clarifai 10.11.1__py3-none-any.whl → 10.11.2rc1__py3-none-any.whl

clarifai/runners/deepgram_runner.py

@@ -0,0 +1,131 @@
+import os
+import time
+import httpx
+from dotenv import load_dotenv
+import threading
+import queue
+import hashlib
+
+from deepgram import (
+    DeepgramClient,
+    LiveTranscriptionEvents,
+    LiveOptions,
+)
+
+load_dotenv()
+
+os.environ["DG_API_KEY"] = "1d5666523f0f2fbaf3e2db6ae7717f6f87280b5e"
+API_KEY = os.getenv("DG_API_KEY")
+
+from clarifai_grpc.grpc.api import resources_pb2, service_pb2
+from collections.abc import Iterator
+from google.protobuf import json_format
+
+from clarifai.client.runner import Runner
+
+
+class MyRunner(Runner):
+  """A custom runner that adds "Hello World" to the end of the text and replaces the domain of the
+  image URL as an example.
+  """
+
+  def setup_connection(self):
+    print("ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZz")
+    # STEP 2: Create a websocket connection to Deepgram
+    self.dg_connection = self.deepgram.listen.live.v("1")
+
+    output_q = self.output_q
+
+    # STEP 3: Define the event handlers for the connection
+    def on_message(self, result, **kwargs):
+      if result.is_final:
+        sentence = result.channel.alternatives[0].transcript
+        if len(sentence) == 0:
+          return
+        print("SSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSS")
+        print(f"speaker: {sentence}")
+        # put it on a queue as we get responses from deepgram.
+        output_q.put(sentence)
+
+    def on_metadata(self, metadata, **kwargs):
+      print(f"\n\n{metadata}\n\n")
+
+    def on_error(self, error, **kwargs):
+      print(f"\n\n{error}\n\n")
+
+    # STEP 4: Register the event handlers
+    self.dg_connection.on(LiveTranscriptionEvents.Transcript, on_message)
+    self.dg_connection.on(LiveTranscriptionEvents.Metadata, on_metadata)
+    self.dg_connection.on(LiveTranscriptionEvents.Error, on_error)
+
+    # STEP 5: Configure Deepgram options for live transcription
+    self.options = LiveOptions(
+        model="nova-2",
+        language="en-US",
+        smart_format=True,
+    )
+
+    # STEP 6: Start the connection
+    self.dg_connection.start(self.options)
+
+    print("DDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDD")
+    print(self.dg_connection._socket)
+
+  def __init__(self, *args, **kwargs):
+    print("MyRunner init")
+    # STEP 1: Create a Deepgram client using the API key
+    self.deepgram = DeepgramClient(API_KEY)
+    self.output_q = queue.Queue()
+
+    self.setup_connection()
+
+    super().__init__(*args, **kwargs)
+
+  def stream(self, request: service_pb2.PostModelOutputsRequest
+            ) -> Iterator[service_pb2.MultiOutputResponse]:
+    """Example yielding a whole batch of streamed stuff back.
+    """
+
+    assert len(request.inputs) == 1, "This runner only supports one input at a time."
+
+    # Get the next chunk of data from the incoming stream.
+
+    print("Got some audio data")
+
+    data = request.inputs[0].data.audio.base64
+    print(hashlib.md5(data).hexdigest())
+
+    # FIXME(zeiler): this doesnt' work but isn't iportant to our system.
+    if not self.dg_connection._socket:
+      #self.dg_connection.finish()
+      #self.dg_connection.start(self.options)
+      self.setup_connection()
+    self.dg_connection.send(data)
+    print("Sent it to deepgram")
+
+    while True:
+      try:
+        item = self.output_q.get(timeout=0.1)
+        output = resources_pb2.Output()
+        output.data.text.raw = item
+        output.status.code = 10000
+        print("YYYYYYYYYYYYYYYYYYYYYYYYYYYYYY")
+        print("Yielding: ", output.data.text.raw)
+        yield service_pb2.MultiOutputResponse(outputs=[
+            output,
+        ])
+      except queue.Empty:
+        print("Queue is empty, sleeping then breaking")
+        break
+
+  # # STEP 13: Close the connection to Deepgram
+  # dg_connection.finish()
+
+
+if __name__ == '__main__':
+  # Make sure you set these env vars before running the example.
+  # CLARIFAI_PAT
+  # CLARIFAI_USER_ID
+
+  # You need to first create a runner in the Clarifai API and then use the ID here.
+  MyRunner(runner_id="matt-test-runner", base_url="http://q6:32013", num_parallel_polls=1).start()

