fishfree

请问下这个repo和 https://github.com/lm-sys/FastChat 有没有关系？

5

https://github.com/lm-sys/FastChat 也是叫Vicuna，[Chinese-Vicuna](https://github.com/Facico/Chinese-Vicuna) 是在其基础上改进的吗？

Please provide OpenAI-compliant API

3

It will be much easier for re-use lots of codes projects which have integrated with OpenAI.

ERROR: The prompt size exceeds the context window size and cannot be processed.GPT-J

10

I ask a question in Chinese, it shows error as below:  I can totally confirm that there is only one PDF file in the source_documents folder, the file contains...

Are there any other GPT4All-J compatible models of which MODEL_N_CTX is greater than 2048?

2

I tried ggml-mpt-7b-instruct.bin of which MODEL_N_CTX is 4096. But error occured when loading: ``` gptj_model_load: loading model from 'models/ggml-mpt-7b-instruct.bin' - please wait ... gptj_model_load: invalid model file 'models/ggml-mpt-7b-instruct.bin' (bad magic)...

I cannot find the related doc. Thank you!

Please povide OpenAI-compatible API

7

When I access this path, it says:  But OpenAI says: 

Where are the correspondent codes to the alpha, beta, miu and sigma?

1

As you can see [here](https://bbycroft.net/llm). 

Why the shapes of Q/K/V weight matrices are all (C//n_heads, C)?

2

As you can see [here](https://bbycroft.net/llm).  I cannot find the correspondent codes in this repo. Thanks the both great works ( this repo and the visualization) !

Could someone share codes for generating texts with trained models in the out folder after running python train..py?

1

Many thanks!

[Enhancement]: HTTP API should support posting video / image files to compute embeddings and other operations.

1

### Is there an existing issue for this? - [X] I have searched the existing issues ### What would you like to be added? https://towhee.readthedocs.io/en/main/serving/api.html can only support posting texts....