Skip to main content

Large Language Models (LLMs)

Make predictions using LLMs

Large Language Models (LLMs) are a subset of foundation models that have revolutionized natural language understanding and generation tasks. These models are characterized by their vast size, typically containing hundreds of millions to billions of parameters.

LLMs have learned to perform many kinds of prediction tasks. One of the most notable capabilities of LLMs is text generation. Given a prompt or seed text, they can generate coherent and contextually relevant text that appears as if it were written by a human.

info

The initialization code used in the following examples is outlined in detail on the client installation page.

tip

For the third-party models we've wrapped into our platform, like those provided by OpenAI, Anthropic, Cohere, and others, you can also choose to utilize their API keys as an option—in addition to using the default Clarifai keys. You can learn how to add them here.

Text Completion​

Via URL​

Below is an example of how you would provide a prompt text via a URL and autocomplete sentences or phrases using the llama2-7b-chat large language model.

######################################################################################################
# In this section, we set the user authentication, user and app ID, model details, and the URL of 
# the text we want as a prompt. Change these strings to run your own example.
######################################################################################################

# Your PAT (Personal Access Token) can be found in the Account's Security section
PAT = 'YOUR_PAT_HERE'
# Specify the correct user_id/app_id pairings
# Since you're making inferences outside your app's scope
USER_ID = 'meta'
APP_ID = 'Llama-2'
# Change these to whatever model and text URL you want to use
MODEL_ID = 'llama2-7b-chat'
MODEL_VERSION_ID = 'e52af5d6bc22445aa7a6761f327f7129'
TEXT_FILE_URL = 'https://samples.clarifai.com/negative_sentence_12.txt'

############################################################################
# YOU DO NOT NEED TO CHANGE ANYTHING BELOW THIS LINE TO RUN THIS EXAMPLE
############################################################################

from clarifai_grpc.channel.clarifai_channel import ClarifaiChannel
from clarifai_grpc.grpc.api import resources_pb2, service_pb2, service_pb2_grpc
from clarifai_grpc.grpc.api.status import status_code_pb2

channel = ClarifaiChannel.get_grpc_channel()
stub = service_pb2_grpc.V2Stub(channel)

metadata = (('authorization', 'Key ' + PAT),)

userDataObject = resources_pb2.UserAppIDSet(user_id=USER_ID, app_id=APP_ID)

post_model_outputs_response = stub.PostModelOutputs(
    service_pb2.PostModelOutputsRequest(
        user_app_id=userDataObject,  # The userDataObject is created in the overview and is required when using a PAT
        model_id=MODEL_ID,
        version_id=MODEL_VERSION_ID,  # This is optional. Defaults to the latest model version
        inputs=[
            resources_pb2.Input(
                data=resources_pb2.Data(
                    text=resources_pb2.Text(
                        url=TEXT_FILE_URL
                    )
                )
            )
        ]
    ),
    metadata=metadata
)
if post_model_outputs_response.status.code != status_code_pb2.SUCCESS:
    print(post_model_outputs_response.status)
    raise Exception(f"Post model outputs failed, status: {post_model_outputs_response.status.description}")

# Since we have one input, one output will exist here
output = post_model_outputs_response.outputs[0]

print("Completion:\n")
print(output.data.text.raw)
Text Output Example
Completion:

He doesn't have to commute to work. He can telecommute so he can spend more time with his wife and son. Even though he doesn't get a traditional work-life balance, he still has a quality of life. He may choose to work less hours and make less money, which will allow him to have more time for himself and his family. 

There are many advantages and disadvantages of working from home. Some people may find it difficult to focus and be productive. Others may prefer the noise and energy of an office environment. As the world continues to change and technology evolves, working from home may become more common and socially accepted.  

One of the biggest advantages of working from home is the flexibility and convenience it offers. Instead of having to commute to an office, people can work from the comfort of their own homes. This can save time and money on transportation, and it can also reduce the stress and exhaustion that commuting can cause. Many people find that they are more productive and focused when working from home, where they can avoid office distractions and create a workspace that suits their needs.

One of the biggest disadvantages of working from home is lack of social interaction. Many people find that

Via Local Files​

Below is an example of how you would provide a prompt text via a local text file and autocomplete sentences or phrases using the llama2-7b-chat large language model.

