tensorstudio 1.7.0__tar.gz → 1.9.0__tar.gz

{tensorstudio-1.7.0 → tensorstudio-1.9.0}/CHANGELOG.md

@@ -2,6 +2,42 @@
 
 ## Unreleased
 
+## 1.9.0 - 2026-07-07
+
+- Completed the ordered Neural Network Building Blocks roadmap section as one
+  release batch.
+- Added native grouped `conv2d`, native `conv_transpose2d`, and native
+  embedding lookup with autograd support.
+- Added Python-level `Conv1d`, `DepthwiseConv2d`, `ConvTranspose2d`,
+  `BatchNorm1d`, `BatchNorm2d`, `LayerNorm`, `Embedding`,
+  adaptive/global pooling, and additional activation modules.
+- Added `tensorstudio.nn.init` with Xavier, Kaiming, normal, uniform, zero, and
+  one initializers.
+- Added label-smoothing cross entropy, focal loss, KL divergence, negative log
+  likelihood, and cosine embedding loss modules and functional helpers.
+- Added module buffers, buffer-aware `state_dict()` support, and model summary
+  utilities for parameters, shapes, and estimated tensor memory.
+- Expanded tests and docs for the section-5 neural-network API surface.
+
+## 1.8.0 - 2026-07-07
+
+- Completed the ordered Autograd Coverage And Hardening roadmap section as one
+  release batch.
+- Added `retain_graph` support to `Tensor.backward()` and
+  `tensorstudio.autograd.backward()`.
+- Added graph lifecycle hardening: normal backward frees non-leaf graph history,
+  repeated backward through a freed graph raises a clear error, and retained
+  graphs clear intermediate gradients between backward passes.
+- Added Tensor `is_leaf`, `clear_history()`, and `detach_()` controls for
+  explicit graph lifecycle management.
+- Added guarded public in-place methods `zero_()`, `fill_()`, and `add_()`.
+  They reject gradient-tracked mutation while grad mode is enabled and work
+  inside `tensorstudio.no_grad()`.
+- Expanded non-scalar backward and finite-difference gradient tests for stable
+  probability ops, statistics, norms, and batched matrix multiplication.
+- Expanded autograd documentation with a coverage matrix, lifecycle notes, and
+  explicit higher-order-gradient limitations.
+
 ## 1.7.0 - 2026-07-07
 
