tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

don't compile deepspeed or bitsandbytes from source (#837)

Browse Source
This commit is contained in:
Wing Lian
2023-11-08 19:49:55 -05:00
committed by GitHub
parent 641e6f7e51
commit f544ab2bed
25 changed files with 29 additions and 65 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
docker/Dockerfile
View File

@@ -21,9 +21,9 @@ WORKDIR /workspace/axolotl
# If AXOLOTL_EXTRAS is set, append it in brackets # If AXOLOTL_EXTRAS is set, append it in brackets
RUN sed -i "s/torch==.*/torch==$PYTORCH_VERSION/" requirements.txt RUN sed -i "s/torch==.*/torch==$PYTORCH_VERSION/" requirements.txt
RUN if [ "$AXOLOTL_EXTRAS" != "" ] ; then \ RUN if [ "$AXOLOTL_EXTRAS" != "" ] ; then \
pip install -e .[flash-attn,$AXOLOTL_EXTRAS]; \ pip install -e .[deepspeed,flash-attn,$AXOLOTL_EXTRAS]; \
else \ else \
pip install -e .[flash-attn]; \ pip install -e .[deepspeed,flash-attn]; \
fi fi
# fix so that git fetch/pull from remote works # fix so that git fetch/pull from remote works

44
docker/Dockerfile-base
View File

@@ -10,8 +10,10 @@ ENV PATH="/root/miniconda3/bin:${PATH}"
ARG PYTHON_VERSION="3.9" ARG PYTHON_VERSION="3.9"
ARG PYTORCH_VERSION="2.0.1" ARG PYTORCH_VERSION="2.0.1"
ARG CUDA="118" ARG CUDA="118"
ARG TORCH_CUDA_ARCH_LIST="7.0 7.5 8.0 8.6 9.0+PTX"
ENV PYTHON_VERSION=$PYTHON_VERSION ENV PYTHON_VERSION=$PYTHON_VERSION
ENV TORCH_CUDA_ARCH_LIST=$TORCH_CUDA_ARCH_LIST
RUN apt-get update \ RUN apt-get update \
&& apt-get install -y wget git build-essential ninja-build git-lfs libaio-dev && rm -rf /var/lib/apt/lists/* \ && apt-get install -y wget git build-essential ninja-build git-lfs libaio-dev && rm -rf /var/lib/apt/lists/* \
@@ -29,45 +31,7 @@ WORKDIR /workspace
RUN python3 -m pip install --upgrade pip && pip3 install packaging && \ RUN python3 -m pip install --upgrade pip && pip3 install packaging && \
python3 -m pip install --no-cache-dir -U torch==${PYTORCH_VERSION}+cu${CUDA} deepspeed-kernels --extra-index-url https://download.pytorch.org/whl/cu$CUDA python3 -m pip install --no-cache-dir -U torch==${PYTORCH_VERSION}+cu${CUDA} deepspeed-kernels --extra-index-url https://download.pytorch.org/whl/cu$CUDA
FROM base-builder AS deepspeed-builder RUN git lfs install --skip-repo && \
pip3 install awscli && \
ARG TORCH_CUDA_ARCH_LIST="7.0 7.5 8.0 8.6 9.0+PTX"
WORKDIR /workspace
RUN git clone https://github.com/microsoft/DeepSpeed.git && \
cd DeepSpeed && \
MAX_CONCURRENCY=8 DS_BUILD_SPARSE_ATTN=0 DS_BUILD_OPS=1 DS_BUILD_EVOFORMER_ATTN=0 python3 setup.py bdist_wheel
FROM base-builder AS bnb-builder
WORKDIR /workspace
ARG CUDA="118"
ENV CUDA=$CUDA
ARG MAX_JOBS="-1"
ENV MAX_JOBS=$MAX_JOBS
RUN git clone https://github.com/TimDettmers/bitsandbytes.git && \
cd bitsandbytes && \
CUDA_VERSION=$CUDA make cuda11x && \
python setup.py bdist_wheel
FROM base-builder
ARG TORCH_CUDA_ARCH_LIST="7.0 7.5 8.0 8.6 9.0+PTX"
ENV TORCH_CUDA_ARCH_LIST=$TORCH_CUDA_ARCH_LIST
RUN mkdir -p /workspace/builds
COPY --from=bnb-builder /workspace/bitsandbytes /workspace/builds/bitsandbytes
RUN mkdir -p /workspace/wheels/bitsandbytes
COPY --from=deepspeed-builder /workspace/DeepSpeed/dist/deepspeed-*.whl wheels
COPY --from=bnb-builder /workspace/bitsandbytes/dist/bitsandbytes-*.whl wheels
COPY --from=bnb-builder /workspace/bitsandbytes/bitsandbytes/libbitsandbytes*.so wheels/bitsandbytes
RUN pip3 install wheels/deepspeed-*.whl
RUN cd /workspace/builds/bitsandbytes && python3 setup.py install
RUN git lfs install --skip-repo
RUN pip3 install awscli && \
# The base image ships with `pydantic==1.8.2` which is not working # The base image ships with `pydantic==1.8.2` which is not working
pip3 install -U --no-cache-dir pydantic==1.10.10 pip3 install -U --no-cache-dir pydantic==1.10.10