clarifai/runners/deepgram_runner.py~

@@ -0,0 +1,130 @@
+import os
+import time
+import httpx
+from dotenv import load_dotenv
+import threading
+import queue
+import hashlib
+
+from deepgram import (
+    DeepgramClient,
+    LiveTranscriptionEvents,
+    LiveOptions,
+)
+
+load_dotenv()
+
+os.environ["DG_API_KEY"] = "1d5666523f0f2fbaf3e2db6ae7717f6f87280b5e"
+API_KEY = os.getenv("DG_API_KEY")
+
+from clarifai_grpc.grpc.api import resources_pb2, service_pb2
+from collections.abc import Iterator
+from google.protobuf import json_format
+
+from clarifai.client.runner import Runner
+
+
+class MyRunner(Runner):
+  """A custom runner that adds "Hello World" to the end of the text and replaces the domain of the
+  image URL as an example.
+  """
+
+  def setup_connection(self):
+    print("ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZz")
+    # STEP 2: Create a websocket connection to Deepgram
+    self.dg_connection = self.deepgram.listen.live.v("1")
+
+    output_q = self.output_q
+
+    # STEP 3: Define the event handlers for the connection
+    def on_message(self, result, **kwargs):
+      if result.is_final:
+        sentence = result.channel.alternatives[0].transcript
+        if len(sentence) == 0:
+          return
+        print("SSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSS")
+        print(f"speaker: {sentence}")
+        # put it on a queue as we get responses from deepgram.
+        output_q.put(sentence)
+
+    def on_metadata(self, metadata, **kwargs):
+      print(f"\n\n{metadata}\n\n")
+
+    def on_error(self, error, **kwargs):
+      print(f"\n\n{error}\n\n")
+
+    # STEP 4: Register the event handlers
+    self.dg_connection.on(LiveTranscriptionEvents.Transcript, on_message)
+    self.dg_connection.on(LiveTranscriptionEvents.Metadata, on_metadata)
+    self.dg_connection.on(LiveTranscriptionEvents.Error, on_error)
+
+    # STEP 5: Configure Deepgram options for live transcription
+    self.options = LiveOptions(
+        model="nova-2",
+        language="en-US",
+        smart_format=True,
+    )
+
+    # STEP 6: Start the connection
+    self.dg_connection.start(self.options)
+
+    print("DDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDD")
+    print(self.dg_connection._socket)
+
+  def __init__(self, *args, **kwargs):
+    print("MyRunner init")
+    # STEP 1: Create a Deepgram client using the API key
+    self.deepgram = DeepgramClient(API_KEY)
+    self.output_q = queue.Queue()
+
+    self.setup_connection()
+
+    super().__init__(*args, **kwargs)
+
+  def stream(self, request: service_pb2.PostModelOutputsRequest
+            ) -> Iterator[service_pb2.MultiOutputResponse]:
+    """Example yielding a whole batch of streamed stuff back.
+    """
+
+    assert len(request.inputs) == 1, "This runner only supports one input at a time."
+
+    # Get the next chunk of data from the incoming stream.
+
+    print("Got some audio data")
+
+    data = request.inputs[0].data.audio.base64
+    print(hashlib.md5(data).hexdigest())
+
+    if not self.dg_connection._socket:
+      self.dg_connection.finish()
+      self.dg_connection.start(self.options)
+      #self.setup_connection()
+    self.dg_connection.send(data)
+    print("Sent it to deepgram")
+
+    while True:
+      try:
+        item = self.output_q.get(timeout=0.1)
+        output = resources_pb2.Output()
+        output.data.text.raw = item
+        output.status.code = 10000
+        print("YYYYYYYYYYYYYYYYYYYYYYYYYYYYYY")
+        print("Yielding: ", output.data.text.raw)
+        yield service_pb2.MultiOutputResponse(outputs=[
+            output,
+        ])
+      except queue.Empty:
+        print("Queue is empty, sleeping then breaking")
+        break
+
+  # # STEP 13: Close the connection to Deepgram
+  # dg_connection.finish()
+
+
+if __name__ == '__main__':
+  # Make sure you set these env vars before running the example.
+  # CLARIFAI_PAT
+  # CLARIFAI_USER_ID
+
+  # You need to first create a runner in the Clarifai API and then use the ID here.
+  MyRunner(runner_id="matt-test-runner", base_url="http://q6:32013", num_parallel_polls=1).start()

