--- title: Qwen2.5-14B-Instruct-FP8-A16 canonical_url: "https://www.modelscope.cn/models/okwinds/Qwen2.5-14B-Instruct-FP8-A16" md_url: "https://www.modelscope.cn/models/okwinds/Qwen2.5-14B-Instruct-FP8-A16.md" repository: okwinds/Qwen2.5-14B-Instruct-FP8-A16 chinese_name: "Qwen2.5-14B-Instruct-FP8-A16 量化高精校准" last_updated: 2025-01-05 license: qwen pipeline_tag: text-generation tasks: - text-generation model_type: - qwen2 architectures: - Qwen2ForCausalLM parameters: 14.8B tensor_type: - F8_E4M3 - BF16 - F32 library_name: - safetensors - pytorch frameworks: - pytorch language: - en - zh inference_backends: - "deploy_task text/emb" - "lmdeploy 0.9.1" - "lmdeploy_turbomind 0.9.1" - "sglang 0.5.2" - "vllm 0.9.2" downloads: 381 stars: 5 tags: - Instruct - Chat --- # Qwen2.5-14B-Instruct-FP8-A16 > Qwen2.5-14B-Instruct-FP8-A16 - okwinds 在 ModelScope 开源的模型。通义千问2.5-14B-Instruct-FP8-A16量化高精校准 okwinds/Qwen2.5-14B-Instruct-FP8-A16 是 ModelScope 魔搭社区上的 14.8B 参数text-generation模型，采用 qwen 许可，可用 deploy_task text/emb、lmdeploy 0.9.1、lmdeploy_turbomind 0.9.1 部署。 - **Repository**: okwinds/Qwen2.5-14B-Instruct-FP8-A16 - **License**: qwen - **Tasks**: text-generation - **Parameters**: 14.8B - **Inference backends**: deploy_task text/emb, lmdeploy 0.9.1, lmdeploy_turbomind 0.9.1, sglang 0.5.2, vllm 0.9.2 - **Tags**: Instruct, Chat - **Downloads**: 381 - **Stars**: 5 - **Last updated**: 2025-01-05 Source: https://www.modelscope.cn/models/okwinds/Qwen2.5-14B-Instruct-FP8-A16 --- # 通义千问2.5-14B-Instruct-FP8-A16量化高精校准 原模型 [qwen/Qwen2.5-14B-Instruct](https://www.modelscope.cn/models/qwen/Qwen2.5-14B-Instruct) ```python # download model from modelscope import snapshot_download model_dir = snapshot_download('okwinds/Qwen2.5-14B-Instruct-FP8-A16') ``` ## Qwen2.5 同系列量化模型 | **Model** | GPTQ Compression & FP8 | GGUF | **MLX** (macOS) | | ------------ |----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| -------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | *3B params* | - | [3B-GGUF-V3-LOT](https://www.modelscope.cn/models/okwinds/Qwen2.5-3B-Instruct-GGUF-V3-LOT) | [3B-MLX-4bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-3B-Instruct-MLX-4bit)
[3B-MLX-8bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-3B-Instruct-MLX-8bit) | | *7B params* | [7B-Int4-W4A16](https://www.modelscope.cn/models/okwinds/Qwen2.5-7B-Instruct-Int4-W4A16)
[7B-Int8-W8A16](https://www.modelscope.cn/models/okwinds/Qwen2.5-7B-Instruct-Int8-W8A16)
[7B-FP8-A16](https://www.modelscope.cn/models/okwinds/Qwen2.5-7B-Instruct-FP8-A16)
[Coder-7B-Instruct-FP8](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-7B-Instruct-FP8) | [7B-GGUF-V3-LOT](https://www.modelscope.cn/models/okwinds/Qwen2.5-7B-Instruct-GGUF-V3-LOT) | [7B-MLX-4bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-7B-Instruct-MLX-4bit)
[7B-MLX-8bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-7B-Instruct-MLX-8bit) | | *Coder 7B params* | [Coder-7B-Int4-W4A16](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-7B-Instruct-Int4-W4A16)
[Coder-7B-Int8-W8A16](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-7B-Instruct-Int8-W8A16)
[Coder-7B-FP8-A16](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-7B-Instruct-FP8-A16) | [Coder-7B-GGUF-V3-LOT](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-7B-Instruct-GGUF-V3-LOT) | [Coder-7B-MLX-4bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-7B-Instruct-MLX-4bit)