2
examples/cerebras/btlm-ft.yml
View File

@@ -14,7 +14,7 @@ datasets:
- path: mhenrichsen/alpaca_2k_test - path: mhenrichsen/alpaca_2k_test
type: alpaca type: alpaca
dataset_prepared_path: last_prepared_run dataset_prepared_path: last_prepared_run
val_set_size: 0.01 val_set_size: 0.05
adapter: adapter:
lora_model_dir: lora_model_dir:

2
examples/cerebras/qlora.yml
View File

@@ -7,7 +7,7 @@ datasets:
- path: teknium/GPT4-LLM-Cleaned - path: teknium/GPT4-LLM-Cleaned
type: alpaca type: alpaca
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
adapter: qlora adapter: qlora
lora_model_dir: lora_model_dir:
sequence_len: 2048 sequence_len: 2048

2
examples/code-llama/13b/lora.yml
View File

@@ -11,7 +11,7 @@ datasets:
- path: mhenrichsen/alpaca_2k_test - path: mhenrichsen/alpaca_2k_test
type: alpaca type: alpaca
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
output_dir: ./lora-out output_dir: ./lora-out
sequence_len: 4096 sequence_len: 4096

2
examples/code-llama/13b/qlora.yml
View File

@@ -11,7 +11,7 @@ datasets:
- path: mhenrichsen/alpaca_2k_test - path: mhenrichsen/alpaca_2k_test
type: alpaca type: alpaca
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
output_dir: ./qlora-out output_dir: ./qlora-out
adapter: qlora adapter: qlora

2
examples/code-llama/34b/lora.yml
View File

@@ -11,7 +11,7 @@ datasets:
- path: mhenrichsen/alpaca_2k_test - path: mhenrichsen/alpaca_2k_test
type: alpaca type: alpaca
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
output_dir: ./lora-out output_dir: ./lora-out
sequence_len: 4096 sequence_len: 4096

2
examples/code-llama/34b/qlora.yml
View File

@@ -11,7 +11,7 @@ datasets:
- path: mhenrichsen/alpaca_2k_test - path: mhenrichsen/alpaca_2k_test
type: alpaca type: alpaca
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
output_dir: ./qlora-out output_dir: ./qlora-out
adapter: qlora adapter: qlora

2
examples/code-llama/7b/lora.yml
View File

@@ -11,7 +11,7 @@ datasets:
- path: mhenrichsen/alpaca_2k_test - path: mhenrichsen/alpaca_2k_test
type: alpaca type: alpaca
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
output_dir: ./lora-out output_dir: ./lora-out
sequence_len: 4096 sequence_len: 4096

2
examples/code-llama/7b/qlora.yml
View File

@@ -11,7 +11,7 @@ datasets:
- path: mhenrichsen/alpaca_2k_test - path: mhenrichsen/alpaca_2k_test
type: alpaca type: alpaca
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
output_dir: ./qlora-out output_dir: ./qlora-out
adapter: qlora adapter: qlora

2
examples/falcon/config-7b-lora.yml
View File

@@ -12,7 +12,7 @@ datasets:
- path: teknium/GPT4-LLM-Cleaned - path: teknium/GPT4-LLM-Cleaned
type: alpaca:chat type: alpaca:chat
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
adapter: lora adapter: lora
lora_model_dir: lora_model_dir:
sequence_len: 2048 sequence_len: 2048

2
examples/falcon/config-7b-qlora.yml
View File

@@ -18,7 +18,7 @@ datasets:
- Chain-of-Thought/formatted_cot_data/gsm8k_train.json - Chain-of-Thought/formatted_cot_data/gsm8k_train.json
type: "alpaca:chat" type: "alpaca:chat"
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
# enable QLoRA # enable QLoRA
adapter: qlora adapter: qlora
lora_model_dir: lora_model_dir:

2
examples/falcon/config-7b.yml
View File

@@ -12,7 +12,7 @@ datasets:
- path: teknium/GPT4-LLM-Cleaned - path: teknium/GPT4-LLM-Cleaned
type: alpaca:chat type: alpaca:chat
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
adapter: adapter:
lora_model_dir: lora_model_dir:
sequence_len: 2048 sequence_len: 2048

2
examples/gptj/qlora.yml
View File

@@ -7,7 +7,7 @@ datasets:
- path: teknium/GPT4-LLM-Cleaned - path: teknium/GPT4-LLM-Cleaned
type: alpaca type: alpaca
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
adapter: qlora adapter: qlora
lora_model_dir: lora_model_dir:
sequence_len: 2048 sequence_len: 2048

