diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 2a54341..7c54103 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -1,7 +1,7 @@
 name: CI
 on:
   push: {branches: [main]}
-  pull_request_target:
+  pull_request_target: {paths: ['*-*/**']}
   schedule: [{cron: '0 11 * * 6'}]  # M H d m w (Sat 11:00)
 jobs:
   setup:
@@ -15,7 +15,6 @@ jobs:
     - id: list
       name: List modified models
       run: |
-        echo "debug: author_association: ${{ github.event.pull_request.author_association }}"
         if test "${{ github.event_name }}" = schedule; then
           "modified=$(ls -d *-* | sort | jq -Rsc 'split("\n")[:-1]')" >> $GITHUB_OUTPUT
         else
diff --git a/a2t-whisper/build.sh b/a2t-whisper/build.sh
index 2ae4aab..d0f7b4f 100755
--- a/a2t-whisper/build.sh
+++ b/a2t-whisper/build.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 set -e
-export VERSION=1.0.1
+export VERSION=1.0.2
 source "$(dirname "${BASH_SOURCE[0]}")/../utils.sh"
 
 build_cpu ghcr.io/premai-io/audio-to-text-whisper-tiny-cpu     tiny     ${@:1}
diff --git a/a2t-whisper/docker/cpu/Dockerfile b/a2t-whisper/docker/cpu/Dockerfile
index 6128016..163e2f9 100644
--- a/a2t-whisper/docker/cpu/Dockerfile
+++ b/a2t-whisper/docker/cpu/Dockerfile
@@ -1,31 +1,17 @@
 FROM python:3.10-slim-bullseye
-
-ARG MODEL_ID
-
 WORKDIR /usr/src/app/
 
-ENV PYTHONDONTWRITEBYTECODE 1
-ENV PYTHONUNBUFFERED 1
-
-RUN apt-get update && \
-    apt-get install -y --no-install-recommends \
-    build-essential \
-    git \
-    wget \
-    libatlas-base-dev \
-    ffmpeg
-
-COPY requirements.txt /usr/src/app/
-
-RUN pip3 install --no-cache-dir -r requirements.txt
+RUN apt update -qq && apt install -yqq --no-install-recommends \
+    build-essential git wget libatlas-base-dev ffmpeg \
+    && rm -rf /var/lib/apt/lists/*
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
 
+ARG MODEL_ID
 COPY download.py .
-
-RUN python3 download.py --model $MODEL_ID
+RUN python download.py --model $MODEL_ID
 
 COPY . .
-
 ENV MODEL_ID=$MODEL_ID
 ENV DEVICE=cpu
-
-CMD python3 main.py
+CMD python main.py
diff --git a/a2t-whisper/docker/gpu/Dockerfile b/a2t-whisper/docker/gpu/Dockerfile
index f190d5e..2d89f63 100644
--- a/a2t-whisper/docker/gpu/Dockerfile
+++ b/a2t-whisper/docker/gpu/Dockerfile
@@ -1,20 +1,15 @@
 FROM huggingface/transformers-pytorch-gpu:latest
-
-ARG MODEL_ID
-
 WORKDIR /usr/src/app/
 
-COPY requirements.txt /usr/src/app/
-
-RUN pip3 install --no-cache-dir -r requirements.txt
+RUN cd $(dirname $(which python3)) && ln -s python3 python
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
 
+ARG MODEL_ID
 COPY download.py .
-
-RUN python3 download.py --model $MODEL_ID
+RUN python download.py --model $MODEL_ID
 
 COPY . .
-
 ENV MODEL_ID=$MODEL_ID
 ENV DEVICE=cuda
-
-CMD python3 main.py
+CMD python main.py
diff --git a/cdr-replit/build.sh b/cdr-replit/build.sh
index ed25682..04a99be 100755
--- a/cdr-replit/build.sh
+++ b/cdr-replit/build.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 set -e
-export VERSION=1.0.0
+export VERSION=1.0.1
 source "$(dirname "${BASH_SOURCE[0]}")/../utils.sh"
 
 build_gpu ghcr.io/premai-io/coder-replit-code-v1-3b-gpu replit/replit-code-v1-3b ${@:1}
diff --git a/cdr-replit/docker/gpu/Dockerfile b/cdr-replit/docker/gpu/Dockerfile
index ff29261..8f52f45 100644
--- a/cdr-replit/docker/gpu/Dockerfile
+++ b/cdr-replit/docker/gpu/Dockerfile
@@ -1,21 +1,15 @@
 FROM huggingface/transformers-pytorch-gpu:4.28.1
-
-ARG MODEL_ID
-
-RUN pip install "accelerate>=0.16.0,<1"
-
 WORKDIR /usr/src/app/
 
-COPY requirements.txt ./
-
-RUN pip install --no-cache-dir -r ./requirements.txt --upgrade pip
+RUN cd $(dirname $(which python3)) && ln -s python3 python
+COPY requirements.txt .
+RUN pip install "accelerate>=0.16.0,<1"
+RUN pip install --no-cache-dir -r requirements.txt
 
+ARG MODEL_ID
 COPY download.py .
-
-RUN python3 download.py --model $MODEL_ID
+RUN python download.py --model $MODEL_ID
 
 COPY . .
-
 ENV MODEL_ID=$MODEL_ID
-
-CMD python3 main.py
+CMD python main.py
diff --git a/cdr-replit/models.py b/cdr-replit/models.py
index 24c7c0b..812ce8b 100644
--- a/cdr-replit/models.py
+++ b/cdr-replit/models.py
@@ -26,9 +26,7 @@ def generate(
             num_return_sequences=1,
             eos_token_id=cls.tokenizer.eos_token_id,
         )
-        return cls.tokenizer.decode(
-            tokens[0], skip_special_tokens=True, clean_up_tokenization_spaces=False
-        )
+        return cls.tokenizer.decode(tokens[0], skip_special_tokens=True, clean_up_tokenization_spaces=False)
 
