Upload 4 files

README.md

@@ -4,7 +4,7 @@ emoji: 👀😻
 colorFrom: red
 colorTo: purple
 sdk: gradio
-sdk_version: 5.24.0
+sdk_version: 5.31.0
 app_file: app.py
 pinned: true
 license: apache-2.0

genimage.py

@@ -1,9 +1,16 @@
 import spaces
+import gradio as gr
 import torch
-import gc
+import gc, os, uuid, json
+from PIL import PngImagePlugin
 
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
+if os.getenv("SPACES_ZERO_GPU", None):
+    torch.backends.cudnn.deterministic = True
+    torch.backends.cudnn.benchmark = False
+    torch.backends.cuda.matmul.allow_tf32 = True
+    torch.set_float32_matmul_precision("high") # https://pytorch.org/blog/accelerating-generative-ai-3/
 
 
 def load_pipeline():
@@ -12,7 +19,7 @@ def load_pipeline():
         "John6666/rae-diffusion-xl-v2-sdxl-spo-pcm",
         custom_pipeline="lpw_stable_diffusion_xl",
         #custom_pipeline="nyanko7/sdxl_smoothed_energy_guidance",
-        torch_dtype=torch.float16,
+        torch_dtype=torch.bfloat16 if torch.cuda.is_available() else torch.float32,
     )
     pipe.to("cpu")
     return pipe
@@ -48,10 +55,6 @@ def token_auto_concat_embeds(pipe, positive, negative):
 
 
 def save_image(image, metadata, output_dir):
-    import os
-    import uuid
-    import json
-    from PIL import PngImagePlugin
     filename = str(uuid.uuid4()) + ".png"
     os.makedirs(output_dir, exist_ok=True)
     filepath = os.path.join(output_dir, filename)
@@ -66,8 +69,8 @@ pipe = load_pipeline()
 
 
 @torch.inference_mode()
-@spaces.GPU
-def generate_image(prompt, neg_prompt):
+@spaces.GPU(duration=10)
+def generate_image(prompt, neg_prompt, progress=gr.Progress(track_tqdm=True)):
     pipe.to(device)
     prompt += ", anime, masterpiece, best quality, very aesthetic, absurdres"
     neg_prompt += ", bad hands, bad feet, lowres, (bad), text, error, fewer, extra, missing, worst quality, jpeg artifacts, low quality, watermark, unfinished, displeasing, oldest, early, chromatic aberration, signature, extra digits, artistic error, username, scan, [abstract], photo, deformed, disfigured, low contrast, photo, deformed, disfigured, low contrast"
@@ -91,7 +94,6 @@ def generate_image(prompt, neg_prompt):
             output_type="pil",
             clip_skip=2,
         ).images
-        pipe.to("cpu")
         if images:
             image_paths = [
                 save_image(image, metadata, "./outputs")
@@ -100,9 +102,9 @@ def generate_image(prompt, neg_prompt):
         return image_paths
     except Exception as e:
         print(e)
-        pipe.to("cpu")
         return []
     finally:
+        pipe.to("cpu")
         torch.cuda.empty_cache()
         gc.collect()
 

