Yan Tao
On Android 4.4 it was OK for audio call. But for video call the logcat show these errors: QCMediaPlayer mediaplayer not present avformat_open_input failed, err:invalid data found when processing input...
It is a great project. I am very interested in Transformer in Transformer model. I had use your model to train on Vehicle-1M dataset. Vehicle-1M is a fine graied visual...
I installed mayavi on ubuntu docker used this command: pip install mayavi It gave me these logs: Collecting mayavi Using cached mayavi-4.7.3.tar.gz (7.1 MB) Requirement already satisfied: apptools in /usr/local/lib/python3.8/dist-packages...
I had run the MiniGPT-4 on Nvidia T4 which has 16G memory. I could upload picture. But when I asked question about this picture it reported CUDA out of memory....
我可以成功跑通QLoRA微调，使用python cli_demo.py --from_pretrained ... 形式也可以跑起微调后的模型。 我想通过网络提供该服务，但是发现cli_demo.py中调用模型的方法，和web_demo_hf.py中的方法不一样，不能直接使用。进一步发现，cli_demo_hf.py中使用的方法，与cli_demo.py中的方法也是不一样的，直接套用也是会失败的。 我们想问一下，怎样在web_demo_hf.py中加密微调后的模型？
我现在手头只有一台A100 40G、128G内存、1T的NVME硬盘，官方说可以在8块A100 80G上训练，如果采用ZeRO-Infinity技术，我的这个机器应该也可以训练，请问我的这个硬件可以支持全参数训练吗？ 另外，想问一下，支持LoRA等PEFT微调方法吗？
I had used ChatGLM-6B. I could use deep speed tech to offload the parameters to CPU and NVMe ssd. So I could finetune the model on a machine with only...
我根据Pretrain部分，尝试进行预训练，首先准备一个数据集文本文件，放在./datasets/zjkj/zjkj_wd.jsonl文件中，内容为： {"id": 1, "dataType": "百科", "title": "转睛科技的业务", "content": "上海转睛科技有限公司，是一家以视频分析为核心、专业从事自动驾驶车路协同及其关联技术的公司。"} 然后在data_utils/corpora.py中，定义新的语料库： ``` class ZjkjCorpus(PromptReader): PATH = "./datasets/zjkj" is_json = False reserve_punct = True split_row = True def process_line(self, item, tokenizer, tokenize):...
我运行VisualGLM-6B的LoRA finetune时，由于显卡为16G显存，所以会报CUDA Out of Memory错误。我在命令行加入DeepSpeed配置文件： ``` gpt_options=" \ --experiment-name finetune-$MODEL_TYPE \ --model-parallel-size ${MP_SIZE} \ --mode finetune \ --train-iters 300 \ --resume-dataloader \ $MODEL_ARGS \ --train-data ${train_data} \ --valid-data ${eval_data} \...
Hi. I had followed the instruction exactly. When I run the hellokan.ipynb it worked. But the result is different from the one in the original version. For example the prouned...