[Coder-7B-MLX-8bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-7B-Instruct-MLX-8bit) | | *14B params* | [14B-Int4-W4A16](https://www.modelscope.cn/models/okwinds/Qwen2.5-14B-Instruct-Int4-W4A16)
[14B-Int8-W8A16](https://www.modelscope.cn/models/okwinds/Qwen2.5-14B-Instruct-Int8-W8A16)
[14B-FP8-A16](https://www.modelscope.cn/models/okwinds/Qwen2.5-14B-Instruct-FP8-A16) | [14B-GGUF-V3-LOT](https://www.modelscope.cn/models/okwinds/Qwen2.5-14B-Instruct-GGUF-V3-LOT) | [14B-MLX-4bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-14B-Instruct-MLX-4bit)
[14B-MLX-8bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-14B-Instruct-MLX-8bit) | | *Coder 14B params* | [Coder-14B-Int4-W4A16](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-14B-Instruct-Int4-W4A16)
[Coder-14B-Int8-W8A16](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-14B-Instruct-Int8-W8A16) | [Coder-14B-GGUF-V3-LOT](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-14B-Instruct-GGUF-V3-LOT) | [Coder-14B-MLX-4bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-14B-Instruct-MLX-4bit)
[Coder-14B-MLX-8bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-14B-Instruct-MLX-8bit) | | *32B params* | [32B-Int4-W4A16](https://www.modelscope.cn/models/okwinds/Qwen2.5-32B-Instruct-Int4-W4A16)
[32B-Int8-W8A16](https://www.modelscope.cn/models/okwinds/Qwen2.5-32B-Instruct-Int8-W8A16)
[32B-FP8-A16](https://www.modelscope.cn/models/okwinds/Qwen2.5-32B-Instruct-FP8-A16) | [32B-GGUF-V3-LOT](https://www.modelscope.cn/models/okwinds/Qwen2.5-32B-Instruct-GGUF-V3-LOT) | [32B-MLX-2bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-32B-Instruct-MLX-2bit)
[32B-MLX-4bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-32B-Instruct-MLX-4bit)
[32B-MLX-8bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-32B-Instruct-MLX-8bit) | | *Coder 32B params* | [Coder-32B-Int4-W4A16](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-32B-Instruct-Int4-W4A16)
[Coder-32B-FP8](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-32B-Instruct-FP8) | [Coder-32B-GGUF-V3-LOT](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-32B-Instruct-GGUF-V3-LOT) | [Coder-32B-MLX-4bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-32B-Instruct-MLX-4bit)
[Coder-32B-MLX-8bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-Coder-32B-Instruct-MLX-8bit) | | *QwQ-Preview 32B params* | [QwQ-32B-Preview-Int4-W4A16](https://www.modelscope.cn/models/okwinds/QwQ-32B-Preview-Int4-W4A16)
[QwQ-32B-Preview-FP8](https://www.modelscope.cn/models/okwinds/QwQ-32B-Preview-FP8) | [QwQ-32B-GGUF-V3-LOT](https://www.modelscope.cn/models/okwinds/QwQ-32B-Preview-GGUF-V3-LOT) | [QwQ-32B-MLX-4bit](https://www.modelscope.cn/models/okwinds/QwQ-32B-Preview-MLX-4bit)
[QwQ-32B-MLX-8bit](https://www.modelscope.cn/models/okwinds/QwQ-32B-Preview-MLX-8bit) | | *72B params* | - | [72B-GGUF-V3-LOT](https://www.modelscope.cn/models/okwinds/Qwen2.5-72B-Instruct-GGUF-V3-LOT) | [72B-MLX-2bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-72B-Instruct-MLX-2bit)
[72B-MLX-4bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-72B-Instruct-MLX-4bit)
[72B-MLX-8bit](https://www.modelscope.cn/models/okwinds/Qwen2.5-72B-Instruct-MLX-8bit) | ## 模型概述 > 本模型已升级，请使用精度更高的最新版本，原版本归档至 tag v1.0.0。
