Merge branch 'master' into video-inference

Author: deigen

CHANGELOG.md

@@ -1,3 +1,21 @@
+## [[11.1.4]](https://github.com/Clarifai/clarifai-python/releases/tag/11.1.4) - [PyPI](https://pypi.org/project/clarifai/11.1.4/) - 2025-02-12
+
+### Changed
+
+ - Introduce 3 times when you can download checkpoints [(#515)] (https://github.com/Clarifai/clarifai-python/pull/515)
+
+## [[11.1.3]](https://github.com/Clarifai/clarifai-python/releases/tag/11.1.3) - [PyPI](https://pypi.org/project/clarifai/11.1.3/) - 2025-02-11
+
+### Changed
+
+ - Fix dependency parsing [(#514)] (https://github.com/Clarifai/clarifai-python/pull/514)
+
+## [[11.1.2]](https://github.com/Clarifai/clarifai-python/releases/tag/11.1.2) - [PyPI](https://pypi.org/project/clarifai/11.1.2/) - 2025-02-10
+
+### Changed
+
+ - User new base images and fix clarifai version [(#513)] (https://github.com/Clarifai/clarifai-python/pull/513)
+
 ## [[11.1.1]](https://github.com/Clarifai/clarifai-python/releases/tag/11.1.1) - [PyPI](https://pypi.org/project/clarifai/11.1.1/) - 2025-02-06
 
 ### Changed

clarifai/__init__.py

@@ -1 +1 @@
-__version__ = "11.1.1"
+__version__ = "11.1.4"

clarifai/cli/model.py

@@ -15,21 +15,24 @@ def model():
     required=True,
     help='Path to the model directory.')
 @click.option(
-    '--download_checkpoints',
-    is_flag=True,
+    '--stage',
+    required=False,
+    type=click.Choice(['runtime', 'build', 'upload'], case_sensitive=True),
+    default="upload",
+    show_default=True,
     help=
-    'Flag to download checkpoints before uploading and including them in the tar file that is uploaded. Defaults to False, which will attempt to download them at docker build time.',
+    'The stage we are calling download checkpoints from. Typically this would "upload" and will download checkpoints if config.yaml checkpoints section has when set to "upload". Other options include "runtime" to be used in load_model or "upload" to be used during model upload. Set this stage to whatever you have in config.yaml to force downloading now.'
 )
 @click.option(
     '--skip_dockerfile',
     is_flag=True,
     help=
     'Flag to skip generating a dockerfile so that you can manually edit an already created dockerfile.',
 )
-def upload(model_path, download_checkpoints, skip_dockerfile):
+def upload(model_path, stage, skip_dockerfile):
   """Upload a model to Clarifai."""
   from clarifai.runners.models.model_builder import upload_model
-  upload_model(model_path, download_checkpoints, skip_dockerfile)
+  upload_model(model_path, stage, skip_dockerfile)
 
 
 @model.command()
@@ -44,14 +47,23 @@ def upload(model_path, download_checkpoints, skip_dockerfile):
     required=False,
     default=None,
     help=
-    'Option path to write the checkpoints to. This will place them in {out_path}/ If not provided it will default to {model_path}/1/checkpoints where the config.yaml is read..'
+    'Option path to write the checkpoints to. This will place them in {out_path}/1/checkpoints If not provided it will default to {model_path}/1/checkpoints where the config.yaml is read.'
+)
+@click.option(
+    '--stage',
+    required=False,
+    type=click.Choice(['runtime', 'build', 'upload'], case_sensitive=True),
+    default="build",
+    show_default=True,
+    help=
+    'The stage we are calling download checkpoints from. Typically this would be in the build stage which is the default. Other options include "runtime" to be used in load_model or "upload" to be used during model upload. Set this stage to whatever you have in config.yaml to force downloading now.'
 )
-def download_checkpoints(model_path, out_path):
+def download_checkpoints(model_path, out_path, stage):
   """Download checkpoints from external source to local model_path"""
 
   from clarifai.runners.models.model_builder import ModelBuilder
   builder = ModelBuilder(model_path, download_validation_only=True)
-  builder.download_checkpoints(out_path)
+  builder.download_checkpoints(stage=stage, checkpoint_path_override=out_path)
 
 
 @model.command()
@@ -159,6 +171,18 @@ def run_locally(model_path, port, mode, keep_env, keep_image):
     click.echo(f"Failed to starts model server locally: {e}", err=True)
 
 
+@model.command()
+@click.option(
+    '--model_path',
+    type=click.Path(exists=True),
+    required=True,
+    help='Path to the model directory.')
+def local_dev(model_path):
+  """Run the model as a local dev runner to help debug your model connected to the API. You must set several envvars such as CLARIFAI_PAT, CLARIFAI_RUNNER_ID, CLARIFAI_NODEPOOL_ID, CLARIFAI_COMPUTE_CLUSTER_ID. """
+  from clarifai.runners.server import serve
+  serve(model_path)
+
+
 @model.command()
 @click.option(
     '--config',

clarifai/client/model.py

@@ -27,7 +27,7 @@
 from clarifai.runners.utils import video_utils
 from clarifai.urls.helper import ClarifaiUrlHelper
 from clarifai.utils.logging import logger
-from clarifai.utils.misc import BackoffIterator
+from clarifai.utils.misc import BackoffIterator, status_is_retryable
 from clarifai.utils.model_train import (find_and_replace_key, params_parser,
                                         response_to_model_params, response_to_param_info,
                                         response_to_templates)
@@ -439,7 +439,7 @@ def predict(self,
     while True:
       response = self._grpc_request(self.STUB.PostModelOutputs, request)
 