     @classmethod
     def get_model(cls):
diff --git a/cdr-t5/build.sh b/cdr-t5/build.sh
index 57d5c25..bf862be 100755
--- a/cdr-t5/build.sh
+++ b/cdr-t5/build.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 set -e
-export VERSION=1.0.0
+export VERSION=1.0.1
 source "$(dirname "${BASH_SOURCE[0]}")/../utils.sh"
 
 build_cpu ghcr.io/premai-io/coder-codet5p-220m-py-cpu Salesforce/codet5p-220m-py ${@:1}
diff --git a/cdr-t5/docker/cpu/Dockerfile b/cdr-t5/docker/cpu/Dockerfile
index 7e8da1c..15db572 100644
--- a/cdr-t5/docker/cpu/Dockerfile
+++ b/cdr-t5/docker/cpu/Dockerfile
@@ -1,20 +1,14 @@
 FROM python:3.10-slim-bullseye
-
-ARG MODEL_ID
-
 WORKDIR /usr/src/app/
 
-COPY requirements.txt ./
-
-RUN pip install --no-cache-dir -r ./requirements.txt --upgrade pip
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
 
+ARG MODEL_ID
 COPY download.py .
-
-RUN python3 download.py --model $MODEL_ID
+RUN python download.py --model $MODEL_ID
 
 COPY . .
-
 ENV MODEL_ID=$MODEL_ID
 ENV DEVICE=cpu
-
 CMD python main.py
diff --git a/cdr-t5/models.py b/cdr-t5/models.py
index e468064..427b843 100644
--- a/cdr-t5/models.py
+++ b/cdr-t5/models.py
@@ -15,18 +15,14 @@ def generate(
         stop: str = "",
         **kwargs,
     ):
-        inputs = cls.tokenizer.encode(prompt, return_tensors="pt").to(
-            os.getenv("DEVICE", "cpu")
-        )
+        inputs = cls.tokenizer.encode(prompt, return_tensors="pt").to(os.getenv("DEVICE", "cpu"))
         outputs = cls.model.generate(inputs, max_length=max_tokens)
         return cls.tokenizer.decode(outputs[0], skip_special_tokens=True)
 
     @classmethod
     def get_model(cls):
         if cls.model is None:
-            cls.tokenizer = AutoTokenizer.from_pretrained(
-                os.getenv("MODEL_ID", "Salesforce/codet5p-220m-py")
-            )
+            cls.tokenizer = AutoTokenizer.from_pretrained(os.getenv("MODEL_ID", "Salesforce/codet5p-220m-py"))
             cls.model = T5ForConditionalGeneration.from_pretrained(
                 os.getenv("MODEL_ID", "Salesforce/codet5p-220m-py")
             ).to(os.getenv("DEVICE", "cpu"))
diff --git a/cht-dolly-v2/build.sh b/cht-dolly-v2/build.sh
index 0628bcc..94d04b2 100755
--- a/cht-dolly-v2/build.sh
+++ b/cht-dolly-v2/build.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 set -e
-export VERSION=1.0.3
+export VERSION=1.0.4
 source "$(dirname "${BASH_SOURCE[0]}")/../utils.sh"
 
 build_gpu ghcr.io/premai-io/chat-dolly-v2-12b-gpu databricks/dolly-v2-12b ${@:1}
diff --git a/cht-dolly-v2/docker/gpu/Dockerfile b/cht-dolly-v2/docker/gpu/Dockerfile
index 008ebc0..e90ec37 100644
--- a/cht-dolly-v2/docker/gpu/Dockerfile
+++ b/cht-dolly-v2/docker/gpu/Dockerfile
@@ -1,22 +1,16 @@
 FROM huggingface/transformers-pytorch-gpu:4.28.1
-
-ARG MODEL_ID
-
-RUN pip install "accelerate>=0.16.0,<1"
-
 WORKDIR /usr/src/app/
 
-COPY requirements.txt ./
-
-RUN pip install --no-cache-dir -r ./requirements.txt --upgrade pip
+RUN cd $(dirname $(which python3)) && ln -s python3 python
+COPY requirements.txt .
+RUN pip install "accelerate>=0.16.0,<1"
+RUN pip install --no-cache-dir -r requirements.txt
 
+ARG MODEL_ID
 COPY download.py .
-
-RUN python3 download.py --model $MODEL_ID
+RUN python download.py --model $MODEL_ID
 
 COPY . .
-
 ENV MODEL_ID=$MODEL_ID
-ENV DEVICE=auto
-
-CMD python3 main.py
+#ENV DEVICE=auto
+CMD python main.py
diff --git a/cht-falcon/build.sh b/cht-falcon/build.sh
index 8e8cbad..bd4e4d9 100755
--- a/cht-falcon/build.sh
+++ b/cht-falcon/build.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 set -e
-export VERSION=1.0.0
+export VERSION=1.0.1
 source "$(dirname "${BASH_SOURCE[0]}")/../utils.sh"
 
 build_gpu ghcr.io/premai-io/chat-falcon-7b-instruct-gpu tiiuae/falcon-7b-instruct ${@:1}
diff --git a/cht-falcon/docker/gpu/Dockerfile b/cht-falcon/docker/gpu/Dockerfile
index 61696ca..bc7918b 100644
--- a/cht-falcon/docker/gpu/Dockerfile
+++ b/cht-falcon/docker/gpu/Dockerfile
@@ -1,19 +1,14 @@
 FROM huggingface/transformers-pytorch-gpu:4.28.1
-
-ARG MODEL_ID
-
 WORKDIR /usr/src/app/
 
-COPY requirements.txt ./
-
-RUN pip install --no-cache-dir -r ./requirements.txt --upgrade pip
+RUN cd $(dirname $(which python3)) && ln -s python3 python
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
 
+ARG MODEL_ID
 COPY download.py .
-
-RUN python3 download.py --model $MODEL_ID
+RUN python download.py --model $MODEL_ID
 