 - Completed the ordered Core Math Expansion roadmap section as one release

{tensorstudio-1.7.0 → tensorstudio-1.9.0}/CMakeLists.txt

@@ -1,70 +1,70 @@
-cmake_minimum_required(VERSION 3.18)
-
-project(tensorstudio VERSION 1.7.0 LANGUAGES CXX)
-
-find_package(Python COMPONENTS Interpreter Development.Module REQUIRED)
-set(PYBIND11_FINDPYTHON ON)
-find_package(pybind11 CONFIG REQUIRED)
-
-option(TENSORSTUDIO_ENABLE_BLAS "Enable optional CBLAS-backed matmul when available" ON)
-
-set(TENSORSTUDIO_CORE_SOURCES
-  src/core/tensor.cpp
-  src/core/dtype.cpp
-  src/core/shape.cpp
-  src/core/device.cpp
-  src/core/storage.cpp
-  src/core/ops.cpp
-  src/core/autograd.cpp
-  src/core/module.cpp
-  src/core/optim.cpp
-  src/core/random.cpp
-  src/core/serialization.cpp
-  src/core/errors.cpp
-  src/core/perf.cpp
-)
-
-set(TENSORSTUDIO_BINDING_SOURCES
-  src/bindings/bindings.cpp
-  src/bindings/bind_tensor.cpp
-  src/bindings/bind_ops.cpp
-  src/bindings/bind_autograd.cpp
-  src/bindings/bind_nn.cpp
-  src/bindings/bind_optim.cpp
-)
-
-pybind11_add_module(_C MODULE
-  ${TENSORSTUDIO_CORE_SOURCES}
-  ${TENSORSTUDIO_BINDING_SOURCES}
-)
-
-target_include_directories(_C PRIVATE include src/bindings)
-target_compile_features(_C PRIVATE cxx_std_20)
-
-if(TENSORSTUDIO_ENABLE_BLAS)
-  find_package(BLAS)
-  if(BLAS_FOUND)
-    include(CheckIncludeFileCXX)
-    check_include_file_cxx(cblas.h TENSORSTUDIO_HAVE_CBLAS_H)
-    target_link_libraries(_C PRIVATE ${BLAS_LIBRARIES})
-    if(APPLE)
-      target_compile_definitions(_C PRIVATE TENSORSTUDIO_HAS_ACCELERATE=1)
-    elseif(TENSORSTUDIO_HAVE_CBLAS_H)
-      target_compile_definitions(_C PRIVATE TENSORSTUDIO_HAS_CBLAS=1)
-    endif()
-  endif()
-endif()
-
-if(MSVC)
-  target_compile_options(_C PRIVATE /W4 /permissive-)
-else()
-  target_compile_options(_C PRIVATE -Wall -Wextra -Wpedantic)
-endif()
-
-target_compile_definitions(_C PRIVATE TENSORSTUDIO_VERSION="${PROJECT_VERSION}")
-
-install(TARGETS _C
-  LIBRARY DESTINATION tensorstudio
-  RUNTIME DESTINATION tensorstudio
-  ARCHIVE DESTINATION tensorstudio
-)
+cmake_minimum_required(VERSION 3.18)
+
+project(tensorstudio VERSION 1.9.0 LANGUAGES CXX)
+
+find_package(Python COMPONENTS Interpreter Development.Module REQUIRED)
+set(PYBIND11_FINDPYTHON ON)
+find_package(pybind11 CONFIG REQUIRED)
+
+option(TENSORSTUDIO_ENABLE_BLAS "Enable optional CBLAS-backed matmul when available" ON)
+
+set(TENSORSTUDIO_CORE_SOURCES
+  src/core/tensor.cpp
+  src/core/dtype.cpp
+  src/core/shape.cpp
+  src/core/device.cpp
+  src/core/storage.cpp
+  src/core/ops.cpp
+  src/core/autograd.cpp
+  src/core/module.cpp
+  src/core/optim.cpp
+  src/core/random.cpp
+  src/core/serialization.cpp
+  src/core/errors.cpp
+  src/core/perf.cpp
+)
+
+set(TENSORSTUDIO_BINDING_SOURCES
+  src/bindings/bindings.cpp
+  src/bindings/bind_tensor.cpp
+  src/bindings/bind_ops.cpp
+  src/bindings/bind_autograd.cpp
+  src/bindings/bind_nn.cpp
+  src/bindings/bind_optim.cpp
+)
+
+pybind11_add_module(_C MODULE
+  ${TENSORSTUDIO_CORE_SOURCES}
+  ${TENSORSTUDIO_BINDING_SOURCES}
+)
+
+target_include_directories(_C PRIVATE include src/bindings)
+target_compile_features(_C PRIVATE cxx_std_20)
+
+if(TENSORSTUDIO_ENABLE_BLAS)
+  find_package(BLAS)
+  if(BLAS_FOUND)
+    include(CheckIncludeFileCXX)
+    check_include_file_cxx(cblas.h TENSORSTUDIO_HAVE_CBLAS_H)
+    target_link_libraries(_C PRIVATE ${BLAS_LIBRARIES})
+    if(APPLE)
+      target_compile_definitions(_C PRIVATE TENSORSTUDIO_HAS_ACCELERATE=1)
+    elseif(TENSORSTUDIO_HAVE_CBLAS_H)
+      target_compile_definitions(_C PRIVATE TENSORSTUDIO_HAS_CBLAS=1)
+    endif()
+  endif()
+endif()
+
+if(MSVC)
+  target_compile_options(_C PRIVATE /W4 /permissive-)
+else()
+  target_compile_options(_C PRIVATE -Wall -Wextra -Wpedantic)
+endif()
+
+target_compile_definitions(_C PRIVATE TENSORSTUDIO_VERSION="${PROJECT_VERSION}")
+
+install(TARGETS _C
+  LIBRARY DESTINATION tensorstudio
+  RUNTIME DESTINATION tensorstudio
+  ARCHIVE DESTINATION tensorstudio
+)

