update docs
This commit is contained in:
+5
-7
@@ -126,9 +126,7 @@ All models implement the `GenerateModel` trait:
|
||||
|
||||
```rust
|
||||
pub trait GenerateModel {
|
||||
// Initialize model with weights
|
||||
fn init(model_path: &str, config: Option<GenerationConfig>) -> Result<Self>;
|
||||
|
||||
|
||||
// Generate response
|
||||
fn generate(&mut self, prompt: &str, params: GenerationParams) -> Result<String>;
|
||||
|
||||
@@ -167,8 +165,8 @@ This provides:
|
||||
|
||||
- `audio_utils.rs` - Audio processing (WAV, MP3)
|
||||
- `image_utils.rs` - Image processing (resize, encode/decode)
|
||||
- `download.rs` - Model downloading with retry logic
|
||||
- `common.rs` - Common utilities and constants
|
||||
- `tensor_utils.rs` - Tensor utility methods
|
||||
- `mod.rs` - Common utilities and constants
|
||||
|
||||
## Design Patterns
|
||||
|
||||
@@ -269,7 +267,7 @@ Flash Attention optimizes long-sequence processing:
|
||||
|
||||
Enable with:
|
||||
```bash
|
||||
cargo build --features "cuda,flash-attn"
|
||||
cargo build --features cuda,flash-attn
|
||||
```
|
||||
|
||||
### Memory-Mapped Tensors
|
||||
@@ -294,7 +292,7 @@ Dynamic precision selection based on hardware:
|
||||
|
||||
- No external API calls after model download
|
||||
- No telemetry or data collection
|
||||
- Full control over your data
|
||||
- Data remains entirely on the local system
|
||||
|
||||
### Memory Safety
|
||||
|
||||
|
||||
Reference in New Issue
Block a user