Stop Inductor compilation progress from leaking into the Gradio UI (#33)

gr.Progress(track_tqdm=True) monkey-patches tqdm globally, so it was
picking up TorchInductor's own "Inductor Compilation" progress bars
(triggered by torch.compile on the upscale model, and by the quantized
transformer/text_encoder) regardless of their own disable flag. Drop
track_tqdm and report the generation stage's progress explicitly via
callback_on_step_end instead, matching the interpolation/upscale stages.

app.py

@@ -161,11 +161,15 @@ def run_inference(
     seed,
     frame_multiplier,
     upscale_output,
-    progress=gr.Progress(track_tqdm=True),
+    progress=gr.Progress(),
 ):
     print_infer_start(prompt, negative_prompt, seed, steps, guidance_scale, frame_multiplier, upscale_output)
     t_start = _time.perf_counter()
 
+    def _report_generation_progress(pipe_, step_index, timestep, callback_kwargs):
+        progress(0.7 * (step_index + 1) / int(steps), desc=f"Generating ({step_index + 1}/{int(steps)} steps)...")
+        return callback_kwargs
+
     print_stage_start("generation")
     t0 = _time.perf_counter()
     try:
@@ -180,6 +184,7 @@ def run_inference(
             num_inference_steps=int(steps),
             generator=torch.Generator(device="cuda").manual_seed(seed),
             output_type="np",
+            callback_on_step_end=_report_generation_progress,
         )
     except Exception as e:
         print_stage_error("generation", e)
@@ -246,7 +251,7 @@ def generate_video(
     randomize_seed,
     frame_multiplier,
     upscale_output,
-    progress=gr.Progress(track_tqdm=True),
+    progress=gr.Progress(),
 ):
     if input_image is None:
         raise gr.Error("Please upload an input image.")