Update burn dependency from 0.16 to 0.20#1

Open

mike-marcacci wants to merge 6 commits intoTuringWorks:mainfrom

eidola-ai:burn-0-20

Commits on Feb 12, 2026

Update burn dependency from 0.16 to 0.20

mike-marcacci
and
claude
committed

Commits on Feb 17, 2026

Implement Backend::sync() to synchronize MLX GPU work

mike-marcacci
and
claude
committed
Make Mlx backend generic over float precision (f32, f16, bf16, f64)

mike-marcacci
and
claude
committed

Commits on Feb 23, 2026

Implement native MLX quantization support via mlx_rs ops

mike-marcacci
and
claude
committed
Optimize q_reshape/q_swap_dims/q_expand to avoid unnecessary dequant/requant cycles

mike-marcacci
and
claude
committed
Cast dequantize and quantized_matmul results to backend float type
mike-marcacci
committed