Merge pull request #18 from SmudgedWings/end

'ci_check'

.github/workflows/lint.yml

@@ -0,0 +1,32 @@
+name: lint
+
+on:
+  pull_request:
+    branches:
+      - main  
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  lint:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          ref: ${{ github.head_ref }}
+
+      - name: Set up Python 3.10
+        uses: actions/setup-python@v4
+        with:
+          python-version: '3.10'
+
+      - name: Install pre-commit hook
+        run: |
+          pip install pre-commit mmengine
+          pre-commit install
+          
+      - name: Linting
+        run: pre-commit run   # --all-files

.github/workflows/main.yml

@@ -0,0 +1,61 @@
+name: Check
+
+on:
+  pull_request:
+    branches:
+      - main  
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          ref: ${{ github.head_ref }}
+
+      - name: Set up Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: '3.8'  
+
+      - name: Install dependencies
+        run: |
+          pip install -r requirements.txt
+      
+      - name: Download dataset
+        run: |
+          # pwd  # /home/runner/work/llmc/llmc
+          cd tools
+          python download_calib_dataset.py --save_path ../check/datasets/calib --dataset_name pileval
+          python download_eval_dataset.py --save_path ../check/datasets/eval --dataset_name wikitext2
+
+      - name: Download model
+        run: |
+          cd ci_check
+          mkdir -p opt-125m
+          cp model_urls.txt opt-125m/model_urls.txt
+          cd opt-125m
+          wget -i model_urls.txt
+          wget --no-check-certificate https://hf-mirror.com/facebook/opt-125m/resolve/main/pytorch_model.bin 
+
+      - name: Run push_tests
+        run: |
+          cd ci_check    # /home/runner/work/llmc/llmc/ci_check
+          python change_files.py
+          bash run.sh
+
+      - name: Check success
+        if: ${{ success() }}
+        run: echo "All steps completed successfully. Success!"
+
+      - name: Clean up
+        if: ${{ always() }}
+        run: |
+          cd ..
+          rm -rf opt-125m
+          rm -rf check

.pre-commit-config.yaml

@@ -0,0 +1,60 @@
+exclude: |
+    (?x)^(
+      imgs/|
+      docs/|
+      ci_check/|
+      requirements/|
+      scripts/
+    )
+repos:
+  - repo: https://github.com/PyCQA/flake8
+    rev: 5.0.4
+    hooks:
+      - id: flake8
+        exclude: configs/
+  - repo: https://github.com/PyCQA/isort
+    rev: 5.11.5
+    hooks:
+      - id: isort
+        exclude: configs/
+  - repo: https://github.com/pre-commit/mirrors-yapf
+    rev: v0.32.0
+    hooks:
+      - id: yapf
+        exclude: configs/
+  - repo: https://github.com/codespell-project/codespell
+    rev: v2.2.1
+    hooks:
+      - id: codespell
+        exclude: |
+            (?x)^(
+                .*\.jsonl|
+                .*\.md.template|
+                configs/
+            )
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v4.3.0
+    hooks:
+      - id: check-yaml
+      - id: requirements-txt-fixer
+      - id: double-quote-string-fixer
+      - id: check-merge-conflict
+      - id: fix-encoding-pragma
+        args: ["--remove"]
+      - id: mixed-line-ending
+        args: ["--fix=lf"]
+  - repo: https://github.com/executablebooks/mdformat
+    rev: 0.7.9
+    hooks:
+      - id: mdformat
+        args: ["--number", "--table-width", "200"]
+        additional_dependencies:
+          - mdformat-openmmlab
+          - mdformat_frontmatter
+          - linkify-it-py
+        exclude: configs/
+  - repo: https://github.com/myint/docformatter
+    rev: v1.3.1
+    hooks:
+      - id: docformatter
+        args: ["--in-place", "--wrap-descriptions", "88"]

ci_check/awq_w4a16_fakequant_eval.yml

@@ -0,0 +1,32 @@
+base:
+    seed: &seed 42
+model:
+    type: Opt
+    path: /home/runner/work/llmc/llmc/ci_check/opt-125m
+    torch_dtype: auto
+calib:
+    name: pileval
+    download: False
+    path: /home/runner/work/llmc/llmc/check/datasets/calib/pileval
+    n_samples: 4    # 128
+    bs: -1
+    seq_len: 16   # 512
+    preproc: pileval_awq
+    seed: *seed
+eval: 
+    eval_pos: [pretrain, transformed, fake_quant]
+    name: wikitext2 
+    download: False
+    path: /home/runner/work/llmc/llmc/check/datasets/eval/wikitext2
+    bs: 1
+    seq_len: 16       # 2048
+quant:
+    method: Awq
+    weight:
+        bit: 4
+        symmetric: False
+        granularity: per_group
+        group_size: 128
+save:
+    save_trans: False
+    save_path: /home/runner/work/llmc/llmc/save/opt-125m_awq_w4a16

