merge master

turbo-llm · Sep 9, 2024 · cdc591c · cdc591c
2 parents 6696370 + 184e19c
commit cdc591c
Show file tree

Hide file tree

Showing 47 changed files with 7,788 additions and 90 deletions.
diff --git a/.github/workflows/tests.yml b/.github/workflows/tests.yml
@@ -0,0 +1,46 @@
+name: Push
+on: [push]
+
+jobs:
+  tests:
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ['3.10']
+        poetry-version: ['1.2.2']
+        os: [ubuntu-latest]
+    runs-on: ${{ matrix.os }}
+    steps:
+      - uses: actions/checkout@v3
+      - uses: actions/setup-python@v3
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Run image
+        uses: abatilo/[email protected]
+        with:
+          poetry-version: ${{ matrix.poetry-version }}
+      - name: Install dependencies
+        run: poetry install --extras "deepspeed gpu"
+      - name: Run tests
+        run: poetry run make tests
+  linters:
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ['3.10']
+        poetry-version: ['1.2.2']
+        os: [ubuntu-latest]
+    runs-on: ${{ matrix.os }}
+    steps:
+      - uses: actions/checkout@v3
+      - uses: actions/setup-python@v3
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Run image
+        uses: abatilo/[email protected]
+        with:
+          poetry-version: ${{ matrix.poetry-version }}
+      - name: Install dependencies
+        run: poetry install --extras "deepspeed gpu"
+      - name: Run linters
+        run: poetry run make lint
diff --git a/.gitignore b/.gitignore
@@ -53,6 +53,7 @@ coverage.xml
 *.cover
 .hypothesis/
 .pytest_cache/
+!tests/fixtures/models/llama2_tiny_multimodal_clip_mlp/projections/modality_adapters.pt
 
 # Translations
 *.mo
@@ -99,7 +100,6 @@ ENV/
 env.bak/
 venv.bak/
 default_tb/
-poetry.lock
 
 # Spyder project settings
 .spyderproject

diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
@@ -0,0 +1,9 @@
+repos:
+-   repo: local
+    hooks:
+    -   id:  poetry-make-pretty
+        name: make pretty
+        entry: poetry run make pretty
+        language: system
+        types: [python]
+        stages: [commit]
diff --git a/Makefile b/Makefile
@@ -27,7 +27,7 @@ flake:
 	flake8 --max-line-length 119 --jobs $(JOBS) --statistics $${CI:+--format=gl-codeclimate --output=codeclimate-flake8.json} $(CODE)
 
 pylint:
-	pylint --jobs $(JOBS) --rcfile=setup.cfg $${CI:+--output-format=pylint_gitlab.GitlabCodeClimateReporter --output=codeclimate-pylint.json} $(CODE)
+	pylint --jobs $(JOBS) --rcfile=setup.cfg $(CODE)
 
 mypy:
 	mypy --config-file mypy.ini $(CODE) --show-traceback
@@ -39,9 +39,6 @@ ifneq ($(CODE),)
 	unify --in-place --recursive $(CODE) tutorials
 endif
 
-bash-dev:
-	docker run -v $(shell pwd):/app --rm -it ${DEV_IMAGE} bash
-
 lock:
 	rm poetry.lock
 	poetry lock

diff --git a/configs/exp/train/classification/classification.json b/configs/exp/train/classification/classification.json
@@ -86,6 +86,11 @@
         }
     },
     "tokenizer_settings": {},
+    "special_tokens_settings": {
+        "bos_token": "<s>",
+        "eos_token": "</s>",
+        "pad_token": "<pad>"
+    },
     "trainer_settings": {
         "evaluation_strategy": "steps",
         "per_device_train_batch_size": 1,

diff --git a/configs/exp/train/dpo/dpo.json b/configs/exp/train/dpo/dpo.json
@@ -96,6 +96,10 @@
           }
       },
       "tokenizer_settings": {},
+      "special_tokens_settings": {
+        "bos_token": "<|begin_of_text|>",
+        "eos_token": "<|end_of_text|>"
+    },
       "trainer_settings": {
         "evaluation_strategy": "steps",
         "per_device_train_batch_size": 1,

diff --git a/configs/exp/train/multimodal/c_abs.json b/configs/exp/train/multimodal/c_abs.json
@@ -105,6 +105,10 @@
     "tokenizer_settings": {
         "tokenizer_path": "/from_s3/model"
     },
+    "special_tokens_settings": {
+        "bos_token": "<s>",
+        "eos_token": "</s>"
+    },
     "trainer_settings": {
         "evaluation_strategy": "epoch",
         "save_strategy": "epoch",

diff --git a/configs/exp/train/multimodal/llava.json b/configs/exp/train/multimodal/llava.json
diff --git a/configs/exp/train/multimodal/mlp.json b/configs/exp/train/multimodal/mlp.json
@@ -105,6 +105,10 @@
     "tokenizer_settings": {
         "tokenizer_path": "/from_s3/model"
     },
+    "special_tokens_settings": {
+        "bos_token": "<s>",
+        "eos_token": "</s>"
+    },
     "trainer_settings": {
         "evaluation_strategy": "epoch",
         "save_strategy": "epoch",

diff --git a/configs/exp/train/rag/end2end_rag.json b/configs/exp/train/rag/end2end_rag.json
@@ -112,6 +112,10 @@
     "metric_settings": []
   },
   "tokenizer_settings": {},
+  "special_tokens_settings": {
+    "bos_token": "<s>",
+    "eos_token": "</s>"
+  },
   "trainer_settings": {
     "evaluation_strategy": "steps",
     "save_strategy": "steps",

diff --git a/configs/exp/train/rm/rm.json b/configs/exp/train/rm/rm.json
@@ -83,6 +83,10 @@
         }
     },
     "tokenizer_settings": {},
+    "special_tokens_settings": {
+        "bos_token": "<|begin_of_text|>",
+        "eos_token": "<|end_of_text|>"
+    },
     "trainer_settings": {
         "evaluation_strategy": "steps",
         "per_device_train_batch_size": 1,