Upgrade to 1m context

#4
by freegheist - opened
QuantTrio org
edited Sep 2

Correction: We need to verify whether config_1m.json can run properly on the quantized model.

JunHowie changed discussion status to closed
QuantTrio org

We will conduct testing and provide an updated config.json for 1M context in a later update.

JunHowie changed discussion status to open

We will conduct testing and provide an updated config.json for 1M context in a later update.

How is the testing going I was testing by directly changing the config.json with the addition of the config_1m.json. it is really slow.

Correction: We need to verify whether config_1m.json can run properly on the quantized model.

Any news?

QuantTrio org

We tested the 1M-context configuration, but the results were unsatisfactory; therefore, we currently have no plans to roll out this update.

We tested the 1M-context configuration, but the results were unsatisfactory; therefore, we currently have no plans to roll out this update.

Thank you for the quick response. Is this for all the quantized models? I wanted to try them to in the AWQ version. Also can you more specific about what were the results?

Sign up or log in to comment