Update app.py

app.py

@@ -421,23 +421,19 @@ def encode_remote(prompt, references, canvas, num_frames, rewrite_prompt=False):
 
 
 @spaces.GPU(duration=get_duration, size=GPU_SIZE)
+@spaces.GPU(duration=get_duration, size=GPU_SIZE)
 def _generate(prompt_embeds, text_token_tags, references, height, width, num_frames, steps, seed, loras=()):
     """The only thing on GPU time: the two reference encoders, the packed-sequence denoise loop and the decoders.
-
     References cross as paths and are decoded here; only the three generated outputs come back. A `@spaces.GPU`
     argument crosses a process boundary by pickling, a 5 s 1344x768 reference video is 370 MB of expanded frames, and
     the full `PipelineState` still holds the packed latents and the rotary grid on the card.
-
     The adapters are attached here rather than in the caller: `spaces` runs this body in its own worker, so the
     transformer the request sees is the one that has to carry them.
     """
     import torch
-
     if PLACEMENT == "lazy":
         PIPE.to("cuda")
-
-    apply_loras(PIPE.transformer_ref, loras or ())
-
+        apply_loras(PIPE.transformer_ref, loras or ())
     state = PIPE(
         prompt_embeds=prompt_embeds.to("cuda"),
         text_token_tags=text_token_tags,
@@ -452,8 +448,6 @@ def _generate(prompt_embeds, text_token_tags, references, height, width, num_fra
 
 
 def generate(
-    # Every parameter after `prompt` has a default, and the newest ones sit at the end, so a positional API client
-    # written against an older signature keeps working.
     prompt,
     image_1=None,
     audio_path=None,
@@ -472,11 +466,9 @@ def generate(
     steps=28,
     seed=42,
     upsample=False,
-    *lora_fields,
     progress=gr.Progress(track_tqdm=True),
 ):
-    """One request. The LoRA fields are last and default to empty, so a positional API client that predates them is
-    unaffected. `lora_fields` arrives as `reference, strength, reference, strength, ...`."""
+    """One request."""
     if LOAD_ERROR:
         raise gr.Error(LOAD_ERROR)
     if PIPE is None:
@@ -493,8 +485,6 @@ def generate(
     derivable = len(audio_bearing(references)) == 1
     requested = 0 if (match and derivable) else snap_frames(duration)
 
-    loras, lora_labels = collect_loras(lora_fields, progress)
-
     progress(0.0, desc="Upsampling the prompt ..." if upsample else "Reading the prompt and references ...")
     conditioned = time.time()
     try:
@@ -518,7 +508,7 @@ def generate(
     progress(0.1, desc=f"Generating {num_frames / FPS:.1f} s at {width}x{height} ...")
     started = time.time()
     frames, audio, sampling_rate = _generate(
-        prompt_embeds, text_token_tags, references, height, width, num_frames, steps, seed, loras
+        prompt_embeds, text_token_tags, references, height, width, num_frames, steps, seed
     )
     generate_seconds = time.time() - started
 
@@ -532,8 +522,7 @@ def generate(
         f"({num_frames / FPS:.3f} s), {int(steps)} steps · conditioner {condition_seconds:.0f}s "
         f"({plan['num_text_tokens']} tokens{', upsampled' if refined else ''}) · "
         f"denoise + decode {generate_seconds:.0f}s "
-        f"({generate_seconds / int(steps):.1f} s/step) · seed {int(seed)}"
-        f"{' · LoRA ' + ', '.join(lora_labels) if lora_labels else ''}",
+        f"({generate_seconds / int(steps):.1f} s/step) · seed {int(seed)}",
         flush=True,
     )
     return path, refined, gr.update(visible=bool(refined))
@@ -547,11 +536,7 @@ def generate(
 # is gone by the next visit, so a saved path would restore as a dead file rather than as the image.
 
 SETTINGS_VERSION = 1
-SETTINGS_KEYS = (
-    ["prompt", "upsample", "canvas", "match", "duration", "steps", "seed"]
-    + [f"lora_{slot + 1}" for slot in range(LORA_SLOTS)]
-    + [f"lora_{slot + 1}_scale" for slot in range(LORA_SLOTS)]
-)
+SETTINGS_KEYS = ["prompt", "upsample", "canvas", "match", "duration", "steps", "seed"]
 
 
 def save_settings(*values):
@@ -591,41 +576,9 @@ def load_settings(path):
     return updates
 
 
-def _fill_lora_slots(files, *current):
-    """Drop `.safetensors` files on the uploader and their paths land in the first free slots, so a local adapter
-    needs no typing at all."""
-    slots = list(current)
-    for path in files or []:
-        for index, value in enumerate(slots):
-            if not (value or "").strip():
-                slots[index] = path
-                break
-    return [gr.update(value=value) for value in slots]
-
-
-def _add_preset_lora(preset, *current):
-    """Fill the first free LoRA slot with a preset adapter, set its strength to 1.0 and move the steps slider to the
-    preset's recommended count.
-
-    The Turbo presets are tuned for a specific step range, so the steps slider is moved along with the slot — it is the
-    one output beyond the LoRA fields. A slot already holding the same reference is a no-op, so the button can be
-    pressed twice without duplicating, and a full set of slots is left untouched.
-    """
-    reference, steps, _ = LORA_PRESETS[preset]
-    slots = list(current[:LORA_SLOTS])
-    scales = list(current[LORA_SLOTS:])
-    if reference not in [(value or "").strip() for value in slots]:
-        for index, value in enumerate(slots):
-            if not (value or "").strip():
-                slots[index] = reference
-                scales[index] = 1.0
-                break
-    return [*slots, *scales, steps]
-
-
 load_models()
 
