When clipping https://thinkingmachines.ai/blog/defeating-nondeterminism-in-llm-inference/ it fails to capture the full length of the document. It is truncating the document by stopping before the "Batch-invariant attention" section.
When clipping https://thinkingmachines.ai/blog/defeating-nondeterminism-in-llm-inference/ it fails to capture the full length of the document. It is truncating the document by stopping before the "Batch-invariant attention" section.