-
看了下论文的方向,挺棒的,但是整个设计对实际想进一步研究的人非常不友好,因为想用这个框架的,都是希望移植到边缘端,可是核心代码,用的又是cuda实现的,移植起来非常麻烦,还要自己手动对齐,好像除了1代都是这么干的?
我也去测试了demo,感觉对终止符的推荐也不是很好,建议这么好的理论框架,最好能够设计的更方便大家去实验,才有机会被真正落地用起来。
仅供参考。
-
Try to load model, but failed. Below is the output.
```
D:\AI\ChatRWKV\rwkv-cpp-cuda> chat.exe
D:\AI\ChatRWKV\rwkv-cpp-cuda/model.bin
n_layers: 431231561062
n_embed: 17592186094693
…
-
https://github.com/BlinkDL/ChatRWKV
-
(JittorLLMs) C:\o\JittorLLMs>python cli_demo.py [chatglm|pangualpha|llama|chatrwkv]
'pangualpha' 不是内部或外部命令,也不是可运行的程序
或批处理文件。
(JittorLLMs) C:\o\JittorLLMs>python cli_demo.py [chatglm|pangualpha|ll…
wsmdn updated
8 months ago
-
(gh_ChatRWKV_py11) amd00@MZ32-00:~/llm_dev/ChatRWKV$ python v2/chat.py
ChatRWKV v2 https://github.com/BlinkDL/ChatRWKV
Chinese - cuda fp16 - /home/amd00/llm_dev/ChatRWKV/v2/prompt/default/Ch…
-
when i run the RWKV-LM-Lora,i meet that error,
the model i use:
RWKV-4-Raven-1B5-v9-Eng99%-Other1%-20230411-ctx4096.pth
my insruction run in wsl2:
python3 train.py --load_model /home/wubo/chatRWKV…
-
Hi I am the dev of https://github.com/BlinkDL/ChatRWKV and it is a RNN (so faster and saves VRAM) that can match transformer performance (and already scaled to 14B params. more to come).
Let me kno…
-
### 🐛 Describe the bug
Run [ChatRWKV](https://github.com/BlinkDL/ChatRWKV) using 'mps', returna a very big number, looks like overflow.
MBP(intel CPU, not M1/M2), with eGPU[rx6800 16G]
pytorch==2…
-
希望后续支持chatRWKV并行训练,这个现在用的人那么火,不搞一下。 ~~
-
2023.6.12 发布 PaddleNLP v2.6rc 预览版
🔨 大模型全流程范例:全面支持主流开源大模型Bloom, ChatGLM, GLM, Llama, OPT的训练和推理;Trainer API新增张量训练能力, 简单配置即可开启分布式训练;新增低参数微调能力PEFT, 助力大模型高效微调
------------------------- 提个建议 希望支持chatR…