######################################################################################################
# In this section, we set the user authentication, user and app ID, model details, and the location of 
# the text we want as a prompt. Change these strings to run your own example.
######################################################################################################

# Your PAT (Personal Access Token) can be found in the Account's Security section
PAT = 'YOUR_PAT_HERE'
# Specify the correct user_id/app_id pairings
# Since you're making inferences outside your app's scope
USER_ID = 'meta'
APP_ID = 'Llama-2'
# Change these to whatever model and text you want to use
MODEL_ID = 'llama2-7b-chat'
MODEL_VERSION_ID = 'e52af5d6bc22445aa7a6761f327f7129'
TEXT_FILE_LOCATION = 'YOUR_TEXT_FILE_LOCATION_HERE'

############################################################################
# YOU DO NOT NEED TO CHANGE ANYTHING BELOW THIS LINE TO RUN THIS EXAMPLE
############################################################################

from clarifai_grpc.channel.clarifai_channel import ClarifaiChannel
from clarifai_grpc.grpc.api import resources_pb2, service_pb2, service_pb2_grpc
from clarifai_grpc.grpc.api.status import status_code_pb2

channel = ClarifaiChannel.get_grpc_channel()
stub = service_pb2_grpc.V2Stub(channel)

metadata = (('authorization', 'Key ' + PAT),)

userDataObject = resources_pb2.UserAppIDSet(user_id=USER_ID, app_id=APP_ID)

with open(TEXT_FILE_LOCATION, "rb") as f:
    file_bytes = f.read()

post_model_outputs_response = stub.PostModelOutputs(
    service_pb2.PostModelOutputsRequest(
        user_app_id=userDataObject,  # The userDataObject is created in the overview and is required when using a PAT
        model_id=MODEL_ID,
        version_id=MODEL_VERSION_ID,  # This is optional. Defaults to the latest model version
        inputs=[
            resources_pb2.Input(
                data=resources_pb2.Data(
                    text=resources_pb2.Text(
                        raw=file_bytes
                    )
                )
            )
        ]
    ),
    metadata=metadata
)
if post_model_outputs_response.status.code != status_code_pb2.SUCCESS:
    print(post_model_outputs_response.status)
    raise Exception(f"Post model outputs failed, status: {post_model_outputs_response.status.description}")

# Since we have one input, one output will exist here
output = post_model_outputs_response.outputs[0]

print("Completion:\n")
print(output.data.text.raw)

Via Raw Text​

Below is an example of how you would provide a raw text prompt and autocomplete sentences or phrases using the llama2-7b-chat large language model.

######################################################################################################
# In this section, we set the user authentication, user and app ID, model details, and the raw
# text we want as a prompt. Change these strings to run your own example.
######################################################################################################

# Your PAT (Personal Access Token) can be found in the Account's Security section
PAT = 'YOUR_PAT_HERE'
# Specify the correct user_id/app_id pairings
# Since you're making inferences outside your app's scope
USER_ID = 'meta'
APP_ID = 'Llama-2'
# Change these to whatever model and text you want to use
MODEL_ID = 'llama2-7b-chat'
MODEL_VERSION_ID = 'e52af5d6bc22445aa7a6761f327f7129'
RAW_TEXT = 'I love your product very much'

############################################################################
# YOU DO NOT NEED TO CHANGE ANYTHING BELOW THIS LINE TO RUN THIS EXAMPLE
############################################################################

from clarifai_grpc.channel.clarifai_channel import ClarifaiChannel
from clarifai_grpc.grpc.api import resources_pb2, service_pb2, service_pb2_grpc
from clarifai_grpc.grpc.api.status import status_code_pb2

channel = ClarifaiChannel.get_grpc_channel()
stub = service_pb2_grpc.V2Stub(channel)

metadata = (('authorization', 'Key ' + PAT),)

userDataObject = resources_pb2.UserAppIDSet(user_id=USER_ID, app_id=APP_ID)

post_model_outputs_response = stub.PostModelOutputs(
    service_pb2.PostModelOutputsRequest(
        user_app_id=userDataObject,  # The userDataObject is created in the overview and is required when using a PAT
        model_id=MODEL_ID,
        version_id=MODEL_VERSION_ID,  # This is optional. Defaults to the latest model version
        inputs=[
            resources_pb2.Input(
                data=resources_pb2.Data(
                    text=resources_pb2.Text(
                        raw=RAW_TEXT
                    )
                )
            )
        ]
    ),
    metadata=metadata
)
if post_model_outputs_response.status.code != status_code_pb2.SUCCESS:
    print(post_model_outputs_response.status)
    raise Exception(f"Post model outputs failed, status: {post_model_outputs_response.status.description}")