llmenv.py

@@ -14,6 +14,7 @@ llm_models = {
     "mn-12b-lyra-v2a1-q5_k_m.gguf": ["HalleyStarbun/MN-12B-Lyra-v2a1-Q5_K_M-GGUF", MessagesFormatterType.CHATML],
     "L3-8B-Tamamo-v1.i1-Q5_K_M.gguf": ["mradermacher/L3-8B-Tamamo-v1-i1-GGUF", MessagesFormatterType.LLAMA_3],
     "MN-Chinofun-12B-2.i1-Q4_K_M.gguf": ["mradermacher/MN-Chinofun-12B-2-i1-GGUF", MessagesFormatterType.MISTRAL],
+    "Shirayukihime-12B.Q4_K_M.gguf": ["mradermacher/Shirayukihime-12B-GGUF", MessagesFormatterType.CHATML],
     "Dolphin-5.1-7b.Q5_K_M.gguf": ["mradermacher/Dolphin-5.1-7b-GGUF", MessagesFormatterType.MISTRAL],
     "Captain-Eris-Diogenes_Twighlight-V0.420-12B.i1-Q4_K_M.gguf": ["mradermacher/Captain-Eris-Diogenes_Twighlight-V0.420-12B-i1-GGUF", MessagesFormatterType.MISTRAL],
     "Captain-Eris-Diogenes_Twilight-V0.420-12B.i1-Q4_K_M.gguf": ["mradermacher/Captain-Eris-Diogenes_Twilight-V0.420-12B-i1-GGUF", MessagesFormatterType.MISTRAL],
@@ -144,6 +145,9 @@ llm_models = {
     "MagnaMellRei-v1-12B.Q4_K_M.gguf": ["mradermacher/MagnaMellRei-v1-12B-GGUF", MessagesFormatterType.CHATML],
     "HMS-Slerp-12B.Q4_K_M.gguf": ["mradermacher/HMS-Slerp-12B-GGUF", MessagesFormatterType.CHATML],
     "SnowElf-12B-v2.Q4_K_M.gguf": ["mradermacher/SnowElf-12B-v2-GGUF", MessagesFormatterType.CHATML],
+    "KnowledgeCore-12B.Q4_K_M.gguf": ["mradermacher/KnowledgeCore-12B-GGUF", MessagesFormatterType.CHATML],
+    "BianCang-Qwen2.5-14B-Instruct.Q4_K_M.gguf": ["mradermacher/BianCang-Qwen2.5-14B-Instruct-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "Gemma-2-Llama-Swallow-9b-pt-v0.1.Q4_K_M.gguf": ["mradermacher/Gemma-2-Llama-Swallow-9b-pt-v0.1-GGUF", MessagesFormatterType.ALPACA],
     "Protestant-Christian-Bible-Expert-v2.0-12B.Q4_K_M.gguf": ["mradermacher/Protestant-Christian-Bible-Expert-v2.0-12B-GGUF", MessagesFormatterType.MISTRAL],
     "openbuddy-qwen2.5llamaify-14b-v23.1-200k.i1-Q4_K_M.gguf": ["mradermacher/openbuddy-qwen2.5llamaify-14b-v23.1-200k-i1-GGUF", MessagesFormatterType.OPEN_CHAT],
     "Trinas_Nectar-8B-model_stock.i1-Q4_K_M.gguf": ["mradermacher/Trinas_Nectar-8B-model_stock-i1-GGUF", MessagesFormatterType.MISTRAL],
@@ -158,6 +162,36 @@ llm_models = {
     #"": ["", MessagesFormatterType.OPEN_CHAT],
     #"": ["", MessagesFormatterType.CHATML],
     #"": ["", MessagesFormatterType.PHI_3],
+    "TutorRL-7B.Q5_K_M.gguf": ["mradermacher/TutorRL-7B-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "Omega-Darker_The-Final-Directive-14B.Q4_K_M.gguf": ["mradermacher/Omega-Darker_The-Final-Directive-14B-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "Student-rl_Qwen2.5-7B-Instruct_v0.0.1.Q5_K_M.gguf": ["mradermacher/Student-rl_Qwen2.5-7B-Instruct_v0.0.1-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "Contrl-Stheno-v1-8B.Q5_K_M.gguf": ["mradermacher/Contrl-Stheno-v1-8B-GGUF", MessagesFormatterType.LLAMA_3],
+    "fluxiia_14b.Q4_K_M.gguf": ["mradermacher/fluxiia_14b-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "JwenJwen1-1.5b.Q5_K_M.gguf": ["mradermacher/JwenJwen1-1.5b-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "SecGPT-14B-Reasoning.Q4_K_M.gguf": ["mradermacher/SecGPT-14B-Reasoning-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "Dirty-Muse-Writer-v01-Uncensored-Erotica-NSFW.Q4_K_M.gguf": ["mradermacher/Dirty-Muse-Writer-v01-Uncensored-Erotica-NSFW-GGUF", MessagesFormatterType.ALPACA],
+    "SearchR1-nq_hotpotqa_train-qwen2.5-7b-em-ppo.Q4_K_M.gguf": ["mradermacher/SearchR1-nq_hotpotqa_train-qwen2.5-7b-em-ppo-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "ForgottenMaid-12B-v2.Q4_K_M.gguf": ["mradermacher/ForgottenMaid-12B-v2-GGUF", MessagesFormatterType.MISTRAL],