ci_check/change_files.py

@@ -0,0 +1,97 @@
+import os
+
+# 文件路径
+cpu_txt_path = 'cpu.txt'
+
+def modify_file(filepath, modifications):
+    with open(filepath, 'r') as file:
+        lines = file.readlines()
+
+    # 应用修改
+    new_lines = []
+    for line in lines:
+        # 替换操作
+        for search, replace in modifications['modifications']:
+            if search in line:
+                line = line.replace(search, replace)
+        new_lines.append(line)
+
+    # 在文件开头插入新内容
+    with open(filepath, 'w') as file:
+        file.writelines(modifications['header'] + new_lines)
+
+def main():
+    with open(cpu_txt_path, 'r') as file:
+        file_paths = file.readlines()
+
+    for file_path in file_paths:
+        file_path = file_path.strip()
+        if not file_path:
+            continue
+
+        if file_path == '../llmc/__main__.py':
+            modifications = {
+                'header': [
+                    'device_zbl = "cpu"\n',
+                    'use_cuda = (device_zbl != "cpu")\n'
+                ],
+                'modifications': [
+                    ('torch.cuda.empty_cache()', 'if use_cuda: torch.cuda.empty_cache()')
+                ]
+            }
+        elif file_path == '../llmc/compression/quantization/awq.py':
+            modifications = {
+                'header': [
+                    'n_grid_zbl = 1\n'
+                ],
+                'modifications': [
+                    ('n_grid = 20', 'n_grid = n_grid_zbl')
+                ]
+            }
+        elif file_path == '../llmc/compression/quantization/base_blockwise_quantization.py':
+            modifications = {
+                'header': [
+                    'device_zbl = "cpu"\n',
+                    'use_cuda = (device_zbl != "cpu")\n'
+                ],
+                'modifications': [
+                    ('.cuda()', '.to(device_zbl)'),
+                    ('torch.cuda.empty_cache()', 'if use_cuda: torch.cuda.empty_cache()')
+                ]
+            }
+        elif file_path == '../llmc/models/base_model.py':
+            modifications = {
+                'header': [
+                    'device_zbl = "cpu"\n',
+                    'use_cuda = (device_zbl != "cpu")\n'
+                ],
+                'modifications': [
+                    ('.cuda()', '.to(device_zbl)'),
+                    ('self.move_embed_to_device("cuda")', 'self.move_embed_to_device(device_zbl)')
+                ]
+            }
+        elif file_path == '../llmc/eval/eval_ppl.py':
+            modifications = {
+                'header': [
+                    'device_zbl = "cpu"\n',
+                    'use_cuda = (device_zbl != "cpu")\n',
+                    'nsamples_zbl = 1\n'
+                ],
+                'modifications': [
+                    ('.cuda()', '.to(device_zbl)'),
+                    ('torch.cuda.empty_cache()', 'if use_cuda: torch.cuda.empty_cache()'),
+                    ('nlls = []', 'nlls = []; nsamples = nsamples_zbl')
+                ]
+            }
+        else:
+            print(f"File {file_path} not recognized or not specified for modification.")
+            continue
+
+        # 修改文件
+        if os.path.exists(file_path):
+            modify_file(file_path, modifications)
+        else:
+            print(f"File {file_path} does not exist.")
+
+if __name__ == "__main__":
+    main()

ci_check/cpu.txt

@@ -0,0 +1,5 @@
+../llmc/compression/quantization/base_blockwise_quantization.py
+../llmc/__main__.py
+../llmc/eval/eval_ppl.py
+../llmc/compression/quantization/awq.py
+../llmc/models/base_model.py

ci_check/model_urls.txt

@@ -0,0 +1,6 @@
+https://hf-mirror.com/facebook/opt-125m/resolve/main/config.json
+https://hf-mirror.com/facebook/opt-125m/resolve/main/generation_config.json
+https://hf-mirror.com/facebook/opt-125m/resolve/main/merges.txt
+https://hf-mirror.com/facebook/opt-125m/resolve/main/special_tokens_map.json
+https://hf-mirror.com/facebook/opt-125m/resolve/main/tokenizer_config.json
+https://hf-mirror.com/facebook/opt-125m/resolve/main/vocab.json

ci_check/run.sh

@@ -0,0 +1,10 @@
+#!/bin/bash
+
+current_directory=$(pwd)
+llmc=$(echo "$current_directory" | sed 's/\/ci_check$//')
+export PYTHONPATH=$llmc:$PYTHONPATH
+
+
+cd ../scripts
+
+python -m llmc --config ../ci_check/awq_w4a16_fakequant_eval.yml