tensorstudio 1.8.0__tar.gz → 1.10.0__tar.gz

{tensorstudio-1.8.0 → tensorstudio-1.10.0}/CHANGELOG.md

@@ -2,6 +2,41 @@
 
 ## Unreleased
 
+## 1.10.0 - 2026-07-07
+
+- Completed the ordered Training And Project Workflows roadmap section as one
+  release batch.
+- Added `ArrayDataset`, tensor/array/image-folder dataset factories,
+  deterministic train/validation splitting, and dataset metadata summaries.
+- Added `tensorstudio.metrics` with regression, classification, and multilabel
+  metrics for small supervised workflows.
+- Added trainer validation loops, scheduler stepping, callback context support,
+  learning-rate logging, CSV logging, checkpoint callbacks, and early stopping.
+- Added JSON, TOML, and YAML project config loading plus deterministic seeding
+  across TensorStudio, NumPy, and Python random.
+- Expanded full checkpoints with scheduler and epoch state and added
+  `resume_checkpoint()` for continuing training runs.
+- Added generated regression, classification, and vision project templates.
+- Expanded tests and docs for project workflows, metrics, dataset creation,
+  callbacks, configs, templates, and checkpoint resume.
+
+## 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

{tensorstudio-1.8.0 → tensorstudio-1.10.0}/CMakeLists.txt

@@ -1,70 +1,70 @@
-cmake_minimum_required(VERSION 3.18)
-
-project(tensorstudio VERSION 1.8.0 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 1.10.0 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-1.8.0 → tensorstudio-1.10.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: tensorstudio
-Version: 1.8.0
+Version: 1.10.0
 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
@@ -42,6 +42,8 @@ Project-URL: Repository, https://github.com/imattas/TensorStudio
 Project-URL: Issues, https://github.com/imattas/TensorStudio/issues
 Requires-Python: >=3.10
 Requires-Dist: numpy>=1.24
+Requires-Dist: PyYAML>=6
+Requires-Dist: tomli>=2.0; python_version < "3.11"
 Provides-Extra: dev
 Requires-Dist: pytest>=8; extra == "dev"
 Requires-Dist: pytest-cov>=5; extra == "dev"
@@ -71,13 +73,18 @@ 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 `1.8.0` is a CPU-only stable API foundation with native C++
+TensorStudio `1.10.0` 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, seeded random distributions, and a hardened eager autograd
-lifecycle. It is eager-only, intentionally compact, and not a replacement for
-mature ML frameworks.
+lifecycle. The neural-network layer now includes grouped/depthwise/1D/
+transposed convolution, normalization layers, embeddings, richer activations,
+initializers, additional losses, and model summaries. The project layer adds
+dataset factories, deterministic train/validation splitting, metrics, callbacks,
+multi-format configs, checkpoint resume helpers, and starter project templates.
+It is eager-only, intentionally compact, and not a replacement for mature ML
+frameworks.
 
 ## Install
 
@@ -287,6 +294,23 @@ print(model.state_dict().keys())
 print(model.parameter_count())
 ```
 
+The 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
@@ -321,17 +345,22 @@ optimizer.step()
 print(ts.vision.accuracy(model(x), target))
 ```
 
-## DataLoader
+## Data And Metrics
 
 ```python
 import tensorstudio as ts
-from tensorstudio.data import DataLoader, TensorDataset
+from tensorstudio.data import DataLoader, from_arrays, train_val_split
 
-dataset = TensorDataset(ts.arange(6).reshape((6, 1)), ts.arange(6))
-loader = DataLoader(dataset, batch_size=2, shuffle=True, seed=42)
+dataset = from_arrays([[0.0], [1.0], [2.0], [3.0]], [[1.0], [3.0], [5.0], [7.0]])
+train_data, val_data = train_val_split(dataset, val_fraction=0.25, seed=42)
+loader = DataLoader(train_data, batch_size=2, shuffle=True, seed=42)
 
 for features, targets in loader:
     print(features.shape, targets.shape)
+
+prediction = ts.tensor([[0.2, 0.8], [0.7, 0.3]])
+target = ts.tensor([1, 0], dtype="int64")
+print(ts.metrics.accuracy(prediction, target))
 ```
 
 The v1 DataLoader is intentionally single-process so it works cleanly on
@@ -339,24 +368,44 @@ Windows without multiprocessing setup.
 
