SHARK
SHARK copied to clipboard
llama2 - FP16 - IR generation
- fix combine mlir for llama2 for fp16 IR generation by keeping deduplicating constant variable names using appropriate suffix for first llama and second llama
- needs to be tested for IR generation with int4 to see it doesn't break the generation flow for quantised IRs
Converted to draft till sanity check is done for int4 IR generation.