# Since we have one input, one output will exist here
output = post_model_outputs_response.outputs[0]

print("Completion:\n")
print(output.data.text.raw)

Use Hyperparameters to Customize LLMs​

You can use hyperparameters to fine-tune and customize the behavior of LLMs. This allows you to gain precise control over the prediction output of LLMs, shaping their responses to suit your unique needs.

Here are some parameters we support:

  • Temperature—It affects the randomness of the model's output. A higher temperature (e.g., 0.8) will make the output more random and creative, while a lower temperature (e.g., 0.2) will make it more deterministic and focused.

  • Max Tokens—It allow you to limit the length of the generated text. You can set a maximum number of tokens to prevent the output from becoming too long or to fit within specific constraints.

  • Top K—It controls the diversity of the output. It limits the vocabulary to the top_k most likely tokens at each step. A lower value of K (e.g., 10) will make the output more focused, while a higher value (e.g., 50) will make it more diverse.

tip

Most of our models now have new versions that support inference hyperparameters like temperature, top_k, etc. This example illustrates how you can configure them.

#####################################################################################################################
# In this section, we set the user authentication, user and app ID, model details, raw text we want as a prompt,
# and the parameters. Change these values to run your own example.
#####################################################################################################################

# Your PAT (Personal Access Token) can be found in the Account's Security section
PAT = 'YOUR_PAT_HERE'
# Specify the correct user_id/app_id pairings
# Since you're making inferences outside your app's scope
USER_ID = 'meta'
APP_ID = 'Llama-2'
# Change these to whatever model and text you want to use
MODEL_ID = 'llama2-7b-chat'
MODEL_VERSION_ID = 'e52af5d6bc22445aa7a6761f327f7129'
RAW_TEXT = 'I love your product very much'

############################################################################
# YOU DO NOT NEED TO CHANGE ANYTHING BELOW THIS LINE TO RUN THIS EXAMPLE
############################################################################

from clarifai_grpc.channel.clarifai_channel import ClarifaiChannel
from clarifai_grpc.grpc.api import resources_pb2, service_pb2, service_pb2_grpc
from clarifai_grpc.grpc.api.status import status_code_pb2
from google.protobuf.struct_pb2 import Struct

channel = ClarifaiChannel.get_grpc_channel()
stub = service_pb2_grpc.V2Stub(channel)

metadata = (('authorization', 'Key ' + PAT),)

userDataObject = resources_pb2.UserAppIDSet(user_id=USER_ID, app_id=APP_ID)

params = Struct()
params.update({
    "temperature": 0.5,
    "max_tokens": 2048,
    "top_k": 0.95
})

post_model_outputs_response = stub.PostModelOutputs(
    service_pb2.PostModelOutputsRequest(
        user_app_id=userDataObject,  # The userDataObject is created in the overview and is required when using a PAT
        model_id=MODEL_ID,
        version_id=MODEL_VERSION_ID,  # This is optional. Defaults to the latest model version
        inputs=[
            resources_pb2.Input(
                data=resources_pb2.Data(
                    text=resources_pb2.Text(
                        raw=RAW_TEXT
                    )
                )
            )
        ],
        model=resources_pb2.Model(
            model_version=resources_pb2.ModelVersion(
                output_info=resources_pb2.OutputInfo(
                    params=params
                )
            )
        )

    ),
    metadata=metadata
)
if post_model_outputs_response.status.code != status_code_pb2.SUCCESS:
    print(post_model_outputs_response.status)
    raise Exception(f"Post model outputs failed, status: {post_model_outputs_response.status.description}")

# Since we have one input, one output will exist here
output = post_model_outputs_response.outputs[0]

print("Completion:\n")
print(output.data.text.raw)
© 2025 Clarifai, Inc. All rights reserved