Clean up and update build

.env

@@ -1,11 +1,10 @@
-MODEL_DIR=/media/teamgroup/models
-MODEL=dolphin-2.6-mistral-7b-dpo-exl2
-EXTENSIONS_DIR=/home/me/git_repos/extensions/
-CHARACTERS_DIR=/media/teamgroup/characters
-SETTINGS_DIR=/media/teamgroup/settings
-COMMITHASH=1edf321362b3f5b3c2486fca8fd3931628c8b669
+MODEL_DIR=/media/sandisk/models
+MODEL=Hyperion-3.0-Mistral-7B-alpha-exl2-8_0
+EXTENSIONS_DIR=~./git_repos/extensions/
+CHARACTERS_DIR=./characters
+SETTINGS_DIR=./settings
 LOGS_DIR=./logs
-DATASETS_DIR=/media/teamgroup/loras/datasets
-LORAS_DIR=/media/teamgroup/loras
+DATASETS_DIR=./datasets
+LORAS_DIR=./loras
 PRESETS_DIR=./presets
 PROMPTS_DIR=./prompts

.gitignore

@@ -3,3 +3,9 @@ docker_build.log
 nohup.out
 presets/
 prompts/
+characters/
+settings/
+datasets/
+loras/
+dev_build.sd
+release.sh

Dockerfile

@@ -2,7 +2,7 @@
 FROM nvidia/cuda:12.1.1-devel-ubuntu22.04 as builder
 
 # Using conda to transfer python env from builder to runtime later
-COPY --from=continuumio/miniconda3:23.5.2-0 /opt/conda /opt/conda
+COPY --from=continuumio/miniconda3:23.10.0-1 /opt/conda /opt/conda
 ENV PATH=/opt/conda/bin:$PATH
 
 # Update base image
@@ -12,32 +12,30 @@ RUN apt-get update && apt-get upgrade -y \
     && mkdir -p /etc/OpenCL/vendors && echo "libnvidia-opencl.so.1" > /etc/OpenCL/vendors/nvidia.icd
 
 # Create new conda environment
-RUN conda create -y -n textgen python=3.11.5
+RUN conda create -y -n textgen python=3.11.8
 SHELL ["conda", "run", "-n", "textgen", "/bin/bash", "-c"]
 
 ENV CUDA_DOCKER_ARCH=all
 
+ARG CACHEBUST=1
+
 # Installing torch and ninja
-RUN pip3 install torch torchvision torchaudio xformers --index-url https://download.pytorch.org/whl/cu121
+RUN pip3 install torch==2.2.* torchvision==0.17.* torchaudio==2.2.* xformers --index-url https://download.pytorch.org/whl/cu121
 
 RUN pip3 install ninja packaging sentence-transformers
 
 ARG clone_arg
 ARG commit
-ARG CACHEBUST=1
 
 # Pulling latest text-generation-webui branch
 RUN git clone https://github.com/oobabooga/text-generation-webui/ $clone_arg \
     && cd text-generation-webui \
     && if [ -n "$commit" ]; then git checkout $commit; fi \
-    && pip3 install -r requirements.txt
+    && pip3 install -r requirements.txt --upgrade
 
 # Install all the extension requirements
 RUN bash -c 'for i in text-generation-webui/extensions/*/requirements.txt ; do pip3 install -r $i ; done'
 
-# Prepare cache for faster first time runs -- removed until its fixed
-#RUN cd /text-generation-webui/extensions/openai/ && python3 cache_embedding_model.py
-
 RUN conda clean -afy
 
 # Using ubuntu 22.04 for runtime

README.md

@@ -32,7 +32,7 @@ Next, build the image. This will use the latest from the text-generation-webui r
 
 ```sh
 cd text-generation-webui-docker
-docker build -t text-generation-webui-docker:latest .
+docker build -t text-gen-ui-gpu:latest .
 ```
 
 Alternatively you can run the build script, which will also just pull latest unless you have a COMMITHASH variable in a .env file or specify a commit while running the script like below:
@@ -42,12 +42,10 @@ cd text-generation-webui-docker
 ./build.sh 7a3ca2c68f1ca49ac4e4b62f016718556fd3805c
 ```
 
-(note it will call it noneabove1182/text-generation-webui-docker since that's what I use, and tag it as well, may change this later)
-
 # Running the image with docker run
 
 ```sh
-docker run --gpus all -p 7860:7860 -v /media/teamgroup/models:/models -v ./logs:/logs text-generation-webui-docker:latest --model WizardLM-13B-V1.1-GPTQ --chat --listen --listen-port 7860
+docker run --gpus all -p 7860:7860 -v /media/teamgroup/models:/models -v ./logs:/logs text-gen-ui-gpu:latest --model WizardLM-13B-V1.1-GPTQ --chat --listen --listen-port 7860
 ```
 
 # Running the image with docker compose

build.sh

@@ -6,4 +6,4 @@ LATEST_COMMIT=$(curl -s $REPO_URL | grep 'sha' | cut -d\" -f4 | head -n 1)
 
 echo $LATEST_COMMIT
 
-docker build --build-arg commit="$LATEST_COMMIT" -t noneabove1182/text-gen-ui-gpu .
+docker build --build-arg commit="$LATEST_COMMIT" -t text-gen-ui-gpu .

dev_build.sh

@@ -1,3 +1,3 @@
 #!/bin/bash
 
-docker build --build-arg CACHEBUST=$(date +%s) --build-arg clone_arg="--branch dev" -t noneabove1182/text-gen-ui-gpu .
+docker build --build-arg CACHEBUST=$(date +%s) --build-arg clone_arg="--branch gradio4" -t text-gen-ui-gpu .

docker-compose.yml

@@ -18,6 +18,8 @@ services:
       - '5000:5000'
     image: 'noneabove1182/text-gen-ui-gpu:latest'
     environment:
+      - PUID=1000
+      - PGID=1000
       - OPENEDAI_EMBEDDING_DEVICE=cuda
     ulimits:
       memlock: -1
@@ -30,7 +32,7 @@ services:
         reservations:
           devices:
             - driver: nvidia
-              count: 1
+              device_ids: [ '1', '0' ]
               capabilities: [ gpu ]
     command:
       [