Update app.py

app.py

@@ -1,10 +1,8 @@
 """`Plaguekind/Minimax-H3` — the PlagueKind V1.5 ComfyUI workflow for MiniMax-H3, as a Space.
-
 The candidate repository holds no weights: it is a ComfyUI graph over `Comfy-Org/MiniMax-H3`, so what is
 reproduced here is the *graph*, on the `MiniMaxAI/MiniMax-H3` diffusers checkpoint. See `pk_workflow.py` for the
 node-by-node mapping; the short version is euler + `linear_quadratic` at 15 steps, FSR RCAS sharpening at 0.3, and
 FILM 2x frame interpolation to 48 fps.
-
 Deployment is the split one the unquantized MiniMax-H3 needs: 195.9 GiB of bfloat16 does not fit under a Space's
 150 GB storage quota, so the 62.14 GiB Qwen3-VL text encoder runs in a separate Space
 (`multimodalart/qwen3vl-conditioner`) that this one calls per request, and this Space holds the 61.73 GiB
@@ -82,7 +80,6 @@ DEFAULT_STEPS = 15
 
 def snap_frames(seconds: float) -> int:
     """The frame count MiniMax-H3's video VAE can decode: the next `17 * n + 5` at 24 fps.
-
     Identical to the workflow's `ComfyMathExpression`,
     `max(5, round(a*24)) + (5 - (max(5, round(a*24)) % 17)) % 17` — 5 s is 124 frames, i.e. 5.167 s.
     """
@@ -123,7 +120,6 @@ def status() -> str:
 
 def load_models() -> str | None:
     """Load the denoising half at startup, plus FILM.
-
     `MiniMaxH3GeneratorBlocks` declares `transformer`, `vae`, `audio_vae`, the two schedulers and `video_processor`,
     so `load_components` fetches exactly those subfolders — `text_encoder/` and `transformer_ref/` are never
     touched. Both autoencoders carry `_keep_in_fp32_modules` over every module and stay float32: a bfloat16 audio
@@ -138,31 +134,15 @@ def load_models() -> str | None:
     try:
         import torch
         from diffusers import ComponentsManager
+
         from h3_split_blocks import MiniMaxH3GeneratorBlocks
-    
+
         lower_duration_floor()
-    
         manager = ComponentsManager()
         blocks = MiniMaxH3GeneratorBlocks()
-    
-        print(
-            f"[gen] loading {[c.name for c in blocks.expected_components]} from {MODEL_REPO} ...",
-            flush=True,
-        )
-    
-        pipe = blocks.init_pipeline(
-            MODEL_REPO,
-            components_manager=manager,
-            collection="h3",
-        )
-    
+        print(f"[gen] loading {[c.name for c in blocks.expected_components]} from {MODEL_REPO} ...", flush=True)
+        pipe = blocks.init_pipeline(MODEL_REPO, components_manager=manager, collection="h3")
         pipe.load_components(dtype=torch.bfloat16)
-        # Load TenStrip fine-tuned checkpoint into the transformer
-        from huggingface_hub import hf_hub_download
-        from safetensors.torch import load_model
-        finetune_path = hf_hub_download(repo_id="TenStrip/10Eros-Max", filename="10Eros_Max_h3_fl2va_bf16_test2_pruned.safetensors")
-        load_model(pipe.transformer, finetune_path, strict=False, device="cpu")
-        pipe.transformer.to(torch.bfloat16)
         pipe.transformer.set_attention_backend(ATTENTION)
 
         # Still startup, still free: an AoTI package carries no weights and opens its archive lazily inside the GPU
@@ -300,7 +280,6 @@ def _generate(
     seed,
 ):
     """The only thing on GPU time: the denoise loop, the two decoders and the workflow's post chain.
-
     The mp4 is muxed here rather than in the caller: a `@spaces.GPU` return crosses a process boundary by pickling,
     and a 2x-interpolated 124-frame clip is several hundred MB of frames against a few MB of h264.
     """
@@ -502,19 +481,16 @@ def _fit_keyframe(image_path, current_canvas):
 
 load_models()
 