-      if response.status.code == status_code_pb2.MODEL_DEPLOYING and \
+      if status_is_retryable(response.status.code) and \
               time.time() - start_time < 60 * 10:  # 10 minutes
         self.logger.info(f"{self.id} model is still deploying, please wait...")
         time.sleep(next(backoff_iterator))
@@ -722,7 +722,7 @@ def generate(self,
         break
       stream_response = self._grpc_request(self.STUB.GenerateModelOutputs, request)
       for response in stream_response:
-        if response.status.code == status_code_pb2.MODEL_DEPLOYING and \
+        if status_is_retryable(response.status.code) and \
                 time.time() - start_time < 60 * 10:
           self.logger.info(f"{self.id} model is still deploying, please wait...")
           time.sleep(next(backoff_iterator))
@@ -967,7 +967,7 @@ def stream(self,
         break
       stream_response = self._grpc_request(self.STUB.StreamModelOutputs, request)
       for response in stream_response:
-        if response.status.code == status_code_pb2.MODEL_DEPLOYING and \
+        if status_is_retryable(response.status.code) and \
                 time.time() - start_time < 60 * 10:
           self.logger.info(f"{self.id} model is still deploying, please wait...")
           time.sleep(next(backoff_iterator))

clarifai/client/workflow.py

@@ -15,7 +15,7 @@
 from clarifai.errors import UserError
 from clarifai.urls.helper import ClarifaiUrlHelper
 from clarifai.utils.logging import logger
-from clarifai.utils.misc import BackoffIterator
+from clarifai.utils.misc import BackoffIterator, status_is_retryable
 from clarifai.workflows.export import Exporter
 
 
@@ -99,7 +99,7 @@ def predict(self, inputs: List[Input], workflow_state_id: str = None):
     while True:
       response = self._grpc_request(self.STUB.PostWorkflowResults, request)
 
-      if response.status.code == status_code_pb2.MODEL_DEPLOYING and \
+      if status_is_retryable(response.status.code) and \
           time.time() - start_time < 60*10:  # 10 minutes
         self.logger.info(f"{self.id} Workflow is still deploying, please wait...")
         time.sleep(next(backoff_iterator))

clarifai/runners/dockerfile_template/Dockerfile.template

@@ -1,30 +1,11 @@
-# syntax=docker/dockerfile:1
-#############################
-# User specific requirements installed in the pip_packages
-#############################
-FROM --platform=$TARGETPLATFORM ${BUILDER_IMAGE} as pip_packages
+# syntax=docker/dockerfile:1.13-labs
+FROM --platform=$TARGETPLATFORM ${FINAL_IMAGE} as final
 
 COPY --link requirements.txt /home/nonroot/requirements.txt
 
 # Update clarifai package so we always have latest protocol to the API. Everything should land in /venv
-RUN pip install --no-cache-dir -r /home/nonroot/requirements.txt && \
-    (pip install --upgrade --upgrade-strategy only-if-needed --no-deps --no-cache-dir clarifai clarifai-grpc clarifai-protocol || true)
-#############################
-
-#############################
-# Downloader dependencies image
-#############################
-FROM --platform=$TARGETPLATFORM ${DOWNLOADER_IMAGE} as downloader
-
-# make sure we have the latest clarifai package.
-RUN (pip install --upgrade --upgrade-strategy only-if-needed --no-cache-dir clarifai clarifai-grpc clarifai-protocol || true)
-#####
-
-
-#############################
-# Final runtime image
-#############################
-FROM --platform=$TARGETPLATFORM ${RUNTIME_IMAGE} as final
+RUN ["pip", "install", "--no-cache-dir", "-r", "/home/nonroot/requirements.txt"]
+RUN ["pip", "show", "clarifai"]
 
 # Set the NUMBA cache dir to /tmp
 # Set the TORCHINDUCTOR cache dir to /tmp
@@ -34,28 +15,16 @@ ENV NUMBA_CACHE_DIR=/tmp/numba_cache \
     HOME=/tmp \
     DEBIAN_FRONTEND=noninteractive
 
-#####
-# Copy the python requirements needed to download checkpoints
-#####
-COPY --link=true --from=downloader /venv /venv
-#####
-
 #####
 # Copy the files needed to download
 #####
 # This creates the directory that HF downloader will populate and with nonroot:nonroot permissions up.
 COPY --chown=nonroot:nonroot downloader/unused.yaml /home/nonroot/main/1/checkpoints/.cache/unused.yaml
 
 #####
-# Download checkpoints
+# Download checkpoints if config.yaml has checkpoints.when = "build"
 COPY --link=true config.yaml /home/nonroot/main/
-RUN ["python", "-m", "clarifai.cli", "model", "download-checkpoints", "--model_path", "/home/nonroot/main", "--out_path", "/home/nonroot/main"]
-#####
-
-
-#####
-# Copy the python packages from the previous stage.
-COPY --link=true --from=pip_packages /venv /venv
+RUN ["python", "-m", "clarifai.cli", "model", "download-checkpoints", "--model_path", "/home/nonroot/main", "--out_path", "/home/nonroot/main/1/checkpoints", "--stage", "build"]
 #####
 
 # Copy in the actual files like config.yaml, requirements.txt, and most importantly 1/model.py