 ## Projects And Training
 
-`tensorstudio.project` provides project folders, JSON config, reusable trainers,
-safe NPZ weight files, and trusted full checkpoints:
+`tensorstudio.project` provides project folders, JSON/TOML/YAML config loading,
+deterministic seeding, reusable trainers, validation loops, callbacks, safe NPZ
+weight files, trusted full checkpoints, and generated starter templates:
 
 ```python
 import tensorstudio as ts
 from tensorstudio import nn, optim
-from tensorstudio.data import DataLoader, TensorDataset
-from tensorstudio.project import Project, ProjectConfig, Trainer, save_state_dict
+from tensorstudio.data import DataLoader, from_arrays, train_val_split
+from tensorstudio.project import (
+    CSVLogger,
+    CheckpointCallback,
+    LrLogger,
+    Project,
+    ProjectConfig,
+    Trainer,
+    save_state_dict,
+    seed_everything,
+)
 
-x = ts.tensor([[0.0], [1.0], [2.0], [3.0]])
-y = ts.tensor([[1.0], [3.0], [5.0], [7.0]])
+seed_everything(7)
+dataset = from_arrays([[0.0], [1.0], [2.0], [3.0]], [[1.0], [3.0], [5.0], [7.0]])
+train_data, val_data = train_val_split(dataset, val_fraction=0.25, seed=7)
 
 model = nn.Linear(1, 1)
-loader = DataLoader(TensorDataset(x, y), batch_size=2)
-trainer = Trainer(model, optim.SGD(model.parameters(), lr=0.05), nn.MSELoss())
+optimizer = optim.SGD(model.parameters(), lr=0.05)
+trainer = Trainer(model, optimizer, nn.MSELoss(), metric_fn=ts.metrics.mean_squared_error)
 project = Project("runs/linear", ProjectConfig(name="linear-regression", seed=7))
 
-history = trainer.fit(loader, epochs=50)
+history = trainer.fit(
+    DataLoader(train_data, batch_size=2),
+    epochs=50,
+    validation_loader=DataLoader(val_data, batch_size=1),
+    callbacks=[
+        LrLogger(),
+        CSVLogger(project.logs_dir / "history.csv"),
+        CheckpointCallback(project.checkpoints_dir / "best.tsmodel", save_best_only=True),
+    ],
+)
 save_state_dict(model, project.checkpoint_path("weights"))
 print(history.last)
 ```
@@ -391,11 +440,11 @@ Run the loose local regression thresholds with:
 python benchmark_all.py --check-thresholds
 ```
 
-On one Windows CPython 3.10 development run reporting `1.8.0`, with
+On one Windows CPython 3.10 development run reporting `1.10.0`, 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 47 local cases and lost 51. The
+available on that machine; TensorStudio won 38 local cases and lost 60. 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
@@ -407,15 +456,15 @@ Snapshot from that local run:
 
 | operation | shape | TensorStudio | NumPy | JAX CPU dispatch | TS vs NumPy | TS vs JAX |
 |---|---:|---:|---:|---:|---:|---:|
-| `sigmoid` | `(32,)` | 0.0155 ms | 0.0044 ms | 0.0761 ms | 0.2860x | 4.9227x |
-| `mean` | `(32,)` | 0.0158 ms | 0.0082 ms | 0.0126 ms | 0.5215x | 0.7968x |
-| `sum_axis1` | `(16, 16)` | 0.0161 ms | 0.0028 ms | 0.0136 ms | 0.1714x | 0.8417x |
-| `chain_relu` | `(128,)` | 0.0861 ms | 0.0056 ms | 0.0944 ms | 0.0646x | 1.0971x |
-| `matmul` | `(256, 256)` | 2.1900 ms | 0.4380 ms | 0.2610 ms | 0.2000x | 0.1192x |
-| `conv2d_3x3_padding1` | `(1, 1, 8, 8)` | 0.1995 ms | 1.2390 ms | 0.1086 ms | 6.2118x | 0.5443x |
-| `max_pool2d_2x2` | `(1, 1, 16, 16)` | 0.0286 ms | 0.1684 ms | n/a | 5.8921x | n/a |
-| `avg_pool2d_2x2` | `(1, 1, 16, 16)` | 0.0286 ms | 0.5569 ms | n/a | 19.4803x | n/a |
-| `elementwise_backward` | `(1024,)` | 2.7758 ms | n/a | n/a | n/a | n/a |
+| `sigmoid` | `(32,)` | 0.0150 ms | 0.0046 ms | 0.0723 ms | 0.3056x | 4.8092x |
+| `mean` | `(32,)` | 0.0154 ms | 0.0078 ms | 0.0111 ms | 0.5097x | 0.7198x |
+| `sum_axis1` | `(16, 16)` | 0.0159 ms | 0.0030 ms | 0.0129 ms | 0.1904x | 0.8133x |
+| `chain_relu` | `(128,)` | 0.0859 ms | 0.0056 ms | 0.0888 ms | 0.0647x | 1.0337x |
+| `matmul` | `(256, 256)` | 2.0731 ms | 0.4017 ms | 0.2497 ms | 0.1938x | 0.1204x |
+| `conv2d_3x3_padding1` | `(1, 1, 8, 8)` | 0.2028 ms | 1.3615 ms | 0.1219 ms | 6.7128x | 0.6012x |
+| `max_pool2d_2x2` | `(1, 1, 16, 16)` | 0.0279 ms | 0.1740 ms | n/a | 6.2464x | n/a |
+| `avg_pool2d_2x2` | `(1, 1, 16, 16)` | 0.0275 ms | 0.5906 ms | n/a | 21.4893x | n/a |
+| `elementwise_backward` | `(1024,)` | 2.7546 ms | n/a | n/a | n/a | n/a |
 
 Speedup is `competitor median / TensorStudio median`, so values above `1.0x`
 favor TensorStudio.
@@ -516,8 +565,9 @@ 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-1.8.0 → tensorstudio-1.10.0}/README.md

@@ -7,13 +7,18 @@
 TensorStudio is a compact C++ tensor and autograd engine with a Python API for
 learning, experimentation, and lightweight ML workloads.
 
-TensorStudio `1.8.0` is a CPU-only stable API foundation with native C++
+TensorStudio `1.10.0` 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, seeded random distributions, and a hardened eager autograd
-lifecycle. It is eager-only, intentionally compact, and not a replacement for
-mature ML frameworks.
+lifecycle. The neural-network layer now includes grouped/depthwise/1D/
+transposed convolution, normalization layers, embeddings, richer activations,
+initializers, additional losses, and model summaries. The project layer adds
+dataset factories, deterministic train/validation splitting, metrics, callbacks,
+multi-format configs, checkpoint resume helpers, and starter project templates.
+It is eager-only, intentionally compact, and not a replacement for mature ML
+frameworks.
 
 ## Install
 
@@ -223,6 +228,23 @@ print(model.state_dict().keys())
 print(model.parameter_count())
 ```
 