更多精度详情，请关注 Benchmark 部分数据。 通义千问2.5-14B-Instruct-FP8-A16 是一个基于 通义千问2.5-14B-Instruct 的 FP8 量化并校准的模型。 - **模型名称:** Qwen2.5-14B-Instruct-FP8-A16 - **模型架构:** Qwen2.5 - **权重量化:** FP8 ### 量化模型 该模型通过将 [Qwen2.5-14B-Instruct](https://www.modelscope.cn/models/qwen/Qwen2.5-14B-Instruct) 的权重量化为 FP8 数据类型而实现。 量化过程将每个参数从 16bit 减少到 8bit，将模型占用磁盘空间大小，以及推理时加载模型需要的GPU显存空间，减少到了大约为原模型的1/2。 FP8（Floating Point 8-bit）是一种8位浮点数数据格式，由Nvidia、Arm、Intel联合推出，旨在加速深度学习的训练和推理过程。 本模型的量化，采用 F8_E4M3 数据类型仅对 transformer 中的 Linear 层权重进行 static 量化，其他层均保持为原模型数据类型，采用混合精度的计算方式，尽可能减少模型量化后所带来的精度损失。 并且，在量化过程中，进行了数据校准，以进一步提升模型生成精度。 > 注意：本模型需要 compute capability > 8.0（Ampere、Ada Lovelace、Hopper 架构）的 Nvidia GPU 来支持 FP8 及**混合精度**计算。 ## 部署推理 #### 推荐使用 vLLM （ version ≥ 0.6.1.post2 ） Openai api 兼容模式 ```bash >>> vllm serve "/home/gavin/llm/Qwen2.5-14B-Instruct-FP8-A16" --host 0.0.0.0 --port 8000 --quantization fp8 --gpu-memory-utilization 0.9 --served-model-name "Qwen2.5-14B-Instruct-FP8-A16" ``` ## Benchmark 本模型的最新版本对精度进行了更广泛的校准。尤其对推理，CoT，指令遵循，Agent等能力做了强化提升。 | Benchmark | Qwen2.5-14B-Instruct | FP8-A16-v1.0.0 | Recovery-v1.0.0 | _FP8-A16-master(New)_ | Recovery-master(New) | | ------------------- | -------------------- | -------------- | --------------- | -------------------------------------------------------------------- | ------------------------------------------------------------------- | | **Overall** | 72.59 | 71.19 | 98.08% | 72.26 | 99.55% | | Exam | 82.67 | 82.02 | 99.22% | 82.22 | 99.46% | | Language | 57.40 | 56.66 | 98.71% | 55.66 | 96.97% | | Knowledge | 59.94 | 59.28 | 98.90% | 60.95 | 101.69% | | Understanding | 72.66 | 72.42 | 99.67% | 72.81 | 100.21% | | Coding | 80.57 | 78.66 | 97.62% | 79.46 | 98.62% | | Reasoning | 71.88 | 71.30 | 99.20% | 72.04 | 100.22% | | Instruct Follow | 82.99 | 78.00 | 93.99% | 82.67 | 99.61% | | **Exam** | - | - | - | - | - | | C-Eval | 81.38 | 80.81 | 99.30% | 81.23 | 99.82% | | MMLU | 79.44 | 78.93 | 99.36% | 79.06 | 99.52% | | CMMLU | 82.75 | 82.00 | 99.09% | 82.57 | 99.78% | | GaokaoBench | 87.09 | 86.33 | 99.13% | 86.04 | 98.79% | | **Language** | - | - | - | - | - | | WiC | 56.74 | 56.90 | 100.28% | 56.90 | 100.28% | | WSC | 81.73 | 78.85 | 96.48% | 75.00 | 91.77% | | Afqmc-DEV | 70.71 | 70.37 | 99.52% | 70.62 | 99.87% | | Tydiqa-Goldp | 20.42 | 20.51 | 100.44% | 20.11 | 98.48% | | **Knowledge** | - | - | - | - | - | | BoolQ | 84.04 | 84.43 | 100.46% | 84.31 | 100.32% | | TriviaQA | 56.89 | 56.04 | 98.51% | 56.11 | 98.63% | | GPQA Diamond | 38.89 | 37.37 | 96.09% | 42.42 | 109.08% | | **Understanding** | - | - | - | - | - | | C3 | 93.92 | 93.37 | 99.41% | 94.25 | 100.35% | | Race-Middle | 93.11 | 92.90 | 99.77% | 93.31 | 100.21% | | Race-High | 89.79 | 89.59 | 99.78% | 89.88 | 100.10% | | LCSTS | 13.82 | 13.81 | 99.93% | 13.81 | 99.93% | | **Coding** | - | - | - | - | - | | Human Eval | 82.93 | 79.88 | 96.32% | 81.10 | 97.79% | | Sanitized MBPP | 78.21 | 77.43 | 99.00% | 77.82 | 99.50% | | **Reasoning** | - | - | - | - | - | | Ocnli | 68.31 | 67.15 | 98.30% | 66.95 | 98.01% | | COPA | 98.00 | 98.00 | 100.00% | 98.00 | 100.00% | | MATH | 51.82 | 50.84 | 98.11% | 52.22 | 100.77% | | GSM8K | 89.84 | 89.54 | 99.67% | 90.60 | 100.85% | | Winogrande | 78.85 | 78.30 | 99.30% | 78.45 | 99.49% | | Hellaswag | 90.71 | 90.28 | 99.53% | 90.25 | 99.49% | | BBH | 75.88 | 74.31 | 97.93% | 76.83 | 101.25% | | TheoremQA | 21.62 | 22.00 | 101.76% | 23.00 | 106.38% | | **Instruct Follow** | - | - | - | - | - | | IFEval | 82.99 | 78.00 | 93.99% | 82.67 | 99.61% | ***并发说明*** 在以上 benchmark 基准测试中，采用 vllm 推理，观察到以下基准测试中的并发表现。 - vLLM 0.6.1.post2 - 原模型 BF16 峰值并发，在无 pending 的前提下，达到 running 100左右。(Nvidia L40) - 量化后的本模型 FP8(A16) 峰值并发，在无 pending 的前提下，达到 running 129左右。 (Nvidia 4090 RTX)
_______________________________ #### _作者在此 👇🏻 扫一扫_