-INTRO = """# MiniMax-H3 Reference Custom Lora
+INTRO = """# MiniMax-H3 Reference
 
 <div align="center">
   <a href="https://huggingface.co/MiniMaxAI/MiniMax-H3" target="_blank" rel="noopener"><strong>[ model ]</strong></a> &nbsp;
@@ -636,28 +589,6 @@ INTRO = """# MiniMax-H3 Reference Custom Lora
 **MiniMax-H3** is a 33B parameter state of the art video generation model that produces video and a
 fully synchronized soundtrack (ambience, foley, speech). Bring your own subject, voice or camera move as a
 reference.
-
-This Space ships with **pre-wired Turbo LoRA presets** from
-[`larryvrh/MiniMax-H3-Turbo-Lora`](https://huggingface.co/larryvrh/MiniMax-H3-Turbo-Lora) that render joint video +
-soundtrack in **4–8 steps** instead of the usual ~20 (a ~5× speedup). It also accepts **ComfyUI-trained LoRAs** —
-adapters in the ComfyUI checkpoint naming are remapped to diffusers' module names on the fly, so a `.safetensors`
-trained against the ComfyUI base works without a separate conversion step.
-"""
-
-LORA_HELP = """Each slot takes a Hugging Face repo (`owner/repo`), a file inside one
-(`owner/repo/name.safetensors`), a file URL, or a local path — or just drop the files below. A strength of `0`
-switches a slot off without clearing it. Adapters have to be trained against the `transformer_ref/` partition.
-
-**Turbo LoRA presets** — from
-[`larryvrh/MiniMax-H3-Turbo-Lora`](https://huggingface.co/larryvrh/MiniMax-H3-Turbo-Lora), a few-step distillation that
-renders joint video + soundtrack in **4–8 steps** instead of the usual ~20 (a ~5× speedup). Pick one from the dropdown
-and **Add to a free slot** to fill a slot, set its strength to `1.0` and move the steps slider to the recommended
-count. Keep strength at `1.0`; only nudge it if a specific clip misbehaves (smear → up, over-sharp → down).
-"""
-
-SETTINGS_HELP = """Saves the prompt, the canvas, the sliders and the LoRA slots — everything typed rather than
-uploaded. Images, audio and video are not saved: gradio keeps them in a temporary folder that is gone by the next
-visit, so a saved path would come back as a dead file.
 """
 
 CSS = """
@@ -700,40 +631,6 @@ with gr.Blocks(title="MiniMax-H3 Reference Custom Lora") as demo:
                 with gr.Tab("Video"):
                     video = gr.Video(label="Motion & camera, 2–15 s. Its soundtrack comes along.")
             run = gr.Button("Generate", variant="primary")
-
-            with gr.Accordion("LoRA", open=False):
-                gr.Markdown(LORA_HELP)
-                with gr.Row():
-                    lora_preset = gr.Dropdown(
-                        label="Turbo LoRA presets",
-                        choices=list(LORA_PRESETS),
-                        value=list(LORA_PRESETS)[0],
-                        scale=4,
-                    )
-                    lora_preset_add = gr.Button("Add to a free slot", size="sm", variant="secondary", scale=1)
-                lora_references, lora_scales = [], []
-                for slot in range(LORA_SLOTS):
-                    with gr.Row():
-                        lora_references.append(
-                            gr.Textbox(label=f"LoRA {slot + 1}", placeholder="owner/repo", scale=3)
-                        )
-                        lora_scales.append(
-                            gr.Slider(
-                                label="Strength",
-                                minimum=LORA_MIN_SCALE,
-                                maximum=LORA_MAX_SCALE,
-                                step=0.05,
-                                value=1.0,
-                                scale=2,
-                            )
-                        )
-                lora_upload = gr.File(
-                    label="Drop .safetensors here to fill the slots",
-                    file_count="multiple",
-                    file_types=[".safetensors"],
-                    type="filepath",
-                )
-
             with gr.Accordion("Advanced options", open=False):
                 canvas = gr.Dropdown(label="Canvas", choices=list(CANVASES), value=DEFAULT_CANVAS)
                 match = gr.Checkbox(label="Match the reference soundtrack", value=True, visible=False)
@@ -790,16 +687,13 @@ with gr.Blocks(title="MiniMax-H3 Reference Custom Lora") as demo:
     )
 
     # Same order as `SETTINGS_KEYS`.
-    settings_fields = [prompt, upsample, canvas, match, duration, steps, seed, *lora_references, *lora_scales]
+    settings_fields = [prompt, upsample, canvas, match, duration, steps, seed]
     save.click(save_settings, settings_fields, settings_download, api_name=False)
     settings_upload.upload(load_settings, settings_upload, settings_fields, api_name=False)
-
-    # Same order as `generate`'s signature: the five leading columns first, then the remaining image slots, then the
-    # LoRA fields the `*lora_fields` tail collects.
+    # Same order as `generate`'s signature: the five leading columns first, then the remaining image slots.
     request = [
-        prompt, images[0], audio, video, canvas, *images[1:], match, duration, steps, seed, upsample, *lora_inputs
+        prompt, images[0], audio, video, canvas, *images[1:], match, duration, steps, seed, upsample
     ]
-
     run.click(generate, request, [result, upsampled, upsampled_panel], api_name="generate")