+The 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
@@ -257,17 +279,22 @@ optimizer.step()
 print(ts.vision.accuracy(model(x), target))
 ```
 
-## DataLoader
+## Data And Metrics
 
 ```python
 import tensorstudio as ts
-from tensorstudio.data import DataLoader, TensorDataset
+from tensorstudio.data import DataLoader, from_arrays, train_val_split
 
-dataset = TensorDataset(ts.arange(6).reshape((6, 1)), ts.arange(6))
-loader = DataLoader(dataset, batch_size=2, shuffle=True, seed=42)
+dataset = from_arrays([[0.0], [1.0], [2.0], [3.0]], [[1.0], [3.0], [5.0], [7.0]])
+train_data, val_data = train_val_split(dataset, val_fraction=0.25, seed=42)
+loader = DataLoader(train_data, batch_size=2, shuffle=True, seed=42)
 
 for features, targets in loader:
     print(features.shape, targets.shape)
+
+prediction = ts.tensor([[0.2, 0.8], [0.7, 0.3]])
+target = ts.tensor([1, 0], dtype="int64")
+print(ts.metrics.accuracy(prediction, target))
 ```
 
 The v1 DataLoader is intentionally single-process so it works cleanly on
@@ -275,24 +302,44 @@ Windows without multiprocessing setup.
 
 ## Projects And Training
 
-`tensorstudio.project` provides project folders, JSON config, reusable trainers,
-safe NPZ weight files, and trusted full checkpoints:
+`tensorstudio.project` provides project folders, JSON/TOML/YAML config loading,
+deterministic seeding, reusable trainers, validation loops, callbacks, safe NPZ
+weight files, trusted full checkpoints, and generated starter templates:
 
 ```python
 import tensorstudio as ts
 from tensorstudio import nn, optim
-from tensorstudio.data import DataLoader, TensorDataset
-from tensorstudio.project import Project, ProjectConfig, Trainer, save_state_dict
+from tensorstudio.data import DataLoader, from_arrays, train_val_split
+from tensorstudio.project import (
+    CSVLogger,
+    CheckpointCallback,
+    LrLogger,
+    Project,
+    ProjectConfig,
+    Trainer,
+    save_state_dict,
+    seed_everything,
+)
 
