ferrotorch 0.3.0

PyTorch in Rust — deep learning framework built on ferray
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90

#[cfg(not(target_env = "msvc"))]
#[global_allocator]
static GLOBAL: mimalloc::MiMalloc = mimalloc::MiMalloc;

pub use ferrotorch_core::*;

/// Prelude module — import everything commonly needed.
pub mod prelude {
    pub use ferrotorch_core::*;
    pub use ferrotorch_nn::{BatchNorm2d, Dropout, LayerNorm};
    pub use ferrotorch_nn::{Conv2d, Linear, Module, Parameter, Sequential};
    pub use ferrotorch_nn::{CrossEntropyLoss, MSELoss};
    pub use ferrotorch_nn::{GELU, ReLU, SiLU, Sigmoid, Softmax, Tanh};
    pub use ferrotorch_nn::{GRU, LSTM};
    pub use ferrotorch_optim::{Adam, AdamW, Optimizer, Sgd};
}

/// Neural network modules and layers.
pub mod nn {
    pub use ferrotorch_nn::*;
}

/// Optimizers and learning rate schedulers.
pub mod optim {
    pub use ferrotorch_optim::*;
}

/// Data loading, datasets, samplers, and transforms.
pub mod data {
    pub use ferrotorch_data::*;
}

/// Computer vision models, datasets, and transforms.
pub mod vision {
    pub use ferrotorch_vision::*;
}

/// Training loop, Learner, callbacks, and metrics.
#[cfg(feature = "train")]
pub mod train {
    pub use ferrotorch_train::*;
}

/// Model serialization: ONNX export, PyTorch import, safetensors, GGUF.
#[cfg(feature = "serialize")]
pub mod serialize {
    pub use ferrotorch_serialize::*;
}

/// JIT tracing, IR graph, optimization passes, and code generation.
#[cfg(feature = "jit")]
pub mod jit {
    pub use ferrotorch_jit::*;
}

/// Probability distributions for sampling and variational inference.
#[cfg(feature = "distributions")]
pub mod distributions {
    pub use ferrotorch_distributions::*;
}

/// Performance profiling and Chrome trace export.
#[cfg(feature = "profiler")]
pub mod profiler {
    pub use ferrotorch_profiler::*;
}

/// Model hub for downloading and caching pretrained models.
#[cfg(feature = "hub")]
pub mod hub {
    pub use ferrotorch_hub::*;
}

/// CUDA GPU backend with PTX kernels and cuBLAS.
#[cfg(feature = "gpu")]
pub mod gpu {
    pub use ferrotorch_gpu::*;
}

/// Portable GPU compute via CubeCL (CUDA + WGPU + ROCm).
#[cfg(feature = "cubecl")]
pub mod cubecl {
    pub use ferrotorch_cubecl::*;
}

/// Distributed training: DDP, collective ops, TCP backend.
#[cfg(feature = "distributed")]
pub mod distributed {
    pub use ferrotorch_distributed::*;
}