clarifai/runners/dockerfile_template/Dockerfile.cpu.template

@@ -0,0 +1,31 @@
+ARG PYTHON_VERSION=${PYTHON_VERSION}
+FROM public.ecr.aws/docker/library/python:${PYTHON_VERSION}-slim-bookworm as build
+
+# Set the working directory to /app
+WORKDIR /app
+
+COPY requirements.txt .
+# Install requirements and cleanup before leaving this line.
+# Note(zeiler): this could be in a future template as {{model_python_deps}}
+RUN python -m pip install -r requirements.txt && rm -rf /root/.cache
+
+# Install Clarifai SDK
+RUN python -m pip install clarifai
+
+# These will be set by the templaing system.
+ENV CLARIFAI_PAT=${CLARIFAI_PAT}
+ENV CLARIFAI_USER_ID=${CLARIFAI_USER_ID}
+ENV CLARIFAI_RUNNER_ID=${CLARIFAI_RUNNER_ID}
+ENV CLARIFAI_NODEPOOL_ID=${CLARIFAI_NODEPOOL_ID}
+ENV CLARIFAI_COMPUTE_CLUSTER_ID=${CLARIFAI_COMPUTE_CLUSTER_ID}
+ENV CLARIFAI_API_BASE=${CLARIFAI_API_BASE}
+
+# Copy the current folder into /app/model_dir that the SDK will expect.
+COPY . /app/model_dir/${name}
+
+# Add the model directory to the python path.
+ENV PYTHONPATH "${PYTHONPATH}:/app/model_dir/${name}"
+
+# Finally run the clarifai entrypoint to start the runner loop and local dev server.
+# Note(zeiler): we may want to make this a clarifai CLI call.
+CMD ["python", "-m", "clarifai.runners.server", "--model_path", "/app/model_dir/${name}"]

