GitLab

Signed-off-by: Pavani Majety <pmajety@nvidia.com>
Co-authored-by: mgoin <michael@neuralmagic.com>

Signed-off-by: mgoin <michael@neuralmagic.com>

Signed-off-by: kewang-xlnx <kewang@xilinx.com>
Signed-off-by: kewang2 <kewang2@amd.com>
Co-authored-by: kewang2 <kewang2@amd.com>
Co-authored-by: Michael Goin <michael@neuralmagic.com>

Signed-off-by: youkaichao <youkaichao@gmail.com>

Signed-off-by: Isotr0py <2037008807@qq.com>

Signed-off-by: Roger Wang <ywang@roblox.com>

Signed-off-by: youkaichao <youkaichao@gmail.com>

Signed-off-by: youkaichao <youkaichao@gmail.com>

Signed-off-by: Joe Runde <Joseph.Runde@ibm.com>

Signed-off-by: Aaron Pham <contact@aarnphm.xyz>

Signed-off-by: Prashant Gupta <prashantgupta@us.ibm.com>

Signed-off-by: Muralidhar Andoorveedu <muralidhar.andoorveedu@centml.ai>
Signed-off-by: Murali Andoorveedu <muralidhar.andoorveedu@centml.ai>
Co-authored-by: DarkLight1337 <tlleungac@connect.ust.hk>

Co-authored-by: ElizaWszola <eliza@neuralmagic.com>

Co-authored-by: ElizaWszola <eliza@neuralmagic.com>

Co-authored-by: Robert Shaw <rshaw@neuralmagic.com>
Co-authored-by: ZX <zx@lbx.dev>

Co-authored-by: Robert Shaw <rshaw@neuralmagic>
Co-authored-by: Michael Goin <michael@neuralmagic.com>

Signed-off-by: Muralidhar Andoorveedu <muralidhar.andoorveedu@centml.ai>

Co-authored-by: rsnm2 <rshaw@neuralmagic.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-neuralmagic@users.noreply.github.com>

The 2nd PR for #4532.

This PR supports loading FP8 kv-cache scaling factors from a FP8 checkpoint (with .kv_scale parameter).

[Kernel] Support MoE Fp8 Checkpoints for Mixtral (Static Weights with Dynamic/Static Activations) (#4527)

Follow on to #4332 to enable FP8 checkpoint loading for Mixtral and supersedes #4436.

This PR enables the following checkpoint loading features for Mixtral:

Supports loading fp8 checkpoints for Mixtral, such as this "nm-testing/Mixtral-8x7B-Instruct-v0.1-FP8" test model
Supports static or dynamic activation quantization with static weight quantization (all per tensor)
Supports different scales for each expert weight
Supports Fp8 in QKV layer
Notes:

The Expert Gate/Router always runs at half / full precision for now.
If there are different weight scales between QKV layer (for separate QKV weights), they are re-quantized using layer.weight_scale.max() so we can have a single gemm for performance.

Remove the device="cuda" declarations in mixtral as promised in #4343