 COPY . .
-
 ENV MODEL_ID=$MODEL_ID
-
-CMD python3 main.py
+CMD python main.py
diff --git a/cht-falcon/utils.py b/cht-falcon/utils.py
index 5d726a2..a2494e6 100644
--- a/cht-falcon/utils.py
+++ b/cht-falcon/utils.py
@@ -16,10 +16,7 @@ def __call__(self, input_ids, scores, **kwargs) -> bool:
         generated_text = self.tokenizer.decode(input_ids[0])
         generated_text = generated_text.replace(self.prompt, "")
         # Check if the target sequence appears in the generated text
-        return any(
-            target_sequence in generated_text
-            for target_sequence in self.target_sequences
-        )
+        return any(target_sequence in generated_text for target_sequence in self.target_sequences)
 
     def __len__(self) -> int:
         return len(self.target_sequences)
diff --git a/cht-gorilla/build.sh b/cht-gorilla/build.sh
index 632b99f..190207b 100755
--- a/cht-gorilla/build.sh
+++ b/cht-gorilla/build.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 set -e
-export VERSION=1.0.0
+export VERSION=1.0.1
 source "$(dirname "${BASH_SOURCE[0]}")/../utils.sh"
 
 build_gpu ghcr.io/premai-io/chat-gorilla-falcon-7b-gpu gorilla-llm/gorilla-falcon-7b-hf-v0 ${@:1}
diff --git a/cht-gorilla/docker/gpu/Dockerfile b/cht-gorilla/docker/gpu/Dockerfile
index 61696ca..bc7918b 100644
--- a/cht-gorilla/docker/gpu/Dockerfile
+++ b/cht-gorilla/docker/gpu/Dockerfile
@@ -1,19 +1,14 @@
 FROM huggingface/transformers-pytorch-gpu:4.28.1
-
-ARG MODEL_ID
-
 WORKDIR /usr/src/app/
 
-COPY requirements.txt ./
-
-RUN pip install --no-cache-dir -r ./requirements.txt --upgrade pip
+RUN cd $(dirname $(which python3)) && ln -s python3 python
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
 
+ARG MODEL_ID
 COPY download.py .
-
-RUN python3 download.py --model $MODEL_ID
+RUN python download.py --model $MODEL_ID
 
 COPY . .
-
 ENV MODEL_ID=$MODEL_ID
-
-CMD python3 main.py
+CMD python main.py
diff --git a/cht-gorilla/tests/test_views.py b/cht-gorilla/tests/test_views.py
index 1ec8b17..10c7a6e 100644
--- a/cht-gorilla/tests/test_views.py
+++ b/cht-gorilla/tests/test_views.py
@@ -11,9 +11,7 @@ def test_chat_gorilla() -> None:
             "/v1/chat/completions",
             json={
                 "model": os.getenv("MODEL_ID", "gorilla-llm/gorilla-falcon-7b-hf-v0"),
-                "messages": [
-                    {"role": "user", "content": "Generate an image of  a cat"}
-                ],
+                "messages": [{"role": "user", "content": "Generate an image of  a cat"}],
             },
         )
         assert response.status_code == 200, response.content
diff --git a/cht-llama-cpp/build.sh b/cht-llama-cpp/build.sh
index f844ac7..fb78f4b 100755
--- a/cht-llama-cpp/build.sh
+++ b/cht-llama-cpp/build.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 set -e
-export VERSION=1.0.4
+export VERSION=1.0.5
 source "$(dirname "${BASH_SOURCE[0]}")/../utils.sh"
 
 build_cpu ghcr.io/premai-io/chat-gpt4all-lora-q4-cpu gpt4all-lora-q4 ${@:1}
diff --git a/cht-llama-cpp/docker/cpu/Dockerfile b/cht-llama-cpp/docker/cpu/Dockerfile
index 14a31cb..a604b8e 100644
--- a/cht-llama-cpp/docker/cpu/Dockerfile
+++ b/cht-llama-cpp/docker/cpu/Dockerfile
@@ -1,22 +1,17 @@
 FROM python:3.10-slim-bullseye
-
-ARG MODEL_ID
-
-RUN apt update && apt install -y libopenblas-dev ninja-build build-essential wget
-RUN python -m pip install --upgrade pip pytest cmake scikit-build setuptools
-
 WORKDIR /usr/src/app/
 
-RUN wget https://prem-models.s3.eu-central-1.amazonaws.com/${MODEL_ID}.bin
-RUN mkdir -p ./ml/models/
-RUN mv ${MODEL_ID}.bin ./ml/models/
-
-COPY requirements.txt ./
+RUN apt update -qq && apt install -yqq --no-install-recommends \
+    libopenblas-dev ninja-build build-essential wget \
+    && rm -rf /var/lib/apt/lists/*
+COPY requirements.txt .
+#RUN pip install pytest cmake scikit-build setuptools
+RUN pip install --no-cache-dir -r ./requirements.txt
 
-RUN pip install --no-cache-dir -r ./requirements.txt --upgrade pip
+ARG MODEL_ID
+RUN mkdir -p ./ml/models/
+RUN wget -O ./ml/models/${MODEL_ID}.bin https://prem-models.s3.eu-central-1.amazonaws.com/${MODEL_ID}.bin
 
 COPY . .
-
 ENV MODEL_ID=$MODEL_ID
-
 CMD python main.py
diff --git a/cht-llama-cpp/routes.py b/cht-llama-cpp/routes.py
index 5e47307..ff84caa 100644
--- a/cht-llama-cpp/routes.py
+++ b/cht-llama-cpp/routes.py
@@ -94,9 +94,7 @@ async def generate_chunk_based_response(body):
 async def chat_completions(body: ChatCompletionInput):
     try:
         if body.stream:
-            return StreamingResponse(
-                generate_chunk_based_response(body), media_type="text/event-stream"
-            )
+            return StreamingResponse(generate_chunk_based_response(body), media_type="text/event-stream")
         return model.generate(
             messages=body.messages,
             temperature=body.temperature,
diff --git a/cht-llama-v2/build.sh b/cht-llama-v2/build.sh
index 935850a..8c2d6ba 100755
--- a/cht-llama-v2/build.sh
+++ b/cht-llama-v2/build.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 set -e
-export VERSION=1.0.1
+export VERSION=1.0.2
 source "$(dirname "${BASH_SOURCE[0]}")/../utils.sh"
 
 build_gpu ghcr.io/premai-io/chat-llama-2-7b-gpu       llama-2-7b-hf       ${@:1}
diff --git a/cht-llama-v2/docker/gpu/Dockerfile b/cht-llama-v2/docker/gpu/Dockerfile
index 50b61a8..5e2f95f 100644
--- a/cht-llama-v2/docker/gpu/Dockerfile
+++ b/cht-llama-v2/docker/gpu/Dockerfile
@@ -1,19 +1,17 @@
 FROM huggingface/transformers-pytorch-gpu:4.28.1
-
-ARG MODEL_ID
-
-RUN apt install -y wget unzip
-
 WORKDIR /usr/src/app/
 
