GitLab

[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.

Co-authored-by: LiuXiaoxuanPKU <llilyliupku@gmail.com>

Co-authored-by: simon-mo <simon.mo@hey.com>

Co-authored-by: SangBin Cho <rkooo567@gmail.com>

Co-authored-by: Zhuohan Li <zhuohan123@gmail.com>

Co-authored-by: Cade Daniel <edacih@gmail.com>

[Bug fix][Core] assert num_new_tokens == 1 fails when SamplingParams.n is not 1 and max_tokens is large & Add tests for preemption (#4451)

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

This PR updates the tuning script for the fused_moe kernel to support FP8 and also adds configurations for TP4. Note that for the configuration I removed num_warps and num_stages for small batch sizes since that improved performance and brought the benchmarks on par with the numbers before in that regime to make sure this is a strict improvement over the status quo.

All the numbers below are for mistralai/Mixtral-8x7B-Instruct-v0.1, 1000 input and 50 output tokens.

Before this PR (with static activation scaling):

qps = 1: 9.8 ms ITL, 0.49s e2e latency
qps = 2: 9.7 ms ITL, 0.49s e2e latency 
qps = 4: 10.1 ms ITL, 0.52s e2e latency
qps = 6: 11.9 ms ITL, 0.59s e2e latency
qps = 8: 14.0 ms ITL, 0.70s e2e latency
qps = 10: 15.7 ms ITL, 0.79s e2e latency

After this PR (with static activation scaling):

qps = 1: 9.8 ms ITL, 0.49s e2e latency
qps = 2: 9.7 ms ITL, 0.49s e2e latency
qps = 4: 10.2 ms ITL, 0.53s e2e latency
qps = 6: 11.9 ms ITL, 0.59s e2e latency
qps = 8: 11.9 ms ITL, 0.59s e2e latency
qps = 10: 12.1 ms ITL, 0.61s e2e latency

Co-authored-by: Lei Wen <wenlei03@qiyi.com>
Co-authored-by: Sage Moore <sagemoore@utexas.edu>

Co-authored-by: Lei Wen <wenlei03@qiyi.com>

Signed-off-by: Travis Johnson <tsjohnso@us.ibm.com>

[Bugfix] Fix the fp8 kv_cache check error that occurs when failing to obtain the CUDA version. (#4173)
Signed-off-by: AnyISalIn <anyisalin@gmail.com>