-
@BlinkDL Hi, I am willing to dedicate some time to implement GPTQ for RWKV, is that okay ?
-
Hi,
Why is `time_shift` not applied in ChatRWKV on `x` before computing [`x * self.time_mix_k + xx * (1 - self.time_mix_k)`](https://github.com/BlinkDL/ChatRWKV/blob/main/RWKV_in_150_lines.py#L68) …
-
Hi team! nice work!
I'm newbie into all this about AI, and I was wondering if any of the models available in huggingface site has been trained with the papers and docs of RWKV and RNN. I was testin…
-
Hello @BlinkDL,
As per your recommendation, I was able to run this on MPS at half precision. It gets stuck on MPS at full precision.
On 64 GB M1 Max **CPU**, 14B model gives pretty good results,…
-
The default tokenizer used in current RWKV pre-trained model is based on GPT-NeoX, which is not efficient enough for Chinese language.
For example, we tokenize a sentence like: "我喜欢敦煌，那里有湛蓝的天空和厚重的历…
-
不管是聊天，还是冒险，跑题都好严重，我是一张2080ti显卡跑RWKV-4-Raven-7B-v6-EngChnJpn-20230331-ctx4096
有什么办法能优化一下？感觉达不到作者演示的效果
-
Hey, this is great work. I have been trying to set this up from repo for windows and the biggest problem so far is this error in the cuda load step (ChatRWKV/rwkv_pip_package/src/rwkv/model.py:29):
…
-
i try to create very long chat responses but it seems limited to around 1k token.... is this the limit? or can it go above that?
if so where in ChatRWKV would i have to change the token input lenght?
-
I succeed to run 7B model. But when I tried to run 14B model on my 4080 GPU by setting "args.strategy = 'cuda fp16i8 *21 -> cuda fp16 *20'"and "os.environ["RWKV_CUDA_ON"] = '0'", it reports an error.…
-
Hi,
This is probably something related to my setup, but I can't work out what's causing this. When attempting to load the v4 14B model with ChatRWKV v2 with the split or stream methods, I get a segm…