clarifai/runners/dockerfile_template/Dockerfile.cuda.template

@@ -0,0 +1,79 @@
+# Build a virtualenv containing necessary system libraries and Python packages
+# for users to install their own packages while also being distroless.
+# * Install python3-venv
+# * Install gcc libpython3-dev to compile C Python modules
+# * In the virtualenv: Update pip setuputils and wheel to support building new packages
+# * Export environment variables to use the virtualenv by default
+# * Create a non-root user with minimal privileges and use it
+ARG TARGET_PLATFORM=linux/amd64
+FROM --platform=$TARGET_PLATFORM public.ecr.aws/docker/library/python:${PYTHON_VERSION}-slim-bookworm as build
+
+ENV DEBIAN_FRONTEND=noninteractive
+RUN apt-get update && \
+    apt-get install --no-install-suggests --no-install-recommends --yes \
+    software-properties-common  \
+    gcc \
+    libpython3-dev && \
+    python${PYTHON_VERSION} -m venv /venv && \
+    /venv/bin/pip install --disable-pip-version-check --upgrade pip setuptools wheel && \
+    apt-get clean && rm -rf /var/lib/apt/lists/*
+
+# Set environment variables to use virtualenv by default
+ENV VIRTUAL_ENV=/venv
+ENV PATH="$VIRTUAL_ENV/bin:$PATH"
+
+#############################
+# User specific requirements
+#############################
+COPY requirements.txt .
+
+# Install requirements and cleanup before leaving this line.
+# Note(zeiler): this could be in a future template as {{model_python_deps}}
+RUN python -m pip install -r requirements.txt && rm -rf /root/.cache
+
+# Install Clarifai SDK
+RUN python -m pip install clarifai
+
+#############################
+# Finally copy everything we built into a distroless image for runtime.
+######################>#######
+ARG TARGET_PLATFORM=linux/amd64
+FROM --platform=$TARGET_PLATFORM gcr.io/distroless/python3-debian12:latest
+# FROM --platform=$TARGET_PLATFORM gcr.io/distroless/python3-debian12:debug
+ARG PYTHON_VERSION=${PYTHON_VERSION}
+# needed to call pip directly
+COPY --from=build /bin/sh /bin/sh
+
+# virtual env
+COPY --from=build /venv /venv
+
+# We have to overwrite the python3 binary that the distroless image uses
+COPY --from=build /usr/local/bin/python${PYTHON_VERSION} /usr/bin/python3
+# And also copy in all the lib files for it.
+COPY --from=build /usr/local/lib/ /usr/lib/
+
+# Set environment variables to use virtualenv by default
+ENV VIRTUAL_ENV=/venv
+ENV PYTHONPATH=${PYTHONPATH}:${VIRTUAL_ENV}/lib/python${PYTHON_VERSION}/site-packages
+
+# These will be set by the templaing system.
+ENV CLARIFAI_PAT=${CLARIFAI_PAT}
+ENV CLARIFAI_USER_ID=${CLARIFAI_USER_ID}
+ENV CLARIFAI_RUNNER_ID=${CLARIFAI_RUNNER_ID}
+ENV CLARIFAI_NODEPOOL_ID=${CLARIFAI_NODEPOOL_ID}
+ENV CLARIFAI_COMPUTE_CLUSTER_ID=${CLARIFAI_COMPUTE_CLUSTER_ID}
+ENV CLARIFAI_API_BASE=${CLARIFAI_API_BASE}
+
+# Set the working directory to /app
+WORKDIR /app
+
+# Copy the current folder into /app/model_dir that the SDK will expect.
+# Note(zeiler): would be nice to exclude checkpoints in case they were pre-downloaded.
+COPY . /app/model_dir/${name}
+
+# Add the model directory to the python path.
+ENV PYTHONPATH=${PYTHONPATH}:/app/model_dir/${name}
+
+# Finally run the clarifai entrypoint to start the runner loop and local dev server.
+# Note(zeiler): we may want to make this a clarifai CLI call.
+CMD ["-m", "clarifai.runners.server", "--model_path", "/app/model_dir/${name}"]