+RUN cd $(dirname $(which python3)) && ln -s python3 python
+RUN apt update -qq && apt install -yqq --no-install-recommends \
+    wget unzip \
+    && rm -rf /var/lib/apt/lists/*
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+ARG MODEL_ID
 RUN wget https://prem-models.s3.eu-central-1.amazonaws.com/llama-v2/${MODEL_ID}.zip
 RUN unzip -o ${MODEL_ID}.zip && rm ${MODEL_ID}.zip
 
 COPY . .
-
-RUN pip install --no-cache-dir -r ./requirements.txt --upgrade pip
-
-
 ENV MODEL_ID=$MODEL_ID
-
-CMD python3 main.py
+CMD python main.py
diff --git a/cht-llama-v2/utils.py b/cht-llama-v2/utils.py
index b837703..981352e 100644
--- a/cht-llama-v2/utils.py
+++ b/cht-llama-v2/utils.py
@@ -16,10 +16,7 @@ def __call__(self, input_ids, scores, **kwargs) -> bool:
         generated_text = self.tokenizer.decode(input_ids[0])
         generated_text = generated_text.replace(self.prompt, "")
         # Check if the target sequence appears in the generated text
-        return any(
-            target_sequence in generated_text
-            for target_sequence in self.target_sequences
-        )
+        return any(target_sequence in generated_text for target_sequence in self.target_sequences)
 
     def __len__(self) -> int:
         return len(self.target_sequences)
diff --git a/cht-mpt/build.sh b/cht-mpt/build.sh
index c620ac9..88da6d5 100755
--- a/cht-mpt/build.sh
+++ b/cht-mpt/build.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 set -e
-export VERSION=1.0.0
+export VERSION=1.0.1
 source "$(dirname "${BASH_SOURCE[0]}")/../utils.sh"
 
 build_gpu ghcr.io/premai-io/chat-mpt-7b-gpu     mosaicml/mpt-7b-chat     ${@:1}
diff --git a/cht-mpt/docker/gpu/Dockerfile b/cht-mpt/docker/gpu/Dockerfile
index 61696ca..bc7918b 100644
--- a/cht-mpt/docker/gpu/Dockerfile
+++ b/cht-mpt/docker/gpu/Dockerfile
@@ -1,19 +1,14 @@
 FROM huggingface/transformers-pytorch-gpu:4.28.1
-
-ARG MODEL_ID
-
 WORKDIR /usr/src/app/
 
-COPY requirements.txt ./
-
-RUN pip install --no-cache-dir -r ./requirements.txt --upgrade pip
+RUN cd $(dirname $(which python3)) && ln -s python3 python
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
 
+ARG MODEL_ID
 COPY download.py .
-
-RUN python3 download.py --model $MODEL_ID
+RUN python download.py --model $MODEL_ID
 
 COPY . .
-
 ENV MODEL_ID=$MODEL_ID
-
-CMD python3 main.py
+CMD python main.py
diff --git a/cht-mpt/utils.py b/cht-mpt/utils.py
index 9603b8a..d2253ed 100644
--- a/cht-mpt/utils.py
+++ b/cht-mpt/utils.py
@@ -16,10 +16,7 @@ def __call__(self, input_ids, scores, **kwargs) -> bool:
         generated_text = self.tokenizer.decode(input_ids[0])
         generated_text = generated_text.replace(self.prompt, "")
         # Check if the target sequence appears in the generated text
-        return any(
-            target_sequence in generated_text
-            for target_sequence in self.target_sequences
-        )
+        return any(target_sequence in generated_text for target_sequence in self.target_sequences)
 
     def __len__(self) -> int:
         return len(self.target_sequences)
diff --git a/cht-petals/build.sh b/cht-petals/build.sh
index 56036dd..87c6491 100755
--- a/cht-petals/build.sh
+++ b/cht-petals/build.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 set -e
-export VERSION=1.0.1
+export VERSION=1.0.2
 source "$(dirname "${BASH_SOURCE[0]}")/../utils.sh"
 
 build_cpu ghcr.io/premai-io/chat-stable-beluga-2-cpu petals-team/StableBeluga2           ${@:1}
diff --git a/cht-petals/docker/cpu/Dockerfile b/cht-petals/docker/cpu/Dockerfile
index 0e1c0ee..6cfb347 100644
--- a/cht-petals/docker/cpu/Dockerfile
+++ b/cht-petals/docker/cpu/Dockerfile
@@ -1,18 +1,17 @@
 FROM python:3.10-slim-bullseye
-
-RUN apt update && apt install -y libopenblas-dev ninja-build build-essential wget git
-RUN python -m pip install --upgrade pip pytest cmake scikit-build setuptools
-
 WORKDIR /usr/src/app/
 
-COPY requirements.txt ./
+RUN apt update -qq && apt install -yqq --no-install-recommends \
+    libopenblas-dev ninja-build build-essential wget git \
+    && rm -rf /var/lib/apt/lists/*
+COPY requirements.txt .
+#RUN pip install pytest cmake scikit-build setuptools
+RUN pip install --no-cache-dir -r requirements.txt
 
-RUN pip install --no-cache-dir -r ./requirements.txt --upgrade pip
-
-COPY download.py .
 ARG MODEL_ID
-ENV MODEL_ID=$MODEL_ID
-RUN python3 download.py --model $MODEL_ID
+COPY download.py .
+RUN python download.py --model $MODEL_ID
 