{tensorstudio-1.7.0 → tensorstudio-1.9.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: tensorstudio
-Version: 1.7.0
+Version: 1.9.0
 Summary: TensorStudio is a compact C++ tensor and autograd engine with a Python API for learning, experimentation, and lightweight ML workloads.
 Keywords: tensor,autograd,machine-learning,cpp,pybind11
 Author: TensorStudio contributors
@@ -71,11 +71,14 @@ Description-Content-Type: text/markdown
 TensorStudio is a compact C++ tensor and autograd engine with a Python API for
 learning, experimentation, and lightweight ML workloads.
 
-TensorStudio `1.7.0` is a CPU-only stable API foundation with native C++
+TensorStudio `1.9.0` is a CPU-only stable API foundation with native C++
 threading, storage reuse, SIMD-friendly typed kernels, and optional
 CBLAS/Accelerate matrix multiplication when available. It adds native stable
 softmax/logsumexp, batched matrix multiplication, statistical reductions,
-boolean reductions, and seeded random distributions. It is eager-only,
+boolean reductions, seeded random distributions, and a hardened eager autograd
+lifecycle. The neural-network layer now includes grouped/depthwise/1D/
+transposed convolution, normalization layers, embeddings, richer activations,
+initializers, additional losses, and model summaries. It is eager-only,
 intentionally compact, and not a replacement for mature ML frameworks.
 
 ## Install
@@ -241,11 +244,20 @@ loss.backward()
 print(x.grad.tolist())
 ```
 
+Reuse a graph explicitly when needed:
+
+```python
+loss = (x * x).sum()
+loss.backward(retain_graph=True)
+loss.backward()
+```
+
 Use `no_grad()` when you want eager computation without recording a graph:
 
 ```python
 with ts.no_grad():
     y = x * 2
+    x.zero_()
 ```
 
 ## Neural Networks
@@ -277,6 +289,23 @@ print(model.state_dict().keys())
 print(model.parameter_count())
 ```
 
+The `1.9.0` neural-network surface also includes initialization helpers,
+normalization layers, embeddings, grouped/depthwise/1D/transposed convolution,
+adaptive/global pooling, richer activations, and model summaries:
+
+```python
+model = nn.Sequential(
+    nn.Conv2d(1, 8, kernel_size=3, padding=1),
+    nn.BatchNorm2d(8),
+    nn.GELU(),
+    nn.GlobalAvgPool2d(),
+    nn.Flatten(),
+    nn.Linear(8, 10),
+)
+nn.init.kaiming_uniform_(model[0].weight, nonlinearity="relu", seed=7)
+print(nn.summary(model, input_shape=(1, 1, 28, 28))["total_parameters"])
+```
+
 ## Vision
 
