2 ms·
"Our implementation is based on HuggingFace transformers where we register a new model opt-llama that supports the Layer-Condensed KV Cache." Not sure what thi
by jsemrau 2y ago
"Our implementation is based on HuggingFace transformers where we register a new model opt-llama that supports the Layer-Condensed KV Cache."
Not sure what this means? Would this work for a Mistral model as well?