 COPY . .
+ENV MODEL_ID=$MODEL_ID
 CMD python main.py
diff --git a/cht-petals/models.py b/cht-petals/models.py
index 583a7a2..8e83a15 100644
--- a/cht-petals/models.py
+++ b/cht-petals/models.py
@@ -58,17 +58,11 @@ def generate(
     @classmethod
     def get_model(cls):
         if cls.model is None:
-            Tokenizer = (
-                LlamaTokenizer
-                if "llama" in os.getenv("MODEL_ID").lower()
-                else AutoTokenizer
-            )
+            Tokenizer = LlamaTokenizer if "llama" in os.getenv("MODEL_ID").lower() else AutoTokenizer
             cls.tokenizer = Tokenizer.from_pretrained(os.getenv("MODEL_ID"))
 
             kwargs = {}
             if "x86_64" in machine():
                 kwargs["torch_dtype"] = torch.float32
-            cls.model = AutoDistributedModelForCausalLM.from_pretrained(
-                os.getenv("MODEL_ID"), **kwargs
-            )
+            cls.model = AutoDistributedModelForCausalLM.from_pretrained(os.getenv("MODEL_ID"), **kwargs)
         return cls.model
diff --git a/cht-xgen/build.sh b/cht-xgen/build.sh
index e6c7c66..71f583b 100755
--- a/cht-xgen/build.sh
+++ b/cht-xgen/build.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 set -e
-export VERSION=1.0.0
+export VERSION=1.0.1
 source "$(dirname "${BASH_SOURCE[0]}")/../utils.sh"
 
 build_gpu ghcr.io/premai-io/chat-xgen-7b-8k-inst-gpu Salesforce/xgen-7b-8k-inst ${@:1}
diff --git a/cht-xgen/docker/gpu/Dockerfile b/cht-xgen/docker/gpu/Dockerfile
index 61696ca..bc7918b 100644
--- a/cht-xgen/docker/gpu/Dockerfile
+++ b/cht-xgen/docker/gpu/Dockerfile
@@ -1,19 +1,14 @@
 FROM huggingface/transformers-pytorch-gpu:4.28.1
-
-ARG MODEL_ID
-
 WORKDIR /usr/src/app/
 
-COPY requirements.txt ./
-
-RUN pip install --no-cache-dir -r ./requirements.txt --upgrade pip
+RUN cd $(dirname $(which python3)) && ln -s python3 python
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
 
+ARG MODEL_ID
 COPY download.py .
-
-RUN python3 download.py --model $MODEL_ID
+RUN python download.py --model $MODEL_ID
 
 COPY . .
-
 ENV MODEL_ID=$MODEL_ID
-
-CMD python3 main.py
+CMD python main.py
diff --git a/dfs-dalle/build.sh b/dfs-dalle/build.sh
index e41e7a3..e0d82bc 100755
--- a/dfs-dalle/build.sh
+++ b/dfs-dalle/build.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 set -e
-export VERSION=1.0.0
+export VERSION=1.1.0
 source "$(dirname "${BASH_SOURCE[0]}")/../utils.sh"
 
 build_gpu ghcr.io/premai-io/diffuser-dalle-mini-gpu - ${@:1} \
diff --git a/dfs-dalle/docker/gpu/Dockerfile b/dfs-dalle/docker/gpu/Dockerfile
index f39cec5..ebc0c0a 100644
--- a/dfs-dalle/docker/gpu/Dockerfile
+++ b/dfs-dalle/docker/gpu/Dockerfile
@@ -1,37 +1,38 @@
-FROM nvidia/cuda:11.6.2-cudnn8-devel-ubuntu20.04
+FROM nvidia/cuda:11.8.0-cudnn8-devel-ubuntu22.04 as build
+WORKDIR /usr/src/app/
+
+RUN apt update -qq && apt install -yqq --no-install-recommends \
+    git python3 python3-dev python3-pip \
+    && rm -rf /var/lib/apt/lists/*
+RUN cd $(dirname $(which python3)) && ln -s python3 python
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt \
+    jax[cuda]==0.3.25 -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
 
-RUN apt-get update && apt-get install -y \
-  git \
-  python3 \
-  python3-pip \
-  && rm -rf /var/lib/apt/lists/*
+FROM nvidia/cuda:11.8.0-cudnn8-runtime-ubuntu22.04
+WORKDIR /usr/src/app/
 
+RUN apt update -qq && apt install -yqq --no-install-recommends \
+    git python3 \
+    && rm -rf /var/lib/apt/lists/*
+RUN cd $(dirname $(which python3)) && ln -s python3 python
+COPY --from=build --link /usr/local/lib/python3.10 /usr/local/lib/python3.10
+COPY --from=build --link /usr/lib/python3.10 /usr/lib/python3.10
+COPY --from=build --link /usr/lib/python3 /usr/lib/python3
 
 ARG DALLE_MODEL_ID
 ARG DALLE_REVISION_ID
-
 ARG VQGAN_MODEL_ID
 ARG VQGAN_REVISION_ID
-
-WORKDIR /usr/src/app/
-
-COPY requirements.txt /usr/src/app/
-
-RUN pip3 install --no-cache-dir -r requirements.txt
-RUN pip3 install jax[cuda11_cudnn82]==0.3.25 -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
-
 COPY download.py .
-
-RUN python3 download.py --dalle-model $DALLE_MODEL_ID --dalle-revision $DALLE_REVISION_ID \
+RUN python download.py \
+    --dalle-model $DALLE_MODEL_ID --dalle-revision $DALLE_REVISION_ID \
     --vqgan-model $VQGAN_MODEL_ID --vqgan-revision $VQGAN_REVISION_ID
 
