ai-agents-replicate-model-vers-user-attempts-to-run-a-public-replicate-model-e-g--7120ca9f

THEN Use the replicate Python library directly instead of the LangChain wrapper for such models. Query the model's latest version from the Replicate API (e.g., GET /v1/models/meta/meta-llama-3-70b-instruct) and specify it, though even the correct version may still fail due to LangChain's unhandled version handling. Monitor LangChain for updates that make version optional. As a temporary workaround, call replicate.run() directly with input parameters as shown in the issue.

import replicate

output = replicate.run(
    "meta/meta-llama-3-70b-instruct",
    input={
        "top_p": 0.9,
        "prompt": prompt,
        "max_tokens": 512,
        "temperature": 0.6,
        "prompt_template": "<|begin_of_text|>...{prompt}...",
        "presence_penalty": 1.15,
        "frequency_penalty": 0.2
    }
)