 TensorStudio includes a practical computer-vision namespace for local image
@@ -381,11 +410,11 @@ Run the loose local regression thresholds with:
 python benchmark_all.py --check-thresholds
 ```
 
-On one Windows CPython 3.10 development run reporting `1.7.0`, with
+On one Windows CPython 3.10 development run reporting `1.9.0`, with
 TensorStudio threads enabled, storage pooling enabled, SSE2 autovectorization
 reported, and no BLAS provider found, TensorStudio beat NumPy on 7 local
 benchmark cases and lost on 96 NumPy-comparable cases. JAX CPU dispatch was
-available on that machine; TensorStudio won 45 local cases and lost 53. The
+available on that machine; TensorStudio won 55 local cases and lost 43. The
 strongest local wins were the simple NumPy convolution/pooling reference loops
 and some small JAX-dispatch-heavy eager cases. NumPy and JAX were faster for
 many elementwise, reduction, matrix multiplication, larger activation, and
@@ -397,15 +426,15 @@ Snapshot from that local run:
 
 | operation | shape | TensorStudio | NumPy | JAX CPU dispatch | TS vs NumPy | TS vs JAX |
 |---|---:|---:|---:|---:|---:|---:|
-| `sigmoid` | `(32,)` | 0.0152 ms | 0.0044 ms | 0.0727 ms | 0.2909x | 4.7701x |
-| `mean` | `(32,)` | 0.0177 ms | 0.0083 ms | 0.0113 ms | 0.4715x | 0.6400x |
-| `sum_axis1` | `(16, 16)` | 0.0163 ms | 0.0031 ms | 0.0123 ms | 0.1901x | 0.7536x |
-| `chain_relu` | `(128,)` | 0.0867 ms | 0.0062 ms | 0.0923 ms | 0.0721x | 1.0641x |
-| `matmul` | `(256, 256)` | 2.2252 ms | 0.4754 ms | 0.2750 ms | 0.2136x | 0.1236x |
-| `conv2d_3x3_padding1` | `(1, 1, 8, 8)` | 0.1915 ms | 1.2845 ms | 0.1081 ms | 6.7064x | 0.5646x |
-| `max_pool2d_2x2` | `(1, 1, 16, 16)` | 0.0277 ms | 0.1577 ms | n/a | 5.7024x | n/a |
-| `avg_pool2d_2x2` | `(1, 1, 16, 16)` | 0.0278 ms | 0.5538 ms | n/a | 19.8919x | n/a |
-| `elementwise_backward` | `(1024,)` | 2.7344 ms | n/a | n/a | n/a | n/a |
+| `sigmoid` | `(32,)` | 0.0154 ms | 0.0048 ms | 0.0855 ms | 0.3098x | 5.5404x |
+| `mean` | `(32,)` | 0.0187 ms | 0.0104 ms | 0.0143 ms | 0.5575x | 0.7616x |
+| `sum_axis1` | `(16, 16)` | 0.0172 ms | 0.0027 ms | 0.0136 ms | 0.1551x | 0.7942x |
+| `chain_relu` | `(128,)` | 0.0957 ms | 0.0057 ms | 0.1030 ms | 0.0596x | 1.0768x |
+| `matmul` | `(256, 256)` | 2.3777 ms | 0.4328 ms | 0.2489 ms | 0.1820x | 0.1047x |
+| `conv2d_3x3_padding1` | `(1, 1, 8, 8)` | 0.2216 ms | 1.3908 ms | 0.1073 ms | 6.2770x | 0.4841x |
+| `max_pool2d_2x2` | `(1, 1, 16, 16)` | 0.0280 ms | 0.1846 ms | n/a | 6.5816x | n/a |
+| `avg_pool2d_2x2` | `(1, 1, 16, 16)` | 0.0281 ms | 0.5595 ms | n/a | 19.8874x | n/a |
+| `elementwise_backward` | `(1024,)` | 2.8255 ms | n/a | n/a | n/a | n/a |
 
 Speedup is `competitor median / TensorStudio median`, so values above `1.0x`
 favor TensorStudio.
@@ -506,8 +535,9 @@ tokens or print secrets.
   exposing a compatible CBLAS/Accelerate interface; otherwise TensorStudio uses
   a portable C++ fallback.
 - No graph compiler or distributed runtime.
-- Convolution and pooling support are currently limited to CPU NCHW
-  `conv2d`, `max_pool2d`, and `avg_pool2d` style workloads.
+- Convolution and pooling support are CPU-only. Native kernels include NCHW
+  `conv2d`, grouped/depthwise convolution, `conv_transpose2d`, `max_pool2d`,
+  `avg_pool2d`, and embedding lookup; they are not CUDA/cuDNN replacements.
 - Vision covers local image-classification utilities, metrics, visualization,
   and compact CNNs. It is not an OpenCV replacement and does not include
   pretrained model zoos, detection/segmentation training stacks, video IO, or

{tensorstudio-1.7.0 → tensorstudio-1.9.0}/README.md

@@ -7,11 +7,14 @@
 TensorStudio is a compact C++ tensor and autograd engine with a Python API for
 learning, experimentation, and lightweight ML workloads.
 
-TensorStudio `1.7.0` is a CPU-only stable API foundation with native C++
+TensorStudio `1.9.0` is a CPU-only stable API foundation with native C++
 threading, storage reuse, SIMD-friendly typed kernels, and optional
 CBLAS/Accelerate matrix multiplication when available. It adds native stable
 softmax/logsumexp, batched matrix multiplication, statistical reductions,
-boolean reductions, and seeded random distributions. It is eager-only,
+boolean reductions, seeded random distributions, and a hardened eager autograd
+lifecycle. The neural-network layer now includes grouped/depthwise/1D/
+transposed convolution, normalization layers, embeddings, richer activations,
+initializers, additional losses, and model summaries. It is eager-only,
 intentionally compact, and not a replacement for mature ML frameworks.
 
 ## Install
@@ -177,11 +180,20 @@ loss.backward()
 print(x.grad.tolist())
 ```
 
