AlpinDale
|
00503b9fc1
feat: non-uniform quantization via `compressed-tensors` for llama
|
4 ay önce |
AlpinDale
|
ee2c5d34da
feat: add fp8 channel-wise weight quantization support
|
4 ay önce |
AlpinDale
|
98cb1c4cd1
feat: support fp8 via `llm-compressor`
|
4 ay önce |
AlpinDale
|
e2dbe5f05c
feat: add sparse marlin for compressed tensors
|
5 ay önce |
AlpinDale
|
aba03b4756
feat: dynamic per-token activation quantization
|
5 ay önce |