+    "Socratic-Qwen2.5-7B-v2.Q5_K_M.gguf": ["mradermacher/Socratic-Qwen2.5-7B-v2-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "EsotericSage-12B.i1-Q4_K_M.gguf": ["mradermacher/EsotericSage-12B-i1-GGUF", MessagesFormatterType.MISTRAL],
+    "praxis-bookwriter-llama3.1-8b-sft.Q5_K_M.gguf": ["mradermacher/praxis-bookwriter-llama3.1-8b-sft-GGUF", MessagesFormatterType.LLAMA_3],
+    "LinearWriter-12B.Q4_K_M.gguf": ["mradermacher/LinearWriter-12B-GGUF", MessagesFormatterType.MISTRAL],
+    "Qwen2.5-CoderX-14B-v0.5.Q4_K_M.gguf": ["mradermacher/Qwen2.5-CoderX-14B-v0.5-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "Violet_MagCap-Rebase-12B.Q4_K_M.gguf": ["mradermacher/Violet_MagCap-Rebase-12B-GGUF", MessagesFormatterType.MISTRAL],
+    "VeriReason-Qwen2.5-7b-SFT-Reasoning.i1-Q5_K_M.gguf": ["mradermacher/VeriReason-Qwen2.5-7b-SFT-Reasoning-i1-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "Emilia-Multislerp-12B.Q4_K_M.gguf": ["mradermacher/Emilia-Multislerp-12B-GGUF", MessagesFormatterType.CHATML],
+    "TaskRouter-1.5B.Q5_K_M.gguf": ["mradermacher/TaskRouter-1.5B-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "NeonMaid-12B-v2.Q4_K_M.gguf": ["mradermacher/NeonMaid-12B-v2-GGUF", MessagesFormatterType.CHATML],
+    "Orihime-Gutenberg-12B.Q4_K_M.gguf": ["mradermacher/Orihime-Gutenberg-12B-GGUF", MessagesFormatterType.CHATML],
+    "SearchSimulation_3B.Q5_K_M.gguf": ["mradermacher/SearchSimulation_3B-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "SearchSimulation_14B.Q4_K_M.gguf": ["mradermacher/SearchSimulation_14B-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "NeonMaid-12B.Q4_K_M.gguf": ["mradermacher/NeonMaid-12B-GGUF", MessagesFormatterType.CHATML],
+    "Chief_chain_generator-Qwen2.5-7B-Instruct.Q5_K_M.gguf": ["mradermacher/Chief_chain_generator-Qwen2.5-7B-Instruct-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "Emotion_inferencer-Qwen2.5-7B-Instruct.Q5_K_M.gguf": ["mradermacher/Emotion_inferencer-Qwen2.5-7B-Instruct-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "ForgottenMaid-12B.Q4_K_M.gguf": ["mradermacher/ForgottenMaid-12B-GGUF", MessagesFormatterType.MISTRAL],
+    "AceCoder-Qwen2.5-Coder-7B-Ins-V1.1.Q5_K_M.gguf": ["mradermacher/AceCoder-Qwen2.5-Coder-7B-Ins-V1.1-GGUF", MessagesFormatterType.OPEN_CHAT],
+    "TheDrummer_Rivermind-Lux-12B-v1-Q4_K_M.gguf": ["bartowski/TheDrummer_Rivermind-Lux-12B-v1-GGUF", MessagesFormatterType.MISTRAL],
+    "Qwen2.5-GenX-14B.Q4_K_M.gguf": ["mradermacher/Qwen2.5-GenX-14B-GGUF", MessagesFormatterType.OPEN_CHAT],
     "vanilla-cn-roleplay-0.1.Q5_K_M.gguf": ["mradermacher/vanilla-cn-roleplay-0.1-GGUF", MessagesFormatterType.OPEN_CHAT],
     "EsotericLight-12B.Q4_K_M.gguf": ["mradermacher/EsotericLight-12B-GGUF", MessagesFormatterType.CHATML],
     "llama-3-Daredevil-Mahou-8B.Q5_K_M.gguf": ["mradermacher/llama-3-Daredevil-Mahou-8B-GGUF", MessagesFormatterType.LLAMA_3],

requirements.txt

@@ -3,15 +3,11 @@ huggingface_hub
 hf_xet
 hf_transfer
 scikit-build-core
-#https://github.com/abetlen/llama-cpp-python/releases/download/v0.2.90-cu124/llama_cpp_python-0.2.90-cp310-cp310-linux_x86_64.whl
 https://github.com/abetlen/llama-cpp-python/releases/download/v0.3.4-cu124/llama_cpp_python-0.3.4-cp310-cp310-linux_x86_64.whl
 git+https://github.com/Maximilian-Winter/llama-cpp-agent
 pybind11>=2.12
 torch==2.4.0
 torchvision
-torchao>=0.9.0
-triton
-xformers
 accelerate
 transformers<=4.48.3
 optimum[onnxruntime]
@@ -22,4 +18,4 @@ rapidfuzz
 wrapt-timeout-decorator
 opencv-python
 numpy<2
-pydantic==2.10.6
+pydantic==2.10.6