New huge version

.dvc/.gitignore

@@ -0,0 +1,3 @@
+/config.local
+/tmp
+/cache

.dvc/config

@@ -0,0 +1,11 @@
+[cache]
+    type = hardlink
+    protected = true
+[core]
+    remote = storage
+['remote "nas-remote"']
+    url = //diskstation_105/dvc_storage
+['remote "storage"']
+    url = s3://dvc-storage
+    endpointurl = https://minio.komolov.synology.me
+    region = us-east-1

.dvcignore

@@ -0,0 +1,3 @@
+# Add patterns of files dvc should ignore, which could improve
+# the performance. Learn more at
+# https://dvc.org/doc/user-guide/dvcignore

.gitignore

@@ -182,3 +182,32 @@ yandex8b-chat-vika
 vikhr8b-chat-vika
 judge-model
 models/
+data/
+
+flash-attention
+
+.serena/
+
+.claude/
+
+4b-chat-vika/
+
+best_config.yaml
+
+CLAUDE.md
+
+.github/instructions/
+
+design.md
+
+tasks.md
+/data
+
+node_modules/npm/
+
+node_modules/
+
+.deepeval/
+
+gptchain/
+db/

.pre-commit-config.yaml

@@ -1,37 +1,72 @@
 repos:
   - repo: https://github.com/pre-commit/pre-commit-hooks
-    rev: v4.5.0
+    rev: v4.6.0
     hooks:
       - id: check-ast
+        description: Check Python AST syntax
       - id: check-case-conflict
+        description: Check for case conflicts in filenames
       - id: check-toml
+        description: Validate TOML files
       - id: check-yaml
+        description: Validate YAML files
+        args: ['--unsafe']  # Allow custom YAML tags used by Hydra
       - id: check-json
+        description: Validate JSON files
       - id: detect-private-key
+        description: Detect private keys in code
       - id: end-of-file-fixer
-        exclude: '.*\.sh$'
+        description: Fix missing newlines at end of files
+        exclude: '.*\.(sh|bat)$'
       - id: trailing-whitespace
+        description: Remove trailing whitespace
+        exclude: '.*\.(md|rst)$'  # Preserve markdown line breaks
       - id: check-added-large-files
+        description: Check for large files being committed
+        args: ['--maxkb=10240']  # Allow up to 10MB files
+      - id: check-merge-conflict
+        description: Check for merge conflict markers
+      - id: debug-statements
+        description: Check for debug statements (pdb, etc.)
+        language: python
+      - id: mixed-line-ending
+        description: Check for mixed line endings
 
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.1.5
+    rev: v0.8.4
     hooks:
       - id: ruff
-        args: [--fix, --exit-non-zero-on-fix, --show-fixes, --select, I]
+        name: ruff-linter
+        description: Run Ruff linter with auto-fixes
+        args: [--fix, --show-fixes, --config=pyproject.toml]
       - id: ruff-format
+        name: ruff-formatter
+        description: Run Ruff formatter
+        args: [--config=pyproject.toml]
 
   - repo: https://github.com/psf/black
-    rev: 23.1.0
+    rev: 24.10.0
     hooks:
       - id: black
-        args: [--fast]
-  - repo: https://github.com/PyCQA/flake8
-    rev: 5.0.0
+        name: black-formatter
+        description: Format Python code with Black
+        language_version: python3
+
+
+  - repo: https://github.com/pycqa/bandit
+    rev: 1.7.10
+    hooks:
+      - id: bandit
+        name: bandit-security-checker
+        description: Run Bandit security linter
+        args: [-c, pyproject.toml]
+        exclude: ^(tests/|legacy)
+
+  - repo: https://github.com/codespell-project/codespell
+    rev: v2.3.0
     hooks:
-      - id: flake8
-        additional_dependencies: [pep8-naming, flake8-bugbear ]
-#  - repo: https://github.com/pre-commit/mirrors-mypy
-#    rev: 'v1.15.0'
-#    hooks:
-#      - id: mypy
-#        args: [ --strict ]
+      - id: codespell
+        name: codespell-spell-checker
+        description: Check for common misspellings
+        args: [--write-changes]
+        exclude: ^(\.git/|docs/_build/|models/|data/|README.md)

.readthedocs.yaml

@@ -2,7 +2,7 @@ version: 2
 
 sphinx:
   configuration: docs/conf.py
-  
+
 python:
    install:
      - requirements: requirements.txt
@@ -14,5 +14,3 @@ build:
   jobs:
     pre_install:
       - pip install --upgrade pip
-
-

Dockerfile

@@ -1,4 +1,4 @@
-FROM nvidia/cuda:12.1.0-cudnn8-runtime-ubuntu22.04
+FROM nvidia/cuda:12.4.1-runtime-ubuntu22.04
 
 ENV DEBIAN_FRONTEND=noninteractive
 
@@ -16,8 +16,12 @@ RUN apt-get update && \
         build-essential \
         cmake  \
         curl \
+        wget \
         jq  \
         libcurl4-openssl-dev \
+    apt-transport-https \
+    ca-certificates \
+    gnupg \
         && \
     rm -rf /var/lib/apt/lists/*
 
