memory optimization opportunities
+- **Numerical anomaly detection**: Training issues and anomalies
+- **Training convergence analysis**: Convergence patterns
+- **Gradient information analysis**: Gradient flow health
+- **Layer similarity comparison**: Inter-layer analysis
+- **Detailed change summary**: Comprehensive change patterns
+- **Quantization impact analysis**: Quantization effects
+- **Change magnitude sorting**: Priority-sorted differences
+- **Plus 20+ additional specialized features**
 
 ## API Reference
 
@@ -135,12 +137,9 @@ class DiffOptions:
     recursive: bool = False
     verbose: bool = False
     
-    # ML analysis options  
-    stats: bool = False
-    architecture_comparison: bool = False
-    memory_analysis: bool = False
-    anomaly_detection: bool = False
-    # ... and more
+    # For scientific data (NumPy/MATLAB)
+    stats: bool = False  # Only used for NumPy/MATLAB files
+    # Note: ML analysis runs automatically for PyTorch/Safetensors
 ```
 
 ### Results
@@ -166,9 +165,7 @@ class DiffResult:
 before = "model_baseline.safetensors"
 after = "model_finetuned.safetensors"
 
-result = diffai.diff(before, after, 
-                    stats=True, 
-                    convergence_analysis=True)
+result = diffai.diff(before, after)
 ```
 
 ### MLOps Integration
@@ -176,9 +173,7 @@ result = diffai.diff(before, after,
 # Automated model validation in CI/CD
 def validate_model_changes(old_model, new_model):
     result = diffai.diff(old_model, new_model,
-                        output_format=diffai.OutputFormat.JSON,
-                        anomaly_detection=True,
-                        memory_analysis=True)
+                        output_format=diffai.OutputFormat.JSON)
     
     if result.is_json:
         # Check for critical issues
@@ -206,9 +201,7 @@ def log_model_comparison(run_id1, run_id2):
     
     # Compare with diffai
     result = diffai.diff(model1_path, model2_path,
-                        output_format=diffai.OutputFormat.JSON,
-                        stats=True,
-                        architecture_comparison=True)
+                        output_format=diffai.OutputFormat.JSON)
     
     # Log results to MLflow
     with mlflow.start_run():
@@ -231,10 +224,7 @@ def log_model_comparison_wandb(model1_path, model2_path):
     """Log model comparison to Weights & Biases"""
     
     result = diffai.diff(model1_path, model2_path,
-                        output_format=diffai.OutputFormat.JSON,
-                        stats=True,
-                        memory_analysis=True,
-                        convergence_analysis=True)
+                        output_format=diffai.OutputFormat.JSON)
     
     # Log to wandb
     wandb.log({"model_comparison": result.data})
@@ -251,8 +241,7 @@ def log_model_comparison_wandb(model1_path, model2_path):
 ### Jupyter Notebooks
 ```python
 # Interactive analysis in notebooks
-result = diffai.diff("checkpoint_100.pt", "checkpoint_200.pt", 
-                    stats=True, memory_analysis=True)
+result = diffai.diff("checkpoint_100.pt", "checkpoint_200.pt")
 
 # Display results
 if result.is_json:

{diffai_python-0.3.4 → diffai_python-0.3.5}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "maturin"
 
 [project]
 name = "diffai-python"
-version = "0.3.4"
+version = "0.3.5"
 description = "AI/ML specialized diff tool for deep tensor comparison and analysis"
 readme = "README.md"
 license = "MIT"