README.md

metadata

base_model: databricks/dbrx-instruct

dbrx_moe_fp8_test

• Introduction

  This model was created by applying Quark with calibration samples from Pile dataset.

• Quantization Stragegy

  • Quantized Layers: All linear layers excluding "lm_head" and "router.layer"
  • Weight: FP8 symmetric per-tensor
  • Activation: FP8 symmetric per-tensor
  • KV Cache: FP8 symmetric per-tensor

• Quick Start

1. Download and install Quark
2. Run the quantization script in the example folder using the following command line:

export MODEL_DIR = [local model checkpoint folder] or databricks/dbrx-instruct
# single GPU
python3 quantize_quark.py \ 
        --model_dir $MODEL_DIR \
        --output_dir dbrx_moe_fp8_test \                           
        --quant_scheme w_fp8_a_fp8 \
        --kv_cache_dtype fp8 \
        --num_calib_data 128 \
        --model_export quark_safetensors \
        --no_weight_matrix_merge
# If model size is too large for single GPU, please use multi GPU instead.
python3 quantize_quark.py
        --model_dir $MODEL_DIR \
        --output_dir dbrx_moe_fp8_test\
        --quant_scheme w_fp8_a_fp8 \
        --kv_cache_dtype fp8 \
        --num_calib_data 128 \
        --multi_gpu \
        --model_export quark_safetensors \
        --no_weight_matrix_merge

Deployment

Quark has its own export format and allows FP8 quantized models to be efficiently deployed using the vLLM backend(vLLM-compatible). In the dbrx-instruct model, "transformer.blocks.*.ffn.experts" modules can be divided into experts-num mlps, and if the shape of the weight of w1 in one of the mlps is [dim1, dim2], then the shape of “transformer.blocks.*.ffn.experts.mlp.w1.weight“ in the exported safetensors file is [dim1*experts-num, dim2]. The shapes of "transformer.blocks.*.ffn.experts.mlp.w1.weight_scale" and "transformer.blocks.*.ffn.experts.mlp.w1.input_scale" are [dim1]. Similarly, this also applies to the w2 and v1 of "transformer.blocks.*.ffn.experts.mlp".

Evaluation

Quark currently uses perplexity(PPL) as the evaluation metric for accuracy loss before and after quantization.The specific PPL algorithm can be referenced in the quantize_quark.py. The quantization evaluation results are conducted in pseudo-quantization mode, which may slightly differ from the actual quantized inference accuracy. These results are provided for reference only.

Evaluation scores

Benchmark	dbrx-instruct	dbrx_moe_fp8_test(this model)
Perplexity-wikitext2	4.2275	4.3033

License

Modifications copyright(c) 2024 Advanced Micro Devices,Inc. All rights reserved.

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.