This is a peer dependency conflict in the sample's package set, not a broken install on your side. First run npm install from the bedrock-agent-cdk root directory before cdk bootstrap. If npm install fails with ERESOLVE (aws-cdk-lib@2.87.0 vs @aws-cdk/aws-lambda-python-alpha@2.108.0-alpha.0 wanting …
Set the dtype explicitly to torch.float16 when loading the PEFT adapter: AutoPeftModelForCausalLM.from_pretrained(training_args.output_dir, low_cpu_mem_usage=True, torch_dtype=torch.float16). The sample passed the wrong dtype when loading the adapter; setting torch.float16 fixed it. A maintainer con…
Recreate the OpenSearch Serverless vector index with "engine": "faiss" instead of "nmslib" in the storage configuration, then retry the CreateKnowledgeBase call. FAISS is the only accepted engine type now. The amazon-bedrock-samples repo was fixed the same way. Source: https://github.com/aws-samples…
This is a peer dependency conflict in the sample's package set, not a broken install on your side. First run npm install from the bedrock-agent-cdk root directory before cdk bootstrap. If npm install fails with ERESOLVE (aws-cdk-lib@2.87.0 vs @aws-cdk/aws-lambda-python-alpha@2.108.0-alpha.0 wanting …
Set the dtype explicitly to torch.float16 when loading the PEFT adapter: AutoPeftModelForCausalLM.from_pretrained(training_args.output_dir, low_cpu_mem_usage=True, torch_dtype=torch.float16). The sample passed the wrong dtype when loading the adapter; setting torch.float16 fixed it. A maintainer con…
Recreate the OpenSearch Serverless vector index with "engine": "faiss" instead of "nmslib" in the storage configuration, then retry the CreateKnowledgeBase call. FAISS is the only accepted engine type now. The amazon-bedrock-samples repo was fixed the same way. Source: https://github.com/aws-samples…