Pad declared ZeroGPU duration with a cold-start buffer

Fix step-done log ordering vs text_encoder offload/pre-VAE-decode

Reduce default GPU duration for Fast mode from 30s to 15s

Add VAE decode logging: pre-decode mem snapshot, tiling prediction, peak reset

Raise VAE tile threshold to HIGH_DETAIL.max_dim (2048px) to skip tiling

Load text_encoder as int8 (bitsandbytes) when TEXT_ENCODER_INT8_REPO is set

Skip text_encoder CPU offload in fast mode; type mode as an enum

Add per-phase timing to pipeline call and text_encoder offload

Speed up fast mode: lower resolution cap, skip needless VAE tiling, fewer steps

Log exception type/repr and traceback in infer() error path

Replace vllm-flash-attn3 kernel with native SDPA cuDNN backend; wire up attn processor

Free text encoder before VAE decode and enable VAE tiling to fix OOM

Patch RMSNorm for fp8-resident weights; fix crash in prior fp8 commit

Keep transformer weights fp8-resident instead of upcasting at load

Refactor _infer_gpu into single-responsibility helper functions

Fix misleading "includes compile" log label; add first-touch instrumentation

Revert NF4 quantization: this checkpoint's weights are stored fp8-native

Quantize transformer to NF4 via bitsandbytes to fix root-cause OOM

Gate full-GPU fast path on slice size; it OOMs on this Space's 47GB MIG

Try full pipeline GPU residency before falling back to cpu offload

Use sequential CPU offload instead of moving full pipeline to GPU at once

Use cudaMallocAsync allocator to fix MIG GPU CUDACachingAllocator crash

Add startup environment diagnostics (_log_env)

Load model on CPU at startup to bypass slow ZeroGPU disk packing

Revert FA3 and torch.compile: incompatible with this GPU environment

Fix torch.compile crash on torch.device ConstantVariable bug

Enable FA3 attention and torch.compile for faster inference

Batch log uploads to avoid HF 256-commits/hour rate limit

Log gpu_duration and mode at infer start

Add user-adjustable GPU Duration slider to Advanced Settings

Raise ZeroGPU durations to 20s fast, 60s high-detail

Tighten ZeroGPU durations to p95-aligned values (10s fast, 30s high-detail)

Raise high-detail ZeroGPU duration to 90s (was 60s)

Use per-mode dynamic ZeroGPU duration (30s fast, 60s high-detail)

Reduce ZeroGPU duration from 120s to 60s to halve quota usage

Fix log upload dropped by moving _spawn_log outside @spaces.GPU scope

Reduce ZeroGPU usage by moving CPU work outside GPU scope

Load mode toggle JS via demo.load instead of stripped <script> tag

Add Fast / High-Detail mode toggle (768px vs 2048px output)

Add heartbeat logging during model loading

Revert "Add step-progress bar to inference loader"

Add step-progress bar to inference loader

Fix quick prompt buttons: HTML-escape JSON to avoid breaking onclick attribute

Fix f-string backslash syntax error in build_suggestion_chips_html

Load Quick Prompts from suggestions.json, replace artistic chips with realistic ones

Revert 2K/4K resolution selector (4K too slow for generation)

Add 2K/4K resolution selector to Advanced Settings

Refactor app.py: extract helpers, consolidate imports

Remove /data mount dependency and HF cache inspection

Remove local_files_only from transformer load to fix missing snapshots/