speculative : fix KV cache management - SemanticDiff

Commit

2 years ago

speculative : fix KV cache management

References

#3228 - llama : custom attention mask + parallel decoding + no context swaps

Author

ggerganov

ggerganov

Parents

Loading