Expand description
rustensor
Small Rust-native checkpoint format for storing model metadata and tensor
weights in a single verifiable .rten file.
This crate does not train models, does not run inference and does not aim to replace ONNX, GGUF or SafeTensors.
Re-exports§
pub use crate::checkpoint::MAX_LAYER_COUNT;pub use crate::checkpoint::ModelCheckpoint;pub use crate::error::Result;pub use crate::error::RustensorError;pub use crate::format::CHECKSUM_SIZE;pub use crate::format::FORMAT_VERSION;pub use crate::format::HEADER_SIZE;pub use crate::format::MAGIC_BYTES;pub use crate::format::MAX_PAYLOAD_SIZE;pub use crate::format::RtenHeader;pub use crate::metadata::ModelMetadata;pub use crate::reader::load_checkpoint;pub use crate::weights::MAX_DIMENSIONS;pub use crate::weights::TensorDType;pub use crate::weights::TensorData;pub use crate::weights::WeightLayer;pub use crate::writer::save_checkpoint;pub use crate::integrity::sha256_hex;pub use crate::integrity::sha256_payload;pub use crate::integrity::verify_checksum;
Modules§
- checkpoint
- Checkpoint model definitions.
- error
- Error types used by rustensor.
- format
- RTEN binary format definitions.
- integrity
- Integrity helpers for RTEN payloads.
- metadata
- Model metadata definitions.
- reader
- RTEN file reader.
- weights
- Tensor weight layer definitions.
- writer
- RTEN file writer.