Finetuning code refactor (opea-project#1081)

Signed-off-by: Ye, Xinyu <[email protected]>

.github/workflows/docker/compose/finetuning-compose.yaml

@@ -5,9 +5,9 @@
 services:
   finetuning:
     build:
-      dockerfile: comps/finetuning/Dockerfile
+      dockerfile: comps/finetuning/src/Dockerfile
     image: ${REGISTRY:-opea}/finetuning:${TAG:-latest}
   finetuning-gaudi:
     build:
-      dockerfile: comps/finetuning/Dockerfile.intel_hpu
+      dockerfile: comps/finetuning/src/Dockerfile.intel_hpu
     image: ${REGISTRY:-opea}/finetuning-gaudi:${TAG:-latest}

comps/finetuning/Dockerfile → comps/finetuning/src/Dockerfile

@@ -24,17 +24,17 @@ RUN python -m pip install --no-cache-dir --upgrade pip && \
     python -m pip install --no-cache-dir torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu && \
     python -m pip install --no-cache-dir intel-extension-for-pytorch && \
     python -m pip install --no-cache-dir oneccl_bind_pt --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/cpu/cn/ && \
-    python -m pip install --no-cache-dir -r /home/user/comps/finetuning/requirements.txt
+    python -m pip install --no-cache-dir -r /home/user/comps/finetuning/src/requirements.txt
 
 ENV PYTHONPATH=$PYTHONPATH:/home/user
 
-WORKDIR /home/user/comps/finetuning/
+WORKDIR /home/user/comps/finetuning/src
 
 RUN echo PKGPATH=$(python3 -c "import pkg_resources; print(pkg_resources.get_distribution('oneccl-bind-pt').location)") >> run.sh && \
     echo 'export LD_LIBRARY_PATH=$PKGPATH/oneccl_bindings_for_pytorch/opt/mpi/lib/:$LD_LIBRARY_PATH' >> run.sh && \
     echo 'source $PKGPATH/oneccl_bindings_for_pytorch/env/setvars.sh' >> run.sh && \
     echo ray start --head --dashboard-host=0.0.0.0 >> run.sh && \
     echo export RAY_ADDRESS=http://localhost:8265 >> run.sh && \
-    echo python finetuning_service.py >> run.sh
+    echo python opea_finetuning_microservice.py >> run.sh
 
 CMD bash run.sh

comps/finetuning/Dockerfile.intel_hpu → comps/finetuning/src/Dockerfile.intel_hpu

@@ -19,11 +19,11 @@ USER user
 ENV PATH=$PATH:/home/user/.local/bin
 
 RUN python -m pip install --no-cache-dir --upgrade pip && \
-    python -m pip install --no-cache-dir -r /home/user/comps/finetuning/requirements.txt && \
+    python -m pip install --no-cache-dir -r /home/user/comps/finetuning/src/requirements.txt && \
     python -m pip install --no-cache-dir optimum-habana
 
 ENV PYTHONPATH=$PYTHONPATH:/home/user
 
-WORKDIR /home/user/comps/finetuning/
+WORKDIR /home/user/comps/finetuning/src
 
 ENTRYPOINT ["/bin/bash", "launch.sh"]

comps/finetuning/README.md → comps/finetuning/src/README.md

@@ -53,7 +53,7 @@ Build docker image with below command:
 ```bash
 export HF_TOKEN=${your_huggingface_token}
 cd ../../
-docker build -t opea/finetuning:latest --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy --build-arg HF_TOKEN=$HF_TOKEN -f comps/finetuning/Dockerfile .
+docker build -t opea/finetuning:latest --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy --build-arg HF_TOKEN=$HF_TOKEN -f comps/finetuning/src/Dockerfile .
 ```
 
 #### 2.1.2 Run Docker with CLI
@@ -72,7 +72,7 @@ Build docker image with below command:
 
 ```bash
 cd ../../
-docker build -t opea/finetuning-gaudi:latest --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy -f comps/finetuning/Dockerfile.intel_hpu .
+docker build -t opea/finetuning-gaudi:latest --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy -f comps/finetuning/src/Dockerfile.intel_hpu .
 ```
 
 #### 2.2.2 Run Docker with CLI
@@ -244,8 +244,8 @@ curl http://${your_ip}:8015/v1/finetune/list_checkpoints -X POST -H "Content-Typ
 
