bump cache busters to 33 for debug build

.env

@@ -1,7 +1,18 @@
-B200 Node
-IP: `45.76.247.107`
-user: `root`
-password: `6)Jr)B@dcX[mN?dx`
-folder with the weights: `/root/nvidia-meeting`
+# B200 Node
+IP=45.76.247.107
+USER=root
+PASSWORD=6)Jr)B@dcX[mN?dx
 
-This repo: `https://sweetapi.com/biondizzle/deepseek-v4-quant.git` (always use this http to pull since its a public repo)
+# B200 Paths
+DOCKER_COMPOSE=/root/nvidia-meeting/docker-compose.yml
+WEIGHTS=/root/nvidia-meeting/DeepSeek-V4-Pro-NVFP4
+PATCHES=/root/nvidia-meeting/patches
+REPO=/root/nvidia-meeting/deepseek-v4-quant
+
+# Docker
+IMAGE=atl.vultrcr.com/vllm/vllm-dsv4-nvfp4:latest
+CONTAINER_NAME=nvidia-meeting-vllm-1
+
+# Remotes
+REPO_URL=https://sweetapi.com/biondizzle/deepseek-v4-quant.git
+DEEPGEMM_URL=https://sweetapi.com/biondizzle/DeepGEMM.git

Dockerfile

@@ -8,7 +8,7 @@ RUN apt-get update && apt-get install -y git screen cmake && rm -rf /var/lib/apt
 
 # Clone and build DeepGEMM with NVFP4 mega_moe kernel
 # CACHE_BUSTER: increment to force fresh clone
-RUN git clone -b nvfp4-mega-moe https://sweetapi.com/biondizzle/DeepGEMM.git /root/DeepGEMM && CACHE_BUSTER=32
+RUN git clone -b nvfp4-mega-moe https://sweetapi.com/biondizzle/DeepGEMM.git /root/DeepGEMM && CACHE_BUSTER=33
 
 # Build DeepGEMM with proper CUDA/NVRTC paths
 ENV CPATH="/usr/local/lib/python3.12/dist-packages/flashinfer/data/cutlass/include:/usr/local/lib/python3.12/dist-packages/nvidia/cu13/include:/usr/local/cuda-13.0/include:${CPATH}"
@@ -19,7 +19,7 @@ RUN ln -sf /usr/local/lib/python3.12/dist-packages/nvidia/cu13/lib/libnvrtc.so.1
 RUN cd /root/DeepGEMM && python3 setup.py build_ext --inplace
 
 # Bust cache for patch changes — ARG before COPY ensures layer invalidation
-ARG PATCH_CACHE_BUSTER=32
+ARG PATCH_CACHE_BUSTER=33
 # Copy our DeepSeek V4 patch over vLLM's model file
 COPY patches/deepseek_v4.py /usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/deepseek_v4.py