Adopt naming convention of transformers API (#14)

* Let's replace TPUModelForX with AutoModelForX to better match transformers

* Make AutoModelForX top-level import

* match new names

* Remove tpu from make pip install

* import version before anything else

* Let's use setuptools_scm backend to install deps

* Fix F401

* Let's just use ruff not black and ruff

* Again

.github/workflows/check_code_quality.yml

@@ -45,10 +45,6 @@ jobs:
         source venv/bin/activate
         pip install --upgrade pip
         pip install .[quality]
-    - name: Check style with black
-      run: |
-        source venv/bin/activate
-        black --check .
     - name: Check style with ruff
       run: |
         source venv/bin/activate

Makefile

@@ -51,12 +51,10 @@ tpu-tgi:
 
 # Run code quality checks
 style_check:
-	black --check .
 	ruff .
 
 style:
-	black .
-	ruff . --fix
+	ruff check . --fix
 
 # Utilities to release to PyPi
 build_dist_install_tools:
@@ -70,7 +68,7 @@ pypi_upload: ${PACKAGE_DIST} ${PACKAGE_WHEEL}
 
 # Tests
 test_installs:
-	python -m pip install .[tpu,tests]
+	python -m pip install .[tests]
 
 tests: test_installs
 	python -m pytest -sv tests

optimum/tpu/__init__.py

@@ -12,4 +12,5 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 
-from .version import __version__, VERSION  # noqa: F401
+from .version import __version__, VERSION   # noqa: F401
+from .modeling import AutoModelForCausalLM  # noqa: F401

optimum/tpu/modeling.py

@@ -19,13 +19,13 @@
 from typing import Any
 
 from loguru import logger
-from transformers import AutoModelForCausalLM
+from transformers import AutoModelForCausalLM as BaseAutoModelForCausalLM
 from transformers.utils import is_accelerate_available
 
 
 # TODO: For now TpuModelForCausalLM is just a shallow wrapper of
 # AutoModelForCausalLM, later this could be replaced by a custom class.
-class TpuModelForCausalLM(AutoModelForCausalLM):
+class AutoModelForCausalLM(BaseAutoModelForCausalLM):
 
     @classmethod
     def from_pretrained(
@@ -46,11 +46,11 @@ def from_pretrained(
         else:
             device = "xla"
         if is_accelerate_available():
-            model = AutoModelForCausalLM.from_pretrained(
+            model = BaseAutoModelForCausalLM.from_pretrained(
                 pretrained_model_name_or_path, device_map=device, *model_args, **kwargs
             )
         else:
-            model = AutoModelForCausalLM.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs)
+            model = BaseAutoModelForCausalLM.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs)
             model.to(device)
         # Update config with specific data)
         if task is not None or getattr(model.config, "task", None) is None:

pyproject.toml

@@ -48,6 +48,10 @@ dependencies = [
     "loguru == 0.6.0"
 ]
 
+[build-system]
+requires = ["setuptools>=64", "setuptools_scm>=8"]
+build-backend = "setuptools.build_meta"
+
 [project.optional-dependencies]
 tests = ["pytest", "safetensors"]
 quality = ["black", "ruff", "isort", "hf_doc_builder @ git+https://github.com/huggingface/doc-builder.git"]
@@ -58,8 +62,8 @@ Documentation = "https://hf.co/docs/optimum/tpu"
 Repository = "https://github.com/huggingface/optimum-tpu"
 Issues = "https://github.com/huggingface/optimum-tpu/issues"
 
-[tool.setuptools.dynamic]
-version = {attr = "optimum.tpu.__version__"}
+[tool.setuptools_scm]
+
 
 [tool.setuptools.packages.find]
 include = ["optimum.tpu"]

text-generation-inference/server/text_generation_server/generator.py

@@ -11,7 +11,7 @@
 from loguru import logger
 from transformers import AutoTokenizer, PreTrainedTokenizerBase, StaticCache
 from transformers.generation import GenerationConfig
-from optimum.tpu.modeling import TpuModelForCausalLM
+from optimum.tpu import AutoModelForCausalLM
 from optimum.tpu.generation import TokenSelector
 
 from .pb.generate_pb2 import (
@@ -301,7 +301,7 @@ class TpuGenerator(Generator):
 
     def __init__(
         self,
-        model: TpuModelForCausalLM,
+        model,
         tokenizer: PreTrainedTokenizerBase,
     ):
         self.model = model
@@ -633,7 +633,7 @@ def from_pretrained(
         """
         logger.info("Loading model (this can take a few minutes).")
         start = time.time()
-        model = TpuModelForCausalLM.from_pretrained(model_path)
+        model = AutoModelForCausalLM.from_pretrained(model_path)
         end = time.time()
         logger.info(f"Model successfully loaded in {end - start:.2f} s.")
         tokenizer = AutoTokenizer.from_pretrained(model_path)