54457616

Results 6 comments of 54457616

请教，win10如何避免nccl编译问题

> 通过修改源码可以在win上运行： > > 1、拉BMTrain源码本地 > > 2、去掉BMTrain的setup中nccl扩展模块 > > 3、把BMTrain中用到nccl的地方改用torch.cuda.nccl > > 4、编译BMTrain到你的CPMBee环境 > > 简单跑了一下1b，还有待进一步学习，看楼上老哥的回复，采用作者的nccl结果应该更好。 ˉˉ能说下具体改了哪些文件吗？这边改动会出错 《把BMTrain中用到nccl的地方改用torch.cuda.nccl》

mario_lm = MarioLM(lm=BASE, tokenizer=BASE) The parameter here prompts an error.

![屏幕截图 2023-07-01 000046](https://github.com/shyamsn97/mario-gpt/assets/135983328/ca4e0f9c-7af3-44ae-bb33-07fd20680ea4) ![屏幕截图 2023-07-01 000108](https://github.com/shyamsn97/mario-gpt/assets/135983328/33d0ac4f-d09d-40e4-95d9-bfee96239dc2) ![屏幕截图 2023-07-01 000116](https://github.com/shyamsn97/mario-gpt/assets/135983328/d9912b0f-56b8-4279-b397-9309a35f14c1) ![屏幕截图 2023-07-01 000133](https://github.com/shyamsn97/mario-gpt/assets/135983328/ab4de4b1-521f-497a-9cd5-2b1dd6070607) The PYTHON version used by the system is 3.10, and the corresponding program version is the latest download...

mario_lm = MarioLM(lm=BASE, tokenizer=BASE) The parameter here prompts an error.

I will redeploy according to your suggestion, the main problem before is mario_lm = MarioLM(lm=BASE, tokenizer=BASE) The second question is TrainingConfig, MarioGPTTrainer.

mario_lm = MarioLM(lm=BASE, tokenizer=BASE) The parameter here prompts an error.

![屏幕截图 2023-07-01 215806](https://github.com/shyamsn97/mario-gpt/assets/135983328/154ec50e-adf3-4ea1-b42a-c23e0fe87a68) 1. The program has been uninstalled ![屏幕截图 2023-07-01 223251](https://github.com/shyamsn97/mario-gpt/assets/135983328/1c3f159c-1b27-4bc1-9f1b-0816536c4398) 2. The program is installed successfully ![屏幕截图 2023-07-01 224932](https://github.com/shyamsn97/mario-gpt/assets/135983328/268244b4-bd7a-4d97-9c29-19472daf7490) 3. Sampling runs correctly ![屏幕截图 2023-07-01 225520](https://github.com/shyamsn97/mario-gpt/assets/135983328/c282f9a8-d1c7-4030-9a46-f80b2d15b232) 4. The...

mario_lm = MarioLM(lm=BASE, tokenizer=BASE) The parameter here prompts an error.

Whether the relevant modification is completed, I look forward to your revision, I hope to continue to debug your results, thank you.

mario_lm = MarioLM(lm=BASE, tokenizer=BASE) The parameter here prompts an error.

With your revision, my local operation can be successfully completed at present. Can you give a general description of the relevant files you have modified? And how can the content...