-x = ts.tensor([[0.0], [1.0], [2.0], [3.0]])
-y = ts.tensor([[1.0], [3.0], [5.0], [7.0]])
+seed_everything(7)
+dataset = from_arrays([[0.0], [1.0], [2.0], [3.0]], [[1.0], [3.0], [5.0], [7.0]])
+train_data, val_data = train_val_split(dataset, val_fraction=0.25, seed=7)
 
 model = nn.Linear(1, 1)
-loader = DataLoader(TensorDataset(x, y), batch_size=2)
-trainer = Trainer(model, optim.SGD(model.parameters(), lr=0.05), nn.MSELoss())
+optimizer = optim.SGD(model.parameters(), lr=0.05)
+trainer = Trainer(model, optimizer, nn.MSELoss(), metric_fn=ts.metrics.mean_squared_error)
 project = Project("runs/linear", ProjectConfig(name="linear-regression", seed=7))
 
-history = trainer.fit(loader, epochs=50)
+history = trainer.fit(
+    DataLoader(train_data, batch_size=2),
+    epochs=50,
+    validation_loader=DataLoader(val_data, batch_size=1),
+    callbacks=[
+        LrLogger(),
+        CSVLogger(project.logs_dir / "history.csv"),
+        CheckpointCallback(project.checkpoints_dir / "best.tsmodel", save_best_only=True),
+    ],
+)
 save_state_dict(model, project.checkpoint_path("weights"))
 print(history.last)
 ```
@@ -327,11 +374,11 @@ Run the loose local regression thresholds with:
 python benchmark_all.py --check-thresholds
 ```
 
-On one Windows CPython 3.10 development run reporting `1.8.0`, with
+On one Windows CPython 3.10 development run reporting `1.10.0`, 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 47 local cases and lost 51. The
+available on that machine; TensorStudio won 38 local cases and lost 60. 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
@@ -343,15 +390,15 @@ Snapshot from that local run:
 
 | operation | shape | TensorStudio | NumPy | JAX CPU dispatch | TS vs NumPy | TS vs JAX |
 |---|---:|---:|---:|---:|---:|---:|
-| `sigmoid` | `(32,)` | 0.0155 ms | 0.0044 ms | 0.0761 ms | 0.2860x | 4.9227x |
-| `mean` | `(32,)` | 0.0158 ms | 0.0082 ms | 0.0126 ms | 0.5215x | 0.7968x |
-| `sum_axis1` | `(16, 16)` | 0.0161 ms | 0.0028 ms | 0.0136 ms | 0.1714x | 0.8417x |
-| `chain_relu` | `(128,)` | 0.0861 ms | 0.0056 ms | 0.0944 ms | 0.0646x | 1.0971x |
-| `matmul` | `(256, 256)` | 2.1900 ms | 0.4380 ms | 0.2610 ms | 0.2000x | 0.1192x |
-| `conv2d_3x3_padding1` | `(1, 1, 8, 8)` | 0.1995 ms | 1.2390 ms | 0.1086 ms | 6.2118x | 0.5443x |
-| `max_pool2d_2x2` | `(1, 1, 16, 16)` | 0.0286 ms | 0.1684 ms | n/a | 5.8921x | n/a |
-| `avg_pool2d_2x2` | `(1, 1, 16, 16)` | 0.0286 ms | 0.5569 ms | n/a | 19.4803x | n/a |
-| `elementwise_backward` | `(1024,)` | 2.7758 ms | n/a | n/a | n/a | n/a |
+| `sigmoid` | `(32,)` | 0.0150 ms | 0.0046 ms | 0.0723 ms | 0.3056x | 4.8092x |
+| `mean` | `(32,)` | 0.0154 ms | 0.0078 ms | 0.0111 ms | 0.5097x | 0.7198x |
+| `sum_axis1` | `(16, 16)` | 0.0159 ms | 0.0030 ms | 0.0129 ms | 0.1904x | 0.8133x |
+| `chain_relu` | `(128,)` | 0.0859 ms | 0.0056 ms | 0.0888 ms | 0.0647x | 1.0337x |
+| `matmul` | `(256, 256)` | 2.0731 ms | 0.4017 ms | 0.2497 ms | 0.1938x | 0.1204x |
+| `conv2d_3x3_padding1` | `(1, 1, 8, 8)` | 0.2028 ms | 1.3615 ms | 0.1219 ms | 6.7128x | 0.6012x |
+| `max_pool2d_2x2` | `(1, 1, 16, 16)` | 0.0279 ms | 0.1740 ms | n/a | 6.2464x | n/a |
+| `avg_pool2d_2x2` | `(1, 1, 16, 16)` | 0.0275 ms | 0.5906 ms | n/a | 21.4893x | n/a |
+| `elementwise_backward` | `(1024,)` | 2.7546 ms | n/a | n/a | n/a | n/a |
 
 Speedup is `competitor median / TensorStudio median`, so values above `1.0x`
 favor TensorStudio.
@@ -452,8 +499,9 @@ 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