 COPY . .
-
 ENV DALLE_MODEL_ID=$DALLE_MODEL_ID
 ENV DALLE_REVISION_ID=$DALLE_REVISION_ID
 ENV VQGAN_MODEL_ID=$VQGAN_MODEL_ID
 ENV VQGAN_REVISION_ID=$VQGAN_REVISION_ID
-
 ENV DEVICE=gpu
-
-CMD python3 main.py
+CMD python main.py
diff --git a/dfs-dalle/download.py b/dfs-dalle/download.py
index 14c23b7..a03f009 100644
--- a/dfs-dalle/download.py
+++ b/dfs-dalle/download.py
@@ -7,13 +7,9 @@
 
 parser = argparse.ArgumentParser()
 parser.add_argument("--dalle-model", help="Dalle Model to download")
-parser.add_argument(
-    "--dalle-revision", help="Dalle Revision to download", default=None, required=False
-)
+parser.add_argument("--dalle-revision", help="Dalle Revision to download", default=None, required=False)
 parser.add_argument("--vqgan-model", help="VQGAN Model to download")
-parser.add_argument(
-    "--vqgan-revision", help="VQGAN Revision to download", default=None, required=False
-)
+parser.add_argument("--vqgan-revision", help="VQGAN Revision to download", default=None, required=False)
 args = parser.parse_args()
 
 jax.local_device_count()
@@ -22,14 +18,10 @@
 
 @retry(stop=stop_after_attempt(3), wait=wait_fixed(5))
 def download_model():
-    _ = DalleBart.from_pretrained(
-        args.dalle_model, revision=None, dtype=jax.numpy.float16, _do_init=False
-    )
+    _ = DalleBart.from_pretrained(args.dalle_model, revision=None, dtype=jax.numpy.float16, _do_init=False)
     _ = DalleBartProcessor.from_pretrained(args.dalle_model, revision=None)
 
-    _ = VQModel.from_pretrained(
-        args.vqgan_model, revision=args.vqgan_revision or None, _do_init=False
-    )
+    _ = VQModel.from_pretrained(args.vqgan_model, revision=args.vqgan_revision or None, _do_init=False)
 
 
 download_model()
diff --git a/dfs-dalle/models.py b/dfs-dalle/models.py
index 6a4df28..664c0a0 100644
--- a/dfs-dalle/models.py
+++ b/dfs-dalle/models.py
@@ -67,9 +67,7 @@ def generate(
             img = Image.fromarray(np.asarray(decoded_images[0] * 255, dtype=np.uint8))
             buffered = io.BytesIO()
             img.save(buffered, format="PNG")
-            data.append(
-                {response_format: base64.b64encode(buffered.getvalue()).decode("utf-8")}
-            )
+            data.append({response_format: base64.b64encode(buffered.getvalue()).decode("utf-8")})
 
         return data
 
@@ -78,9 +76,7 @@ def get_model(cls):
         jax.local_device_count()
 
         @partial(jax.pmap, axis_name="batch", static_broadcasted_argnums=(3, 4, 5, 6))
-        def p_generate(
-            tokenized_prompt, key, params, top_k, top_p, temperature, condition_scale
-        ):
+        def p_generate(tokenized_prompt, key, params, top_k, top_p, temperature, condition_scale):
             return cls.model.generate(
                 **tokenized_prompt,
                 prng_key=key,
@@ -107,9 +103,7 @@ def p_decode(indices, params):
             )
             cls.decoder, vqgan_params = VQModel.from_pretrained(
                 os.getenv("VQGAN_MODEL_ID", "dalle-mini/vqgan_imagenet_f16_16384"),
-                revision=os.getenv(
-                    "VQGAN_REVISION_ID", "e93a26e7707683d349bf5d5c41c5b0ef69b677a9"
-                ),
+                revision=os.getenv("VQGAN_REVISION_ID", "e93a26e7707683d349bf5d5c41c5b0ef69b677a9"),
                 _do_init=False,
             )
             cls.processor = DalleBartProcessor.from_pretrained(
diff --git a/dfs-dalle/requirements.txt b/dfs-dalle/requirements.txt
index 575959e..648ad1a 100644
--- a/dfs-dalle/requirements.txt
+++ b/dfs-dalle/requirements.txt
@@ -1,13 +1,13 @@
-fastapi==0.95.0
-uvicorn==0.21.1
-pytest==7.2.2
-requests==2.28.2
-tqdm==4.65.0
-httpx==0.23.3
+fastapi==0.103.1
+uvicorn==0.23.2
+pytest==7.4.2
+requests==2.31.0
+tqdm==4.66.1
+httpx==0.25.0
 python-dotenv==1.0.0
-tenacity==8.2.2
-dalle-mini==0.1.4
+tenacity==8.2.3
+dalle-mini==0.1.5
 git+https://github.com/patil-suraj/vqgan-jax.git
-jaxlib==0.3.25
+jaxlib
 flax==0.6.3
 transformers==4.25.1
diff --git a/dfs-diffusers/build.sh b/dfs-diffusers/build.sh
index 41a1f05..e086542 100755
--- a/dfs-diffusers/build.sh
+++ b/dfs-diffusers/build.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 set -e
-export VERSION=1.0.3
+export VERSION=1.0.4
 source "$(dirname "${BASH_SOURCE[0]}")/../utils.sh"
 
 build_gpu ghcr.io/premai-io/diffuser-stable-diffusion-2-1-gpu             stabilityai/stable-diffusion-2-1         ${@:1}
diff --git a/dfs-diffusers/docker/gpu/Dockerfile b/dfs-diffusers/docker/gpu/Dockerfile
index bb8a63f..a2ed439 100644
--- a/dfs-diffusers/docker/gpu/Dockerfile
+++ b/dfs-diffusers/docker/gpu/Dockerfile
@@ -1,22 +1,17 @@
 FROM huggingface/transformers-pytorch-gpu:latest
-
-ARG MODEL_ID
-ARG REFINER_ID
-
 WORKDIR /usr/src/app/
 
-COPY requirements.txt /usr/src/app/
-
-RUN pip3 install --no-cache-dir -r requirements.txt
+RUN cd $(dirname $(which python3)) && ln -s python3 python
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
 
+ARG MODEL_ID
+ARG REFINER_ID
 COPY download.py .
-
-RUN python3 download.py --model $MODEL_ID --refiner "$REFINER_ID"
+RUN python download.py --model $MODEL_ID --refiner "$REFINER_ID"
 
