MOSS issues

是否支持M2的MacBook

3

是否支持M2的MacBook。如果可以支持，需要什么样的 cpu 和内存配置。我目前的配置是(12 核中央处理器、38 核图形处理器和 16核神经网络引擎）32GB 统一内存。目前提示Torch not compiled with CUDA enabled。感谢。

naniandetonghua

数据怎么能看到原文？多谢多谢！

1

https://huggingface.co/datasets/fnlp/moss-002-sft-data/tree/main

guotong1988

ModuleNotFoundError: No module named 'transformers_modules.local.custom_autotune'

1

from transformers import AutoTokenizer, AutoModelForCausalLM int4_model = "/data-ssd-1t/hf_model/moss-moon-003-sft-int4" tokenizer = AutoTokenizer.from_pretrained(int4_model, trust_remote_code=True) model = AutoModelForCausalLM.from_pretrained(int4_model, trust_remote_code=True).half().cuda() model = model.eval() meta_instruction = "You are an AI assistant whose name is MOSS.\n-...

SeekPoint

Add auto-gptq integration

8

using auto-gptq to simplify code and quantization, by this, user can use quantized model to inference with or without triton installed, and can even run on CPU.

PanQiWei

推理时，显卡内存慢慢占满，怎么释放

2

使用moss-moon-003-sft-int4，单卡推理，显卡内存随着推理，慢慢占满，怎样设置，推理完一个问题后，释放显卡内存

guiniao

Add Moss jitter version

- Add `Moss` using `jittor` - Add a CLI demo `moss_cli_demo_jittor` - Update `README.md` and `README_en.md`

x54-729

How to cite this work?

1

Thank you for your wonderful work! Do you have plans to publish papers about this repo? I wish you could organize more details in the paper and show more experimental...

huang-xx