Update Dockerfile

Dockerfile

@@ -83,9 +83,29 @@ RUN pip install --index-url https://download.pytorch.org/whl/cu128 \
 # ---------------- Toolchain, Triton, FA2 (sem bnb) ----------------
 RUN pip install packaging ninja cmake pybind11 scikit-build cython hf_transfer numpy==1.24.4
 
-# Triton 3.x (sem triton.ops)
+
+
+# ---------------- Toolchain base ----------------
+RUN pip install packaging ninja cmake pybind11 scikit-build cython hf_transfer numpy==1.24.4
+
+# ---------------- Instalação via wheels (HF Hub) ----------------
+# Triton, Apex e Q8 pré-compilados (Python 3.10, manylinux x86_64)
+# Use sempre --no-cache-dir e --no-build-isolation para evitar recompilar
 RUN pip uninstall -y triton || true && \
-    pip install -v --no-build-isolation triton==3.4.0
+    pip install --no-cache-dir --no-build-isolation \
+      https://huggingface.co/carlex3321/aduc-sdr/resolve/main/triton-3.5.0%2Bgit51021fb2-cp310-cp310-linux_x86_64.whl && \
+    pip install --no-cache-dir --no-build-isolation \
+      https://huggingface.co/carlex3321/aduc-sdr/resolve/main/apex-0.1-cp310-cp310-linux_x86_64.whl && \
+    pip install --no-cache-dir --no-build-isolation \
+      https://huggingface.co/carlex3321/aduc-sdr/resolve/main/q8_kernels-0.0.5-cp310-cp310-linux_x86_64.whl
+
+# ---------------- FlashAttention 2.8.x ----------------
+# Mantém fallback em cascata caso a versão mais nova falhe
+RUN pip install --no-cache-dir flash-attn==2.8.3 --no-build-isolation || \
+    pip install --no-cache-dir flash-attn==2.8.2 --no-build-isolation || \
+    pip install --no-cache-dir flash-attn==2.8.1 --no-build-isolation || \
+    pip install --no-cache-dir flash-attn==2.8.0.post2 --no-build-isolation
+    
 
 # FlashAttention 2.8.x
 RUN pip install flash-attn==2.8.3 --no-build-isolation || \
@@ -94,10 +114,10 @@ RUN pip install flash-attn==2.8.3 --no-build-isolation || \
     pip install flash-attn==2.8.0.post2 --no-build-isolation
 
 # Diffusers/Transformers estáveis (sem dev)
-RUN pip install --no-cache-dir diffusers==0.31.0 transformers==4.44.2 accelerate==0.34.2 omegaconf==2.3.0
+RUN pip install --no-cache-dir transformers==4.44.2 accelerate==0.34.2 omegaconf==2.3.0
 
 # Opcional: seu fork de otimizações
-# RUN pip install -U git+https://github.com/carlex22/diffusers-aduc-sdr
+RUN pip install -U git+https://github.com/carlex22/diffusers-aduc-sdr
 
 # ---------------- Repositórios auxiliares ----------------
 RUN git clone https://github.com/bytedance-seed/VINCIE.git  && \