[Admin maintenance] Support new ZeroGPU hardware

Thank you so much for having shared this Space with the community on this demo. We have upgraded the ZeroGPU infra-structure to run on modern blackwell architecture.
For that, we need to upgrade your demo to support that. This PR fixes your demo to work with the new architecture. As this is something we broke on our end, we may merge this PR autonomously. If this breaks unexpectedly or brings unintended consequences, feel free to revert, modify or otherwise. Any issues you can email apolinario@huggingface.co

README.md

@@ -4,7 +4,7 @@ emoji: 😻
 colorFrom: purple
 colorTo: pink
 sdk: gradio
-sdk_version: 4.40.0
+sdk_version: 5.50.0
 app_file: app.py
 pinned: false
 ---

app.py

@@ -1,10 +1,23 @@
-import torch
 import spaces
+import torch
+
+# diffusers 0.29.2 imports FLAX_WEIGHTS_NAME from transformers.utils, removed in
+# transformers 5.x. Shim it back before importing diffusers.
+import transformers.utils as _tu
+if not hasattr(_tu, "FLAX_WEIGHTS_NAME"):
+    _tu.FLAX_WEIGHTS_NAME = "flax_model.msgpack"
+
 from diffusers import StableDiffusionPipeline, DDIMScheduler, AutoencoderKL
-from transformers import AutoFeatureExtractor
+from transformers import CLIPImageProcessor
 from ip_adapter.pipeline_stable_diffusion_extra_cfg import StableDiffusionPipelineCFG
 from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 
+# transformers 5.x's _finalize_model_loading reads self.all_tied_weights_keys,
+# which diffusers 0.29.2's StableDiffusionSafetyChecker (older PreTrainedModel)
+# does not define. Provide an empty default so loading succeeds.
+if not hasattr(StableDiffusionSafetyChecker, "all_tied_weights_keys"):
+    StableDiffusionSafetyChecker.all_tied_weights_keys = {}
+
 from ip_adapter.ip_adapter_instruct import IPAdapterInstruct
 from huggingface_hub import hf_hub_download
 import gradio as gr
@@ -16,7 +29,7 @@ image_encoder_path = "laion/CLIP-ViT-H-14-laion2B-s32B-b79K"
 ip_ckpt = hf_hub_download(repo_id="CiaraRowles/IP-Adapter-Instruct", filename="ip-adapter-instruct-sd15.bin", repo_type="model")
 
 safety_model_id = "CompVis/stable-diffusion-safety-checker"
-safety_feature_extractor = AutoFeatureExtractor.from_pretrained(safety_model_id)
+safety_feature_extractor = CLIPImageProcessor.from_pretrained(safety_model_id)
 safety_checker = StableDiffusionSafetyChecker.from_pretrained(safety_model_id)
 
 device = "cuda"
@@ -47,7 +60,7 @@ ip_model = IPAdapterInstruct(pipe, image_encoder_path, ip_ckpt, device,dtypein=t
 
 cv2.setNumThreads(1)
 
-@spaces.GPU(enable_queue=True)
+@spaces.GPU
 def generate_image(images, prompt, negative_prompt,instruct_query, scale, nfaa_negative_prompt, progress=gr.Progress(track_tqdm=True)):
     faceid_all_embeds = []
     first_iteration = True

requirements.txt

@@ -1,9 +1,9 @@
 diffusers==0.29.2
 transformers
+torchvision
 accelerate
 safetensors
 einops
 onnxruntime-gpu
-spaces==0.19.4
 opencv-python
 git+https://github.com/unity-research/IP-Adapter-Instruct.git