+Reuse a graph explicitly when needed:
+
+```python
+loss = (x * x).sum()
+loss.backward(retain_graph=True)
+loss.backward()
+```
+
 Use `no_grad()` when you want eager computation without recording a graph:
 
 ```python
 with ts.no_grad():
     y = x * 2
+    x.zero_()
 ```
 
 ## Neural Networks
@@ -213,6 +225,23 @@ print(model.state_dict().keys())
 print(model.parameter_count())
 ```
 
+The `1.9.0` neural-network surface also includes initialization helpers,
+normalization layers, embeddings, grouped/depthwise/1D/transposed convolution,
+adaptive/global pooling, richer activations, and model summaries:
+
+```python
+model = nn.Sequential(
+    nn.Conv2d(1, 8, kernel_size=3, padding=1),
+    nn.BatchNorm2d(8),
+    nn.GELU(),
+    nn.GlobalAvgPool2d(),
+    nn.Flatten(),
+    nn.Linear(8, 10),
+)
+nn.init.kaiming_uniform_(model[0].weight, nonlinearity="relu", seed=7)
+print(nn.summary(model, input_shape=(1, 1, 28, 28))["total_parameters"])
+```
+
 ## Vision
 
 TensorStudio includes a practical computer-vision namespace for local image
@@ -317,11 +346,11 @@ Run the loose local regression thresholds with:
 python benchmark_all.py --check-thresholds
 ```
 
-On one Windows CPython 3.10 development run reporting `1.7.0`, with
+On one Windows CPython 3.10 development run reporting `1.9.0`, with
 TensorStudio threads enabled, storage pooling enabled, SSE2 autovectorization
 reported, and no BLAS provider found, TensorStudio beat NumPy on 7 local
 benchmark cases and lost on 96 NumPy-comparable cases. JAX CPU dispatch was
-available on that machine; TensorStudio won 45 local cases and lost 53. The
+available on that machine; TensorStudio won 55 local cases and lost 43. The
 strongest local wins were the simple NumPy convolution/pooling reference loops
 and some small JAX-dispatch-heavy eager cases. NumPy and JAX were faster for
 many elementwise, reduction, matrix multiplication, larger activation, and
@@ -333,15 +362,15 @@ Snapshot from that local run:
 
