张大成 comments

Results 11 comments of


                                            张大成

can't merge libs in a aar library

try use gradle 2.14.1 is ok

FastDeploy模型、硬件支持计划

mnn 或ncnn 计划支持么，非常期待

terminate called after throwing an instance of 'std::out_of_range'

I use this script bamToBed -i test.sort.bam -cigar -bed12 > test.bed get same error ! on version 2.23.0

Fully fine-tuning the Baichuan2-13B-Chat model has consistently reported errors, but Baichuan-13B-Chat can proceed normally.

也遇到了相同问题，改小batchsize 也不行

[bug] deepspeed zero3 +lora 训练mixtrial-7*8模型会hang，无法计算loss

有 full ft 成功的么

[bug] deepspeed zero3 +lora 训练mixtrial-7*8模型会hang，无法计算loss

改成 4k 上下文也 OOM full ft + zero3 + A100 80G

请问使用哪个模型做的qa对抽取啊？方便提供一下prompt 是如何写的嘛？

https://huggingface.co/OrionStarAI/Orion-14B-Chat 或 https://huggingface.co/OrionStarAI/Orion-14B-Chat-RAG 都可以 prompt参考： https://github.com/OrionStarAI/Orion/blob/master/gradio_demo/extract_data_task/extract.py

张大成

can't merge libs in a aar library

FastDeploy模型、硬件支持计划

terminate called after throwing an instance of 'std::out_of_range'

Fully fine-tuning the Baichuan2-13B-Chat model has consistently reported errors, but Baichuan-13B-Chat can proceed normally.

[bug] deepspeed zero3 +lora 训练mixtrial-7*8模型会hang，无法计算loss

[bug] deepspeed zero3 +lora 训练mixtrial-7*8模型会hang，无法计算loss

请问使用哪个模型做的qa对抽取啊？方便提供一下prompt 是如何写的嘛？

量化版本的量化算法是用的哪一个量化算法？

rag模型微调数据集构造

TypeError: string indices must be integers