possible problem in description

#3
by kuliev-vitaly - opened

Comparison of models:
Meta-Llama-3.1-70B-Instruct-quantized.w8a8
Meta-Llama-3.1-70B-Instruct-quantized.w8a16

Description contain table with accuracy. Values are very different. Even for base model Meta-Llama-3.1-70B-Instruct - it should be the same. It looks some error in data o testing.

The benchmarks are identical to what is described in Meta-Llama-3.1-8B-Instruct-quantized.w8a16

Meta-Llama-3.1-70B-Instruct-quantized.w8a16 benchmarks need to be corrected.

Neural Magic org

Corrected. Thanks for catching that.

alexmarques changed discussion status to closed

Sign up or log in to comment