-INTRO = """# PlagueKind · 10Eros-Max · MiniMax-H3
-
+INTRO = """# PlagueKind · MiniMax-H3
 <div align="center">
   <a href="https://huggingface.co/Plaguekind/Minimax-H3" target="_blank" rel="noopener"><strong>[ workflow ]</strong></a> &nbsp;
   <a href="https://huggingface.co/MiniMaxAI/MiniMax-H3" target="_blank" rel="noopener"><strong>[ model ]</strong></a> &nbsp;
   <a href="https://github.com/PlagueKind/Comfyui-PlagueKind-Nodes" target="_blank" rel="noopener"><strong>[ nodes ]</strong></a>
-    <a href="https://huggingface.co/TenStrip/10Eros-Max" target="_blank" rel="noopener"><strong>[ 10Eros-Max ]</strong></a>
 </div>
-
 **MiniMax-H3** is a 33B parameter video generation model that produces video and a fully synchronized soundtrack
 (ambience, foley, speech) in one pass. **PlagueKind's V1.5 workflow** is a tuning of it: euler on a
 `linear_quadratic` sigma grid at 15 steps, FSR **RCAS** sharpening at 0.3, and **FILM** 2x frame interpolation to
-48 fps. Text-to-video, first frame, last frame, or both. TenStrip's **10Eros-Max** is an experimental finetune of MiniMax-H3.
+48 fps. Text-to-video, first frame, last frame, or both.
 """
 
 CSS = """
@@ -523,7 +499,7 @@ CSS = """
 .status p {font-size: 0.8rem; opacity: 0.65; text-align: center;}
 """
 
-with gr.Blocks(title="PlagueKind · 10Eros-Max · MiniMax-H3") as demo:
+with gr.Blocks(title="PlagueKind · MiniMax-H3") as demo:
     gr.Markdown(INTRO)
     gr.Markdown(status(), elem_classes="status")
 
@@ -652,7 +628,6 @@ with gr.Blocks(title="PlagueKind · 10Eros-Max · MiniMax-H3") as demo:
             """
 `Plaguekind/Minimax-H3` ships no weights — it is a ComfyUI graph over `Comfy-Org/MiniMax-H3`. Its nodes map onto
 this Space as:
-
 | ComfyUI node | widget | here |
 |---|---|---|
 | `KSamplerSelect` | `euler` | MiniMax-H3's only sampler; the checkpoint is CFG-distilled, so one forward per step and no negative prompt |
@@ -664,13 +639,11 @@ this Space as:
 | `CreateVideo` | fps `24 * 2` | 48 fps output |
 | `RTXVideoSuperResolution` | 2x `ULTRA` | **not reproduced** |
 | `PathchSageAttentionKJ` | `sageattn_qk_int8_pv_fp8_cuda++` | cuDNN fused attention |
-
 Two deliberate deviations. `RTXVideoSuperResolution` is NVIDIA's NGX super-resolution, shipped as a driver-level
 Windows/RTX component with no Linux Python path, so the 2x upscale is missing — pick a larger **Target dimension**
 instead of upscaling a small one. And SageAttention's `qk_int8_pv_fp8_cuda++` kernel is not built for this pool's
 sm120 cards, so attention runs cuDNN's fused kernel, which is the fastest available here and is numerically the
 faithful one (SageAttention is a quantized approximation).
-
 One upgrade: the workflow loads `minimax_h3_fl2va_pruned_int8_convrot.safetensors` and a
 `qwen3vl_32b_..._int8_convrot` text encoder because that is what fits a consumer card. This Space runs both
 **unquantized bfloat16**, off `MiniMaxAI/MiniMax-H3`, with the 62 GiB text encoder in a
@@ -683,4 +656,4 @@ One upgrade: the workflow loads `minimax_h3_fl2va_pruned_int8_convrot.safetensor
 
 if __name__ == "__main__":
     # `theme` and `css` belong to `launch()` from Gradio 6.0 on; on `Blocks` they warn and are ignored.
-    demo.launch(show_error=True, theme=gr.themes.Citrus(), css=CSS)
+    demo.launch(show_error=True, theme=gr.themes.Citrus(), css=CSS)