@@ -27,24 +31,40 @@ RUN python3.11 -m ensurepip && python3.11 -m pip install --upgrade pip
 
 RUN pip install --upgrade --ignore-installed wheel==0.45.1
 
-WORKDIR /training_model
-COPY pyproject.toml poetry.lock ./
-RUN pip install poetry && poetry config virtualenvs.create false && poetry install --no-root --only main
-
-COPY training_model ./training_model
-COPY testing_model ./testing_model
-COPY data ./data
-COPY main.py .
+# Install Docker CLI so the training container can control Docker to run the RKLLM converter
+RUN mkdir -p /etc/apt/keyrings && \
+    curl -fsSL https://download.docker.com/linux/ubuntu/gpg | gpg --dearmor -o /etc/apt/keyrings/docker.gpg && \
+    echo "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.gpg] https://download.docker.com/linux/ubuntu $(lsb_release -cs) stable" \
+        > /etc/apt/sources.list.d/docker.list && \
+    apt-get update && apt-get install -y docker-ce-cli && rm -rf /var/lib/apt/lists/*
 
 WORKDIR /llama.cpp
 RUN git clone https://github.com/ggml-org/llama.cpp.git . && \
     cmake -B build && \
     cmake --build build --config Release
 
-WORKDIR /training_model
+
+WORKDIR /app
+COPY pyproject.toml poetry.lock ./
+RUN pip install poetry && \
+    pip install pyyaml && \
+    poetry lock --no-interaction --no-ansi || true && \
+    poetry install --no-root --only main --no-interaction --no-ansi
+
+# Create symlink for llama.cpp so training code can find it at expected relative path
+RUN ln -s /llama.cpp ./llama.cpp
+
+# Create symlink for quantize executable to match Windows naming convention
+RUN ln -s /llama.cpp/build/bin/llama-quantize /app/llama.cpp/llama-quantize.exe
+
+COPY training_model ./training_model
+COPY testing_model ./testing_model
+COPY evaluation ./evaluation
+COPY data ./data
+COPY conf ./conf
+COPY main.py .
 COPY run_pipeline.sh .
 
 RUN chmod +x run_pipeline.sh
 
 CMD ["./run_pipeline.sh"]
-#CMD python -m training_model

Dockerfile_cuda

@@ -23,6 +23,11 @@ RUN python3.11 -m ensurepip && python3.11 -m pip install --upgrade pip
 
 RUN pip install --upgrade --ignore-installed wheel==0.45.1
 
+WORKDIR /llama.cpp
+RUN ln -s /usr/local/cuda/lib64/stubs/libcuda.so /usr/local/cuda/lib64/stubs/libcuda.so.1 && \
+    git clone https://github.com/ggml-org/llama.cpp.git . && \
+    LD_LIBRARY_PATH=/usr/local/cuda/lib64/stubs:$LD_LIBRARY_PATH cmake -S . -B build -DGGML_CUDA=ON -DCMAKE_BUILD_TYPE=Release && \
+    LD_LIBRARY_PATH=/usr/local/cuda/lib64/stubs:$LD_LIBRARY_PATH cmake --build build --config Release -- -j"$(nproc)"
 WORKDIR /training_model
 COPY pyproject.toml poetry.lock ./
 RUN pip install poetry && poetry config virtualenvs.create false && poetry install --no-root --only main
@@ -32,12 +37,6 @@ COPY testing_model ./testing_model
 COPY data ./data
 COPY main.py .
 
-WORKDIR /llama.cpp
-RUN ln -s /usr/local/cuda/lib64/stubs/libcuda.so /usr/local/cuda/lib64/stubs/libcuda.so.1 && \
-    git clone https://github.com/ggml-org/llama.cpp.git . && \
-    LD_LIBRARY_PATH=/usr/local/cuda/lib64/stubs:$LD_LIBRARY_PATH cmake -B build -DGGML_CUDA=ON && \
-    LD_LIBRARY_PATH=/usr/local/cuda/lib64/stubs:$LD_LIBRARY_PATH cmake --build build --config Release
-
 WORKDIR /training_model
 
 CMD python -m training_model

Dockerfile_ollama

@@ -2,12 +2,12 @@ FROM ollama/ollama:latest
 
 RUN apt-get update && apt-get install -y curl
 
-VOLUME ["/root/.ollama", "/training_model/models"]
-RUN mkdir -p /training_model/models && chmod 777 /training_model/models
+VOLUME ["/root/.ollama", "/app/models"]
+RUN mkdir -p /app/models && chmod 777 /app/models
 
 EXPOSE 11434
 
-ENV OLLAMA_MODELS="/training_model/models"
+ENV OLLAMA_MODELS="/app/models"
 
 COPY entrypoint.sh /usr/local/bin/entrypoint.sh
 RUN chmod +x /usr/local/bin/entrypoint.sh