feat(tune): stream generate_lora output via generate_streaming

[ohdearquant]

What

Stream tokens from generate_lora via generate_streaming, with sampler controls.

Why

generate_lora (the LoRA-adapter generation binary) previously buffered the full output. It now uses the streaming entry point from PR-1 so adapter-driven generation streams token-by-token like chat_metal.

Files

• crates/tune/src/bin/generate_lora.rs (+119/-24)

Verification

cargo build --release -p lattice-tune --bin generate_lora --features safetensors,inference-hook clean. This binary's compile break (E0599: generate_streaming not in scope) was one of the two regressions that motivated the PR-G gate. Built green in the integrated-tree gate.

Base

Stacked on pr/eng-1-streaming-detok (depends on generate_streaming). Review/merge PR-1 first.

Series

Part of the PR #193 engine-slice (finest split). All engine code lands on main; the macOS app surfaces a subset (Models + Chat) for v0.0.1.

[github-actions]

E2E Parity Report

PASS: all 3 prompts match within first 3 tokens

Prompt	Agreement	First Diff	HF tok/s	Lattice tok/s	Verdict
The capital of France is	3/15	pos 3	0.5	4.2	PASS
In the year 2024, artificial intelligence	10/15	pos 9	0.4	3.7	PASS
`def fibonacci(n):

if n <= 1:
    return n
return` | 15/15 | none | 0.4 | 2.9 | PASS |

The capital of France is

• HF: Paris.
  The capital of France is Paris.
  The capital of France
• Lattice: Paris.
  A: Yes, the capital of France is Paris.

In the year 2024, artificial intelligence

• HF: (AI) has become a significant part of the global economy. It is
• Lattice: (AI) has become a significant part of our daily lives. From personal

def fibonacci(n): if n <= 1: return n return

• HF: fibonacci(n-1) + fibonacci(n-2)

print(fib

• Lattice: fibonacci(n-1) + fibonacci(n-2)

print(fib