 ### 3.4 Leverage fine-tuned model
 
-After fine-tuning job is done, fine-tuned model can be chosen from listed checkpoints, then the fine-tuned model can be used in other microservices. For example, fine-tuned reranking model can be used in [rerankings](../rerankings/src/README.md) microservice by assign its path to the environment variable `RERANK_MODEL_ID`, fine-tuned embedding model can be used in [embeddings](../embeddings/src/README.md) microservice by assign its path to the environment variable `model`, LLMs after instruction tuning can be used in [llms](../llms/src/text-generation/README.md) microservice by assign its path to the environment variable `your_hf_llm_model`.
+After fine-tuning job is done, fine-tuned model can be chosen from listed checkpoints, then the fine-tuned model can be used in other microservices. For example, fine-tuned reranking model can be used in [reranks](../../rerankings/src/README.md) microservice by assign its path to the environment variable `RERANK_MODEL_ID`, fine-tuned embedding model can be used in [embeddings](../../embeddings/src/README.md) microservice by assign its path to the environment variable `model`, LLMs after instruction tuning can be used in [llms](../../llms/text-generation/README.md) microservice by assign its path to the environment variable `your_hf_llm_model`.
 
 ## 🚀4. Descriptions for Finetuning parameters
 
-We utilize [OpenAI finetuning parameters](https://platform.openai.com/docs/api-reference/fine-tuning) and extend it with more customizable parameters, see the definitions at [finetune_config](https://github.com/opea-project/GenAIComps/blob/main/comps/finetuning/finetune_config.py).
+We utilize [OpenAI finetuning parameters](https://platform.openai.com/docs/api-reference/fine-tuning) and extend it with more customizable parameters, see the definitions at [finetune_config](https://github.com/opea-project/GenAIComps/blob/main/comps/finetuning/src/integrations/finetune_config.py).

comps/finetuning/src/integrations/__init__.py

@@ -0,0 +1,2 @@
+# Copyright (C) 2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0

comps/finetuning/finetune_runner.py → comps/finetuning/src/integrations/finetune_runner.py

@@ -6,7 +6,7 @@
 from pydantic_yaml import parse_yaml_raw_as
 from transformers import TrainerCallback, TrainerControl, TrainerState, TrainingArguments
 
-from comps.finetuning.finetune_config import FinetuneConfig
+from comps.finetuning.src.integrations.finetune_config import FinetuneConfig
 
 
 class FineTuneCallback(TrainerCallback):
@@ -29,7 +29,7 @@ def main():
     callback = FineTuneCallback()
     finetune_config["Training"]["callbacks"] = [callback]
 
-    from comps.finetuning.llm_on_ray.finetune.finetune import main as llm_on_ray_finetune_main
+    from comps.finetuning.src.integrations.llm_on_ray.finetune.finetune import main as llm_on_ray_finetune_main
 
     llm_on_ray_finetune_main(finetune_config)
 

comps/finetuning/llm_on_ray/finetune/finetune.py → comps/finetuning/src/integrations/llm_on_ray/finetune/finetune.py

@@ -23,9 +23,9 @@
 from transformers import Trainer, TrainingArguments
 
 from comps import CustomLogger
-from comps.finetuning.finetune_config import FinetuneConfig
-from comps.finetuning.llm_on_ray import common
-from comps.finetuning.llm_on_ray.finetune.data_process import (
+from comps.finetuning.src.integrations.finetune_config import FinetuneConfig
+from comps.finetuning.src.integrations.llm_on_ray import common
+from comps.finetuning.src.integrations.llm_on_ray.finetune.data_process import (
     DPOCollator,
     DPODataProcessor,
     EmbedCollator,
@@ -35,7 +35,7 @@
     TrainDatasetForCE,
     TrainDatasetForEmbedding,
 )
-from comps.finetuning.llm_on_ray.finetune.modeling import BiEncoderModel, CrossEncoder
+from comps.finetuning.src.integrations.llm_on_ray.finetune.modeling import BiEncoderModel, CrossEncoder
 
 logger = CustomLogger("llm_on_ray/finetune")
 
@@ -394,7 +394,7 @@ def get_trainer(config: Dict, model, ref_model, tokenizer, tokenized_dataset, da
         if task == "dpo":
             lora_config = config["General"].get("lora_config", None)
             peft_config = LoraConfig(**lora_config)
-            from comps.finetuning.llm_on_ray.finetune.dpo_trainer import DPOTrainer
+            from comps.finetuning.src.integrations.llm_on_ray.finetune.dpo_trainer import DPOTrainer
 
             trainer = DPOTrainer(
                 model,
@@ -431,7 +431,7 @@ def get_trainer(config: Dict, model, ref_model, tokenizer, tokenized_dataset, da
         if task == "dpo":
             lora_config = config["General"].get("lora_config", None)
             peft_config = LoraConfig(**lora_config)
-            from comps.finetuning.llm_on_ray.finetune.dpo_trainer import GaudiDPOTrainer
+            from comps.finetuning.src.integrations.llm_on_ray.finetune.dpo_trainer import GaudiDPOTrainer
 
             trainer = GaudiDPOTrainer(
                 model,