[bugfix] Fix CI failures and address review comments

CI fixes:
- Add use_embedded_guidance to SamplingConfig in fastvideo/api/schema.py
  (test_inventory_targets_exist_in_typed_schema walked request.sampling but
  field was missing from the dataclass)
- Add true_cfg_scale and use_embedded_guidance to expected_dests in
  inference_schema_parity_inventory.yaml generate section
  (test_cli_dest_inventory_matches_live_parsers found both dests on the
  live parser but not in the inventory)

Gemini/Copilot review fixes:
- flux.py: scale ts0 by *1000 when timestep is float — FLUX passes timestep
  in [0,1] so the old int(item()) produced ts0=0 for every non-zero step
- flux.py: use truthiness guard on controlnet samples instead of is not None
  to prevent ZeroDivisionError on empty-list input
- test_flux_t2i_similarity.py: restrict SSIM parametrize to TORCH_SDPA;
  FLASH_ATTN reference not yet seeded (pending Will confirmation on coverage)
- docs: update README.md and PORT_STATUS.md to reflect PASS evidence and
  committed SSIM references (TORCH_SDPA, A40, 2026-05-11)
- test_flux_dev_pipeline_smoke.py: fix function signature indent (PEP 8)

Author: Mister-Raggs

docs/design/inference_schema_parity_inventory.yaml

@@ -600,7 +600,9 @@ cli:
       - text_encoder_precisions
       - torch_compile_kwargs
       - tp_size
+      - true_cfg_scale
       - trust_remote_code
+      - use_embedded_guidance
       - use_fsdp_inference
       - vae_config.blend_num_frames
       - vae_config.load_decoder

fastvideo/api/schema.py

@@ -120,6 +120,7 @@ class SamplingConfig:
     guidance_scale_2: float | None = None
     guidance_rescale: float = 0.0
     true_cfg_scale: float | None = None
+    use_embedded_guidance: bool | None = None
     boundary_ratio: float | None = None
     sigmas: list[float] | None = None
 

fastvideo/models/dits/flux.py

@@ -507,7 +507,12 @@ def forward(
             get_forward_context()
             forward_context = nullcontext()
         except AssertionError:
-            ts0 = int(timestep[0].item()) if timestep.numel() > 0 else 0
+            if timestep.numel() == 0:
+                ts0 = 0
+            elif torch.is_floating_point(timestep):
+                ts0 = int(round(timestep[0].item() * 1000))
+            else:
+                ts0 = int(timestep[0].item())
             forward_context = set_forward_context(current_timestep=ts0, attn_metadata=None)
 
         with forward_context:
@@ -542,7 +547,7 @@ def forward(
                     image_rotary_emb=image_rotary_emb,
                     joint_attention_kwargs=jkwargs,
                 )
-                if controlnet_block_samples is not None:
+                if controlnet_block_samples:
                     interval = len(self.transformer_blocks) / len(controlnet_block_samples)
                     interval = int(math.ceil(interval))
                     if controlnet_blocks_repeat:
@@ -558,7 +563,7 @@ def forward(
                     image_rotary_emb=image_rotary_emb,
                     joint_attention_kwargs=jkwargs,
                 )
-                if controlnet_single_block_samples is not None:
+                if controlnet_single_block_samples:
                     interval = len(self.single_transformer_blocks) / len(controlnet_single_block_samples)
                     interval = int(math.ceil(interval))
                     hidden_states = hidden_states + controlnet_single_block_samples[idx // interval]

fastvideo/tests/ssim/test_flux_t2i_similarity.py

@@ -93,7 +93,7 @@
     reason="FLUX T2I SSIM test requires CUDA",
 )
 @pytest.mark.parametrize("prompt", TEST_PROMPTS)
-@pytest.mark.parametrize("attention_backend_name", ["TORCH_SDPA", "FLASH_ATTN"])
+@pytest.mark.parametrize("attention_backend_name", ["TORCH_SDPA"])
 @pytest.mark.parametrize("model_id", list(FLUX_MODEL_TO_PARAMS.keys()))
 def test_flux_t2i_similarity(
     prompt: str,

tests/local_tests/flux/PORT_STATUS.md

@@ -33,7 +33,7 @@ Strict-load evidence: PASS — component loader test runs strict load of all com
 | Item | Status |
 |---|---|
 | SSIM test | written (fastvideo/tests/ssim/test_flux_t2i_similarity.py) |
-| Reference images committed | not yet — pending SSIM seeding run |
+| Reference images committed | TORCH_SDPA committed (A40, 2026-05-11); FLASH_ATTN pending |
 
 ## DiT Parity Notes
 

tests/local_tests/flux/README.md

@@ -44,7 +44,7 @@ FASTVIDEO_ATTENTION_BACKEND=TORCH_SDPA \
 pytest tests/local_tests/pipelines/test_flux_dev_pipeline_parity.py -vs
 ```
 
-Status: requires weights — pending local run (see PORT_STATUS.md).
+Status: PASS on A40 (2026-05-11) — requires weights; not run in CI.
 
 ## DiT Parity Test
 
@@ -57,7 +57,7 @@ pytest fastvideo/tests/transformers/test_flux.py -vs
 ```
 
 Status: requires weights — not run in CI.
-Pass evidence: pending (see PORT_STATUS.md).
+Pass evidence: PASS recorded in PORT_STATUS.md (A40, 2026-05-11).
 
 ## SSIM Regression Test
 
@@ -70,4 +70,4 @@ FLUX_T2I_MODEL_DIR=official_weights/FLUX.1-dev \
 pytest fastvideo/tests/ssim/test_flux_t2i_similarity.py -vs
 ```
 
-Status: reference images not yet committed — see PORT_STATUS.md.
+Status: seeded reference images committed (TORCH_SDPA, A40, 2026-05-11) — see PORT_STATUS.md.

tests/local_tests/pipelines/test_flux_dev_pipeline_smoke.py

@@ -26,7 +26,8 @@
 
 
 def test_flux_dev_pipeline_short_run_finite_output(
-        monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
     import torch
 
     if not torch.cuda.is_available():