The Windows ROCm torchao import stub satisfies find_spec and even lets
from torchao.quantization import quantize_ succeed, but its quantize_ is a
no-op: auto would pick int8, leave the transformer dense, and disable
compile as if it were quantized. has_functional_torchao imports the exact
symbols the int8 path uses and rejects the stub via its sentinel; both the
auto picker and the /info advertised modes now use it
- base_precision="auto" only picks int8 when torchao is importable (the int8
quantize has no runtime fallback, unlike fp8); otherwise the middle band falls
back to nf4. Threaded as a parameter so the policy stays pure.
- The dense-mode validation (prequant base / bf16 compute) now applies only to
DiT families: sdxl ignores base_precision entirely, so a leftover value can no
longer fail an SDXL run. The mode-name validity check still runs everywhere.
New base_precision config for the DiT trainers: nf4 (unchanged default) |
bf16 | int8 | fp8 | auto, advertised per family + per machine through
/api/train/diffusion/info (precision_modes, recommended_precision,
supports_compile) so the UI can gate the selector.
- bf16: dense transformer + regional torch.compile (auto-armed). The
measured speed mode: 2.3x nf4 on FLUX (1.81 -> 4.12 steps/s), 2.6x on
Z-Image (2.5 -> 6.38 steps/s) on B200, at dense-weight VRAM
(FLUX 24.7 GB / Z-Image 13.6 GB peak vs 10.4 / 4.7 for nf4).
- int8: torchao weight-only int8 on the frozen base, quantized AFTER
add_adapter (quantizing first trips peft 0.18's TorchaoLoraLinear,
which is incompatible with the torchao 0.16 config API). Runs eager:
inductor rejects the int8 subclass training graph (aliased subclass
outputs), so compile is force-disabled for it.
- fp8: torchao convert_to_float8_training on the frozen linears
(filter skips lora_ modules, proj_out, non-divisible-by-16 dims,
pad_inner_dim), applied after add_adapter, compile auto-armed.
Works and round-trips, but measured SLOWER than compiled bf16 at
LoRA-training shapes (FLUX 3.15 vs 4.12 steps/s; Z-Image similar),
so it is an explicit opt-in and auto never picks it.
- auto: free VRAM (measured before load) + dense-size table -> bf16
when it fits with headroom, int8 in the middle band, else nf4.
Prequant bnb repos always resolve to nf4; dense modes on them are
rejected at validation with a pointer to the family's dense base.
Two crashes found and fixed along the way:
- The cuDNN SDPA backend's training graph fails on the FLUX attention
shapes (torch 2.10 + cu130, B200): mha_graph.execute errors, then the
context degrades into illegal memory accesses. The perf-flag guard now
pins flash/mem-efficient SDPA for the run (mathematically equivalent,
snapshot/restored). nf4 escaped it by routing attention differently.
- Regional compile now uses dynamic=True (the inference layer's proven
default): dynamic=False specialisation fused a gemm_and_bias epilogue
that failed with CUBLAS_STATUS_EXECUTION_FAILED on the FLUX training
graph; dynamic=True is also faster (Z-Image 3.84 -> 6.38 steps/s).
Verified: 98 backend tests green (new test_diffusion_base_precision.py:
validation, auto policy table, fp8 filter, compile gating, /info fields);
per-mode 40-step runs on FLUX + Z-Image with loss means inside the nf4
envelope and adapter round-trip generation through the normal LoRA path
for bf16-, fp8-, and int8-trained adapters.