clarifai/runners/example_llama2.py~

@@ -0,0 +1,72 @@
+from clarifai.client.runner import Runner
+from clarifai_grpc.grpc.api import resources_pb2
+
+# This example requires to run the following before running this example:
+# pip install transformers
+
+# https://huggingface.co/TheBloke/Llama-2-70B-chat-GPTQ
+model_name_or_path = "TheBloke/Llama-2-7B-chat-GPTQ"
+model_basename = "model"
+
+use_triton = False
+
+
+class Llama2Runner(Runner):
+  """A custom runner that runs the LLama2 LLM.
+  """
+
+  def __init__(self, *args, **kwargs):
+    print("Starting to load the model...")
+    st = time.time()
+    self.tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True)
+    self.model = AutoModelForCausalLM.from_pretrained(model_name_or_path, device_map='auto')
+
+    self.logger.info("Loading model complete in (%f seconds), ready to loop for requests." %
+                     (time.time() - st))
+    super(MyRunner, self).__init__(*args, **kwargs)
+
+  def run_input(self, input: resources_pb2.Input,
+                output_info: resources_pb2.OutputInfo) -> resources_pb2.Output:
+    """This is the method that will be called when the runner is run. It takes in an input and
+    returns an output.
+    """
+
+    output = resources_pb2.Output()
+    data = input.data
+    if data.text.raw != "":
+      input_text = data.text.raw
+    elif data.text.url != "":
+      input_text = str(requests.get(data.text.url).text)
+    else:
+      raise Exception("Need to include data.text.raw or data.text.url in your inputs.")
+
+    st = time.time()
+    max_tokens = 4096
+    # # Method 1
+    # input_ids = self.tokenizer(input_text, return_tensors='pt').input_ids.cuda()
+    # out = self.model.generate(inputs=input_ids, temperature=0.7, max_new_tokens=max_tokens)
+    # out_text = self.tokenizer.decode(out[0], skip_special_tokens=True)
+    # output.data.text.raw = out_text.replace(input_text, '')
+
+    # # Method 2
+    pipe = pipeline(
+        "text-generation",
+        model=self.model,
+        tokenizer=self.tokenizer,
+        max_new_tokens=max_tokens,
+        temperature=0.7,
+        top_p=0.95,
+        repetition_penalty=1.15,
+        return_full_text=False)
+    a = pipe(input_text)
+    output.data.text.raw = a[0]['generated_text']
+    return output
+
+
+if __name__ == '__main__':
+  # Make sure you set these env vars before running the example.
+  # CLARIFAI_PAT
+  # CLARIFAI_USER_ID
+
+  # You need to first create a runner in the Clarifai API and then use the ID here.
+  Llama2Runner(runner_id="sdk-llama2-runner").start()

clarifai/runners/matt_example.py

@@ -0,0 +1,89 @@
+from clarifai_grpc.grpc.api import resources_pb2, service_pb2
+from collections.abc import Iterator
+from google.protobuf import json_format
+
+from clarifai.client.runner import Runner
+
+
+class MyRunner(Runner):
+  """A custom runner that adds "Hello World" to the end of the text and replaces the domain of the
+  image URL as an example.
+  """
+
+  def run_input(self, input: resources_pb2.Input, output_info: resources_pb2.OutputInfo,
+                **kwargs) -> resources_pb2.Output:
+    """This is the method that will be called when the runner is run. It takes in an input and
+    returns an output.
+    """
+
+    output = resources_pb2.Output()
+
+    data = input.data
+
+    # Optional use of output_info
+    params_dict = {}
+    if "params" in output_info:
+      params_dict = output_info["params"]
+
+    if data.text.raw != "":
+      output.data.text.raw = data.text.raw + "Hello World" + params_dict.get(
+          "hello", "") + kwargs.get("extra", "")
+    if data.image.url != "":
+      output.data.text.raw = data.image.url.replace("samples.clarifai.com",
+                                                    "newdomain.com" + params_dict.get("domain",))
+    return output
+
+  def generate(self, request: service_pb2.PostModelOutputsRequest
+              ) -> Iterator[service_pb2.MultiOutputResponse]:
+    """Example yielding a whole batch of streamed stuff back.
+    """
+
+    model = request.model
+    output_info = None
+    if request.model.model_version.id != "":
+      output_info = json_format.MessageToDict(
+          model.model_version.output_info, preserving_proto_field_name=True)
+
+    for i in range(10):  # fake something iterating generating 10 times.
+
+      outputs = []
+      for input in request.inputs:
+        # output = self.run_input(input, output_info, extra=f" {i}")
+        output = resources_pb2.Output()
+        output.data.text.raw = f"Generate Hello World {i}"
+        outputs.append(output)
+      resp = service_pb2.MultiOutputResponse(outputs=outputs,)
+      yield resp
+
+  def stream(self, request: service_pb2.PostModelOutputsRequest
+            ) -> Iterator[service_pb2.MultiOutputResponse]:
+    """Example yielding a whole batch of streamed stuff back.
+    """
+
+    model = request.model
+    output_info = None
+    if request.model.model_version.id != "":
+      output_info = json_format.MessageToDict(
+          model.model_version.output_info, preserving_proto_field_name=True)
+
+    for i in range(10):  # fake something iterating generating 10 times.
+
+      outputs = []
+      for input in request.inputs:
+        # output = self.run_input(input, output_info, extra=f" {i}")
+        output = resources_pb2.Output()
+        out_text = input.data.text.raw + f"Stream Hello World {i}"
+        print(out_text)
+        output.data.text.raw = out_text
+        outputs.append(output)
+      resp = service_pb2.MultiOutputResponse(outputs=outputs,)
+      yield resp
+
+
+if __name__ == '__main__':
+  # Make sure you set these env vars before running the example.
+  # CLARIFAI_PAT
+  # CLARIFAI_USER_ID
+
+  # You need to first create a runner in the Clarifai API and then use the ID here.
+  MyRunner(runner_id="matt-test-runner", base_url="http://q6:32013", num_parallel_polls=1).start()

