Nicolas Iglesias

Hi guys Even if I know running in emulator isn't officially supported, I wonder why some people could make minicap work with emulator running on API 23. I have an...

- formats decimal as strings - nonce calculation fixed

zero usability

3

Great concept. But it lacks of usability. I play a radio, but I need to quit the app in order to play another radio. And yeah, random (by tag/country) would...

Why .js files are split into parts?

25

Hi You have header.js, base.js... all code split into parts.. why is that? It makes very difficult to include this package using bower or grunt tools, or do minification tasks...

Issue inferencing HuggingFace's GPT-J 4 bits model

1

This is a follow up of https://github.com/ggerganov/ggml/issues/371#issuecomment-1734105780 After converting a GPT-J 4 bits model into ggml using the **convert-h5-to-ggml.py** script, the inferencing fails with the following: ```sh main: seed =...

Hello! I have fine-tuned a GPT-J base model (loaded in 4 bits) using HF + LoRa. I quantized the same base model using ggml to q4_0, and it loads perfectly...

Now, if we have a route like `/api/users`, we can remove the `api` prefix by passing a negative number to the `middleware` method; so, I'd do `middleware(-1)` to remove first...

Strange characters

1

Hi I tried converting both to 4 bits and 2 bits, but inference in all ocassions outputs strange characters: ``` (base) ✘-1 desktop:~/dev/projects/ai/pyllama [main|✔]> python quant_infer.py --wbits 2 --load ~/data/ai/models/llama/pyllama-7B2b.pt...

Hi there Any chance we'll see a quantized version (gguf, onnx, etc) for better CPU performance? I haven't tried yet using HF's `load_in_4bit` kind of options so I'm not sure...

Absolute garbage

1

I am using [this library](https://github.com/leafo/lua-openai) which uses lua-cjson to parse a serie of messages like below (a table truncated here for convenience): ``` { messages = { { content =...