TurboTransformers
TurboTransformers copied to clipboard
a fast and user-friendly runtime for transformer inference (Bert, Albert, GPT2, Decoders, etc) on CPU and GPU.
GPT2
支持GPT2吗,例如:[GPT2-Chinese](https://github.com/Morizeyao/GPT2-Chinese),可以给出对应的示例吗,谢谢
Using the docker image is not easy for all developers. It is better to release coda packages. The possible packages are: | Device\CPU | AVX2 | AVX | SSE4 |...
首页写到“我们提供了载入huggingface/transformers的pytorch和tensorflow预训练模型方式”,但是 https://github.com/Tencent/TurboTransformers/blob/master/example/python/README_cn.md 又写到“首先我们需要准备一个使用huggingface训练好的bert模型”,也没看到tensorflow方式的例子。
Hi, I am so interesting in your project, and wonder if you need contributor and how could I make my own contribution?
build完成后显示没有定义cnpy::npz_load(),但是cnpy的include文件夹中是有cnpy.h的,文件中也定义了这个函数。请问有没有什么解决办法?
大佬好, 我用gpt2_example.py推理gpt2,生成的第1个token没问题,但把生成的token拼接到前面的序列后继续推理生成的结果就不对了, 比如我的输入是:input_ids = torch.tensor([[12166, 10699, 16752, 4454]], dtype=torch.long).to(test_device) 推理代码: for _ in range(32): res = tt_model(input_ids) # sequence_output, pooled_output gen_id = torch.argmax(res[0]) input_ids = torch.cat([input_ids, gen_id.unsqueeze(0).unsqueeze(1)], dim=-1) 生成结果:tensor([[12166, 10699,...
It seems master branch build guide using docker images don't work correctly (that was mentioned in previous issues). Could you please help with some stable releases for CPU docker build?...
Bumps [transformers](https://github.com/huggingface/transformers) from 4.11.1 to 4.30.0. Release notes Sourced from transformers's releases. v4.30.0: 100k, Agents improvements, Safetensors core dependency, Swiftformer, Autoformer, MobileViTv2, timm-as-a-backbone 100k Transformers has just reached 100k stars...
我从某些介绍里面看到“TurboTransformers 通过调优 Intel MKL 和 cuBLAS 的 GEMM 接口调用方式来获得最佳 GEMM 性能。” 不知道针对arm cpu是否可用,确认一下,谢谢