clarifai/runners/matt_example.py~

@@ -0,0 +1,87 @@
+from clarifai_grpc.grpc.api import resources_pb2, service_pb2
+from collections.abc import Iterator
+from google.protobuf import json_format
+
+from clarifai.client.runner import Runner
+
+
+class MyRunner(Runner):
+  """A custom runner that adds "Hello World" to the end of the text and replaces the domain of the
+  image URL as an example.
+  """
+
+  def run_input(self, input: resources_pb2.Input, output_info: resources_pb2.OutputInfo,
+                **kwargs) -> resources_pb2.Output:
+    """This is the method that will be called when the runner is run. It takes in an input and
+    returns an output.
+    """
+
+    output = resources_pb2.Output()
+
+    data = input.data
+
+    # Optional use of output_info
+    params_dict = {}
+    if "params" in output_info:
+      params_dict = output_info["params"]
+
+    if data.text.raw != "":
+      output.data.text.raw = data.text.raw + "Hello World" + params_dict.get(
+          "hello", "") + kwargs.get("extra", "")
+    if data.image.url != "":
+      output.data.text.raw = data.image.url.replace("samples.clarifai.com",
+                                                    "newdomain.com" + params_dict.get("domain",))
+    return output
+
+  def generate(self, request: service_pb2.PostModelOutputsRequest
+              ) -> Iterator[service_pb2.MultiOutputResponse]:
+    """Example yielding a whole batch of streamed stuff back.
+    """
+
+    model = request.model
+    output_info = None
+    if request.model.model_version.id != "":
+      output_info = json_format.MessageToDict(
+          model.model_version.output_info, preserving_proto_field_name=True)
+
+    for i in range(10):  # fake something iterating generating 10 times.
+
+      outputs = []
+      for input in request.inputs:
+        # output = self.run_input(input, output_info, extra=f" {i}")
+        output = resources_pb2.Output()
+        output.data.text.raw = f"Generate Hello World {i}"
+        outputs.append(output)
+      resp = service_pb2.MultiOutputResponse(outputs=outputs,)
+      yield resp
+
+  def stream(self, request: service_pb2.PostModelOutputsRequest
+            ) -> Iterator[service_pb2.MultiOutputResponse]:
+    """Example yielding a whole batch of streamed stuff back.
+    """
+
+    model = request.model
+    output_info = None
+    if request.model.model_version.id != "":
+      output_info = json_format.MessageToDict(
+          model.model_version.output_info, preserving_proto_field_name=True)
+
+    for i in range(10):  # fake something iterating generating 10 times.
+
+      outputs = []
+      for input in request.inputs:
+        # output = self.run_input(input, output_info, extra=f" {i}")
+        output = resources_pb2.Output()
+        output.data.text.raw = input.data.text.raw + f"Stream Hello World {i}"
+        outputs.append(output)
+      resp = service_pb2.MultiOutputResponse(outputs=outputs,)
+      yield resp
+
+
+if __name__ == '__main__':
+  # Make sure you set these env vars before running the example.
+  # CLARIFAI_PAT
+  # CLARIFAI_USER_ID
+
+  # You need to first create a runner in the Clarifai API and then use the ID here.
+  MyRunner(runner_id="matt-test-runner", base_url="http://q6:32013", num_parallel_polls=1).start()