 | operation | shape | TensorStudio | NumPy | JAX CPU dispatch | TS vs NumPy | TS vs JAX |
 |---|---:|---:|---:|---:|---:|---:|
-| `sigmoid` | `(32,)` | 0.0152 ms | 0.0044 ms | 0.0727 ms | 0.2909x | 4.7701x |
-| `mean` | `(32,)` | 0.0177 ms | 0.0083 ms | 0.0113 ms | 0.4715x | 0.6400x |
-| `sum_axis1` | `(16, 16)` | 0.0163 ms | 0.0031 ms | 0.0123 ms | 0.1901x | 0.7536x |
-| `chain_relu` | `(128,)` | 0.0867 ms | 0.0062 ms | 0.0923 ms | 0.0721x | 1.0641x |
-| `matmul` | `(256, 256)` | 2.2252 ms | 0.4754 ms | 0.2750 ms | 0.2136x | 0.1236x |
-| `conv2d_3x3_padding1` | `(1, 1, 8, 8)` | 0.1915 ms | 1.2845 ms | 0.1081 ms | 6.7064x | 0.5646x |
-| `max_pool2d_2x2` | `(1, 1, 16, 16)` | 0.0277 ms | 0.1577 ms | n/a | 5.7024x | n/a |
-| `avg_pool2d_2x2` | `(1, 1, 16, 16)` | 0.0278 ms | 0.5538 ms | n/a | 19.8919x | n/a |
-| `elementwise_backward` | `(1024,)` | 2.7344 ms | n/a | n/a | n/a | n/a |
+| `sigmoid` | `(32,)` | 0.0154 ms | 0.0048 ms | 0.0855 ms | 0.3098x | 5.5404x |
+| `mean` | `(32,)` | 0.0187 ms | 0.0104 ms | 0.0143 ms | 0.5575x | 0.7616x |
+| `sum_axis1` | `(16, 16)` | 0.0172 ms | 0.0027 ms | 0.0136 ms | 0.1551x | 0.7942x |
+| `chain_relu` | `(128,)` | 0.0957 ms | 0.0057 ms | 0.1030 ms | 0.0596x | 1.0768x |
+| `matmul` | `(256, 256)` | 2.3777 ms | 0.4328 ms | 0.2489 ms | 0.1820x | 0.1047x |
+| `conv2d_3x3_padding1` | `(1, 1, 8, 8)` | 0.2216 ms | 1.3908 ms | 0.1073 ms | 6.2770x | 0.4841x |
+| `max_pool2d_2x2` | `(1, 1, 16, 16)` | 0.0280 ms | 0.1846 ms | n/a | 6.5816x | n/a |
+| `avg_pool2d_2x2` | `(1, 1, 16, 16)` | 0.0281 ms | 0.5595 ms | n/a | 19.8874x | n/a |
+| `elementwise_backward` | `(1024,)` | 2.8255 ms | n/a | n/a | n/a | n/a |
 
 Speedup is `competitor median / TensorStudio median`, so values above `1.0x`
 favor TensorStudio.
@@ -442,8 +471,9 @@ tokens or print secrets.
   exposing a compatible CBLAS/Accelerate interface; otherwise TensorStudio uses
   a portable C++ fallback.
 - No graph compiler or distributed runtime.
-- Convolution and pooling support are currently limited to CPU NCHW
-  `conv2d`, `max_pool2d`, and `avg_pool2d` style workloads.
+- Convolution and pooling support are CPU-only. Native kernels include NCHW
+  `conv2d`, grouped/depthwise convolution, `conv_transpose2d`, `max_pool2d`,
+  `avg_pool2d`, and embedding lookup; they are not CUDA/cuDNN replacements.
 - Vision covers local image-classification utilities, metrics, visualization,
   and compact CNNs. It is not an OpenCV replacement and does not include
   pretrained model zoos, detection/segmentation training stacks, video IO, or