Skip to content
PMetal

Easy API

The easy module provides high-level builders that wrap PMetal’s full pipeline into ergonomic one-liner APIs.

Fine-Tuning

Section titled “Fine-Tuning”
use pmetal::easy;


let result = easy::finetune("Qwen/Qwen3-0.6B", "train.jsonl")
    .lora(16, 32.0)           // rank, alpha
    .learning_rate(2e-4)
    .epochs(3)
    .batch_size(4)
    .output("./output")
    .run()
    .await?;


println!("Final loss: {}", result.final_loss);
println!("Total steps: {}", result.total_steps);

Preference Optimization

Section titled “Preference Optimization”
// DPO
let result = easy::dpo("Qwen/Qwen3-0.6B", "preferences.jsonl")
    .dpo_beta(0.1)
    .reference_model("Qwen/Qwen3-0.6B")
    .run()
    .await?;


// SimPO
let result = easy::simpo("Qwen/Qwen3-0.6B", "preferences.jsonl")
    .run()
    .await?;


// ORPO
let result = easy::orpo("Qwen/Qwen3-0.6B", "preferences.jsonl")
    .run()
    .await?;


// KTO
let result = easy::kto("Qwen/Qwen3-0.6B", "feedback.jsonl")
    .run()
    .await?;

Inference

Section titled “Inference”
// Single generation
let output = easy::infer("Qwen/Qwen3-0.6B")
    .temperature(0.7)
    .max_tokens(256)
    .lora("./output/lora_weights.safetensors")
    .generate("What is 2+2?")
    .await?;


println!("{output}");

Streaming Inference

Section titled “Streaming Inference”
easy::infer("Qwen/Qwen3-0.6B")
    .temperature(0.7)
    .generate_streaming("Tell me a story", |delta| {
        print!("{delta}");
        true // return false to stop early
    })
    .await?;

Available Builders

Section titled “Available Builders”
BuilderDescription
easy::finetune()SFT with LoRA/QLoRA/DoRA
easy::dpo()Direct Preference Optimization
easy::simpo()Simple Preference Optimization
easy::orpo()Odds-Ratio Preference Optimization
easy::kto()Kahneman-Tversky Optimization
easy::infer()Inference with optional LoRA

See Also

Section titled “See Also”