 COPY . .
-
 ENV MODEL_ID=$MODEL_ID
 ENV REFINER_ID=$REFINER_ID
 ENV DEVICE=cuda
-
-CMD python3 main.py
+CMD python main.py
diff --git a/dfs-diffusers/download.py b/dfs-diffusers/download.py
index de68de4..fc7f3c7 100644
--- a/dfs-diffusers/download.py
+++ b/dfs-diffusers/download.py
@@ -1,11 +1,7 @@
 import argparse
 
 import torch
-from diffusers import (
-    DiffusionPipeline,
-    StableDiffusionLatentUpscalePipeline,
-    StableDiffusionPipeline,
-)
+from diffusers import DiffusionPipeline, StableDiffusionLatentUpscalePipeline, StableDiffusionPipeline
 from tenacity import retry, stop_after_attempt, wait_fixed
 
 parser = argparse.ArgumentParser()
@@ -19,12 +15,8 @@
 @retry(stop=stop_after_attempt(3), wait=wait_fixed(5))
 def download_model():
     if "latent" not in args.model:
-        _ = StableDiffusionPipeline.from_pretrained(
-            args.model, torch_dtype=torch.float16
-        )
-    _ = StableDiffusionLatentUpscalePipeline.from_pretrained(
-        args.model, torch_dtype=torch.float16
-    )
+        _ = StableDiffusionPipeline.from_pretrained(args.model, torch_dtype=torch.float16)
+    _ = StableDiffusionLatentUpscalePipeline.from_pretrained(args.model, torch_dtype=torch.float16)
     if args.refiner:
         _ = DiffusionPipeline.from_pretrained(args.refiner, torch_dtype=torch.float16)
 
diff --git a/dfs-diffusers/models.py b/dfs-diffusers/models.py
index 95a7ffe..1712303 100644
--- a/dfs-diffusers/models.py
+++ b/dfs-diffusers/models.py
@@ -49,9 +49,7 @@ def generate(
         if image:
             init_image = Image.open(io.BytesIO(image.file.read())).convert("RGB")
             init_image = (
-                init_image.resize(tuple(map(int, (size.split("x")))))
-                if size
-                else init_image
+                init_image.resize(tuple(map(int, (size.split("x"))))) if size else init_image
             )  # breaks e.g 512x512 -> (512, 512)
             model_fn = partial(model_fn, image=init_image)
         images = model_fn(output_type="latent" if cls.refiner_model else "pil").images
@@ -63,9 +61,7 @@ def generate(
             buffered = io.BytesIO()
             img = img.resize(tuple(map(int, (size.split("x"))))) if size else img
             img.save(buffered, format="PNG")
-            data.append(
-                {response_format: base64.b64encode(buffered.getvalue()).decode()}
-            )
+            data.append({response_format: base64.b64encode(buffered.getvalue()).decode()})
         return data
 
     @classmethod
@@ -85,11 +81,7 @@ def upscale(
         # size = "300x300"
         init_image = Image.open(io.BytesIO(image.file.read())).convert("RGB")
         # print("size found:", init_image.)
-        init_image = (
-            init_image.resize(tuple(map(int, (size.split("x")))))
-            if size
-            else init_image
-        )
+        init_image = init_image.resize(tuple(map(int, (size.split("x"))))) if size else init_image
         images = cls.upscaler_model(
             prompt=prompt,
             image=init_image,
@@ -105,9 +97,7 @@ def upscale(
             buffered = io.BytesIO()
             img = img.resize(tuple(map(int, (size.split("x"))))) if size else img
             img.save(buffered, format="PNG")
-            data.append(
-                {response_format: base64.b64encode(buffered.getvalue()).decode()}
-            )
+            data.append({response_format: base64.b64encode(buffered.getvalue()).decode()})
         return data
 
     @classmethod
@@ -116,11 +106,9 @@ def get_model(cls):
             model_id = os.getenv("MODEL_ID", "stabilityai/stable-diffusion-2-1")
             print("set text img model: ", model_id)
             if "latent" in model_id:
-                cls.upscaler_model = (
-                    StableDiffusionLatentUpscalePipeline.from_pretrained(
-                        model_id, torch_dtype=torch.float16
-                    ).to(os.getenv("DEVICE", "cpu"))
-                )
+                cls.upscaler_model = StableDiffusionLatentUpscalePipeline.from_pretrained(
+                    model_id, torch_dtype=torch.float16
+                ).to(os.getenv("DEVICE", "cpu"))
                 cls.upscaler_model.enable_attention_slicing()
                 return cls.upscaler_model
             elif "xl" in model_id:
@@ -143,14 +131,10 @@ def get_model(cls):
             cls.text_img_model = cls.text_img_model.to(os.getenv("DEVICE", "cpu"))
             cls.text_img_model.enable_attention_slicing()
 
-            cls.img_img_model = StableDiffusionImg2ImgPipeline(
-                **cls.text_img_model.components
-            )
+            cls.img_img_model = StableDiffusionImg2ImgPipeline(**cls.text_img_model.components)
             cls.upscaler_model = StableDiffusionUpscalePipeline(
                 **cls.text_img_model.components,
-                low_res_scheduler=DDPMScheduler.from_config(
-                    cls.text_img_model.scheduler.config
-                ),
+                low_res_scheduler=DDPMScheduler.from_config(cls.text_img_model.scheduler.config),
             )
 
         return cls.text_img_model
diff --git a/dfs-diffusers/routes.py b/dfs-diffusers/routes.py
index f2fcb95..2f62baf 100644
--- a/dfs-diffusers/routes.py
+++ b/dfs-diffusers/routes.py
@@ -22,9 +22,7 @@ def as_form(cls: Type[BaseModel]):
             inspect.Parameter(
                 model_field.alias,
                 inspect.Parameter.POSITIONAL_ONLY,
-                default=Form(...)
-                if model_field.required
-                else Form(model_field.default),
+                default=Form(...) if model_field.required else Form(model_field.default),
                 annotation=model_field.outer_type_,
             )
         )
diff --git a/ebd-all-minilm/build.sh b/ebd-all-minilm/build.sh
index b0af3de..c112287 100755
--- a/ebd-all-minilm/build.sh
+++ b/ebd-all-minilm/build.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 set -e
-export VERSION=1.0.3
+export VERSION=1.0.4
 source "$(dirname "${BASH_SOURCE[0]}")/../utils.sh"
 
 build_cpu ghcr.io/premai-io/embeddings-all-minilm-l6-v2-cpu all-MiniLM-L6-v2 ${@:1}
diff --git a/ebd-all-minilm/docker/cpu/Dockerfile b/ebd-all-minilm/docker/cpu/Dockerfile
index 0f67a9d..d1129c6 100644
--- a/ebd-all-minilm/docker/cpu/Dockerfile
+++ b/ebd-all-minilm/docker/cpu/Dockerfile
@@ -1,30 +1,17 @@
 FROM python:3.10-slim-bullseye
-
-ARG MODEL_ID
-
 WORKDIR /usr/src/app/
 
