Mixtral-8x7B-Instruct-v0.1-WMXFP4FP8-AMXFP4FP8-AMP-KVFP8
Introduction
This model was created by applying Quark with calibration samples from Pile dataset.
Quantization Stragegy
- Quantized Layers: All linear layers excluding "lm_head", "*.gate"
- Weight: Auto Mixed Precision quantized by Quark, each weight has either quantization scheme in candidates of
- FP8 symmetric per-tensor
- OCP Microscaling (MX) FP4
- Activation: Auto Mixed Precision quantized by Quark, each activation input has the same quantization scheme with weight, i.e., in candidates of
- FP8 symmetric per-tensor
- OCP Microscaling (MX) FP4
- KV Cache: FP8 symmetric per-tensor
Quick Start
- Download and install Quark
- [TODO] We will provide example script(s) to run auto mixed precision (AMP) quantizations later.
Deployment
The Quark quantized Auto Mixed Precision (AMP) models are now supported to be easily deployed in vLLM backend (vLLM-compatible).
Evaluation
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
| Quant scheme | arc challenge (โ) (acc) |
gsm8k (โ) (strict-match) |
mmlu (โ) (acc) |
wikitext (โ) (word_perplexity) |
winogrande (โ) (acc) |
|||||
|---|---|---|---|---|---|---|---|---|---|---|
| absolute value | recovery rate | absolute value | recovery rate | absolute value | recovery rate | absolute value | recovery rate | absolute value | recovery rate | |
| FP16 | 0.6271 | 100.0% | 0.6384 | 100.0% | 0.6893 | 100.0% | 5.7259 | 100.0% | 0.7664 | 100.0% |
| FP8 | 0.6220 | 99.2% | 0.6384 | 100.0% | 0.6816 | 98.9% | 5.8092 | 98.6% | 0.7830 | 102.2% |
| AMP | 0.6237 | 99.5% | 0.6133 | 96.1% | 0.6765 | 98.1% | 6.2020 | 92.3% | 0.7569 | 98.8% |
| MXFP4 | 0.5922 | 94.4% | 0.4845 | 75.9% | 0.6364 | 92.3% | 7.0935 | 80.7% | 0.7474 | 97.5% |
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.
- Downloads last month
- 13
Model tree for amd/Mixtral-8x7B-Instruct-v0.1-WMXFP4FP8-AMXFP4FP8-AMP-KVFP8
Base model
mistralai/Mixtral-8x7B-v0.1