2
examples/llama-2/fft_optimized.yml
View File

@@ -11,7 +11,7 @@ datasets:
- path: mhenrichsen/alpaca_2k_test - path: mhenrichsen/alpaca_2k_test
type: alpaca type: alpaca
dataset_prepared_path: last_run_prepared dataset_prepared_path: last_run_prepared
val_set_size: 0.01 val_set_size: 0.05
output_dir: ./out output_dir: ./out
sequence_len: 4096 sequence_len: 4096

2
examples/llama-2/gptq-lora.yml
View File

@@ -15,7 +15,7 @@ datasets:
- path: mhenrichsen/alpaca_2k_test - path: mhenrichsen/alpaca_2k_test
type: alpaca type: alpaca
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
adapter: lora adapter: lora
lora_model_dir: lora_model_dir:
sequence_len: 4096 sequence_len: 4096

2
examples/llama-2/lora.yml
View File

@@ -11,7 +11,7 @@ datasets:
- path: mhenrichsen/alpaca_2k_test - path: mhenrichsen/alpaca_2k_test
type: alpaca type: alpaca
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
output_dir: ./lora-out output_dir: ./lora-out
sequence_len: 4096 sequence_len: 4096

2
examples/llama-2/qlora.yml
View File

@@ -11,7 +11,7 @@ datasets:
- path: mhenrichsen/alpaca_2k_test - path: mhenrichsen/alpaca_2k_test
type: alpaca type: alpaca
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
output_dir: ./qlora-out output_dir: ./qlora-out
adapter: qlora adapter: qlora

2
examples/llama-2/relora.yml
View File

@@ -11,7 +11,7 @@ datasets:
- path: teknium/GPT4-LLM-Cleaned - path: teknium/GPT4-LLM-Cleaned
type: alpaca type: alpaca
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
output_dir: ./relora-out output_dir: ./relora-out
adapter: qlora adapter: qlora

2
examples/llama-2/tiny-llama.yml
View File

@@ -12,7 +12,7 @@ datasets:
- path: mhenrichsen/alpaca_2k_test - path: mhenrichsen/alpaca_2k_test
type: alpaca type: alpaca
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
output_dir: ./lora-out output_dir: ./lora-out
sequence_len: 4096 sequence_len: 4096

2
examples/mistral/config.yml
View File

@@ -11,7 +11,7 @@ datasets:
- path: mhenrichsen/alpaca_2k_test - path: mhenrichsen/alpaca_2k_test
type: alpaca type: alpaca
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
output_dir: ./out output_dir: ./out
sequence_len: 8192 sequence_len: 8192

2
examples/mistral/qlora.yml
View File

@@ -11,7 +11,7 @@ datasets:
- path: mhenrichsen/alpaca_2k_test - path: mhenrichsen/alpaca_2k_test
type: alpaca type: alpaca
dataset_prepared_path: last_run_prepared dataset_prepared_path: last_run_prepared
val_set_size: 0.01 val_set_size: 0.05
output_dir: ./qlora-out output_dir: ./qlora-out
adapter: qlora adapter: qlora

2
examples/openllama-3b/qlora.yml
View File

@@ -9,7 +9,7 @@ datasets:
- path: teknium/GPT4-LLM-Cleaned - path: teknium/GPT4-LLM-Cleaned
type: alpaca type: alpaca
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
adapter: qlora adapter: qlora
lora_model_dir: lora_model_dir:
sequence_len: 1024 sequence_len: 1024

2
examples/xgen-7b/xgen-7b-8k-qlora.yml
View File

@@ -16,7 +16,7 @@ datasets:
- openassistant_best_replies_train.jsonl - openassistant_best_replies_train.jsonl
type: "completion" type: "completion"
dataset_prepared_path: dataset_prepared_path:
val_set_size: 0.01 val_set_size: 0.05
# enable QLoRA # enable QLoRA
adapter: qlora adapter: qlora
lora_model_dir: lora_model_dir:

2
requirements.txt
View File

@@ -3,7 +3,7 @@
torch==2.0.1 torch==2.0.1
auto-gptq==0.4.2 auto-gptq==0.4.2
packaging packaging
peft @ git+https://github.com/huggingface/peft.git peft==0.6.0
transformers @ git+https://github.com/huggingface/transformers.git@acc394c4f5e1283c19783581790b3dc3105a3697 transformers @ git+https://github.com/huggingface/transformers.git@acc394c4f5e1283c19783581790b3dc3105a3697
bitsandbytes>=0.41.1 bitsandbytes>=0.41.1
accelerate @ git+https://github.com/huggingface/accelerate@80da9cfb09bb3cc9f1b385cb55d6b90d025a5fd9 accelerate @ git+https://github.com/huggingface/accelerate@80da9cfb09bb3cc9f1b385cb55d6b90d025a5fd9