-ENV PYTHONDONTWRITEBYTECODE 1
-ENV PYTHONUNBUFFERED 1
-
-RUN apt-get update && \
-    apt-get install -y --no-install-recommends \
-    build-essential \
-    git \
-    wget \
-    libatlas-base-dev
-
-COPY requirements.txt /usr/src/app/
-
-RUN pip3 install --no-cache-dir -r requirements.txt
+RUN apt update -qq && apt install -yqq --no-install-recommends \
+    build-essential git wget libatlas-base-dev \
+    && rm -rf /var/lib/apt/lists/*
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
 
+ARG MODEL_ID
 COPY download.py .
-
-RUN python3 download.py --model $MODEL_ID
+RUN python download.py --model $MODEL_ID
 
 COPY . .
-
 ENV MODEL_ID=$MODEL_ID
 ENV DEVICE=cpu
-
-CMD python3 main.py
+CMD python main.py
diff --git a/ebd-all-minilm/docker/gpu/Dockerfile b/ebd-all-minilm/docker/gpu/Dockerfile
index 9f86b69..2d89f63 100644
--- a/ebd-all-minilm/docker/gpu/Dockerfile
+++ b/ebd-all-minilm/docker/gpu/Dockerfile
@@ -1,20 +1,15 @@
 FROM huggingface/transformers-pytorch-gpu:latest
-
-ARG MODEL_ID
-
 WORKDIR /usr/src/app/
 
-COPY requirements.txt ./
-
-RUN pip install --no-cache-dir -r ./requirements.txt --upgrade pip
+RUN cd $(dirname $(which python3)) && ln -s python3 python
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
 
+ARG MODEL_ID
 COPY download.py .
-
-RUN python3 download.py --model $MODEL_ID
+RUN python download.py --model $MODEL_ID
 
 COPY . .
-
 ENV MODEL_ID=$MODEL_ID
 ENV DEVICE=cuda
-
-CMD python3 main.py
+CMD python main.py
diff --git a/ebd-all-minilm/routes.py b/ebd-all-minilm/routes.py
index 7c0c69a..99983fd 100644
--- a/ebd-all-minilm/routes.py
+++ b/ebd-all-minilm/routes.py
@@ -56,9 +56,7 @@ async def embeddings(body: EmbeddingsInput):
     )
 
 
-@router.post(
-    "/engines/text-embedding-ada-002/embeddings", response_model=EmbeddingsResponse
-)
+@router.post("/engines/text-embedding-ada-002/embeddings", response_model=EmbeddingsResponse)
 async def embeddings_openai(body: EmbeddingsInput):
     if len(body.input) > 0 and isinstance(body.input[0], list):
         encoding = tiktoken.model.encoding_for_model("text-embedding-ada-002")
diff --git a/pyproject.toml b/pyproject.toml
index b62b8cf..bd585ea 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,8 +1,12 @@
 [tool.flake8]
-max_line_length = 99
+max_line_length = 120
 
 [tool.isort]
 profile = "black"
+line_length = 120
+
+[tool.black]
+line_length = 120
 
 [tool.mypy]
 check_untyped_defs = true
diff --git a/t2a-bark/build.sh b/t2a-bark/build.sh
index 3ff7c52..1fa2e6c 100755
--- a/t2a-bark/build.sh
+++ b/t2a-bark/build.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 set -e
-export VERSION=1.0.0
+export VERSION=1.0.1
 source "$(dirname "${BASH_SOURCE[0]}")/../utils.sh"
 
 build_cpu ghcr.io/premai-io/text-to-audio-bark-cpu bark/t2a-bark ${@:1}
diff --git a/t2a-bark/docker/cpu/Dockerfile b/t2a-bark/docker/cpu/Dockerfile
index 2bb4d30..c853a20 100644
--- a/t2a-bark/docker/cpu/Dockerfile
+++ b/t2a-bark/docker/cpu/Dockerfile
@@ -1,32 +1,19 @@
 FROM python:3.10-slim-bullseye
-
-ARG MODEL_ID
-
 WORKDIR /usr/src/app/
 
-ENV PYTHONDONTWRITEBYTECODE 1
-ENV PYTHONUNBUFFERED 1
-
-RUN apt-get update && \
-    apt-get install -y --no-install-recommends \
-    build-essential \
-    git \
-    wget \
-    libatlas-base-dev
-
-COPY requirements.txt /usr/src/app/
-
-RUN pip3 install --no-cache-dir -r requirements.txt
+RUN apt update -qq && apt install -yqq --no-install-recommends \
+    build-essential git wget libatlas-base-dev \
+    && rm -rf /var/lib/apt/lists/*
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
 
+ARG MODEL_ID
 COPY download.py .
-
-RUN python3 download.py
+RUN python download.py
 
 COPY . .
-
 ENV MODEL_ID=$MODEL_ID
 ENV DEVICE=cpu
 ENV SUNO_OFFLOAD_CPU=True
 ENV SUNO_USE_SMALL_MODELS=True
-
-CMD python3 main.py
+CMD python main.py
diff --git a/t2a-bark/docker/gpu/Dockerfile b/t2a-bark/docker/gpu/Dockerfile
index f563189..6a40e30 100644
--- a/t2a-bark/docker/gpu/Dockerfile
+++ b/t2a-bark/docker/gpu/Dockerfile
@@ -1,20 +1,15 @@
 FROM huggingface/transformers-pytorch-gpu:latest
-
-ARG MODEL_ID
-
 WORKDIR /usr/src/app/
 
-COPY requirements.txt /usr/src/app/
-
-RUN pip3 install --no-cache-dir -r requirements.txt
+RUN cd $(dirname $(which python3)) && ln -s python3 python
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
 
+ARG MODEL_ID
 COPY download.py .
-
-RUN python3 download.py
+RUN python download.py
 
 COPY . .
-
 ENV MODEL_ID=$MODEL_ID
 ENV DEVICE=cuda
-
-CMD python3 main.py
+CMD python main.py