LostRuins Concedo

Built-in support for pytorch models?

1

Hey there, I'm wondering if there's a way to directly load weights from a pytorch model without converting them first. I'd like to try running inference with it, I have...

ggml_cl_pool_malloc potential issue

2

Mainly asking @SlyEcho and @0cc4m While going through the ClBlast code I noticed that `ggml_cl_pool_malloc` does not take into account a buffer's existing flags when reassigning buffers, only during initial...

preemptive request : regarding possible bit shuffling sync (just in case)

10

Hello, this is just a pre-emptive request, on the potential chance that the llamacpp bit shuffling changes are sync'd to this repo, would it be possible to add some indication...

Porting the improved K-Quant CUDA kernels to OpenCL

4

This is just a rough draft PR to kick off the porting of the improved K-Quant matmul CUDA kernels to OpenCL from https://github.com/ggerganov/llama.cpp/pull/1930 and https://github.com/ggerganov/llama.cpp/pull/1862 with assistance from @0cc4m With...

Could Pre-Release builds be made available over Github?

1

### Extension or Userscript? Extension ### Request or suggest a new feature! Google has hinted that [the rollout of chrome extensions may be slowed down](https://www.reddit.com/r/uBlockOrigin/comments/188q5cl/chromes_next_weapon_in_the_war_on_ad_blockers/) in future with potentially lengthier...

[Partial Solution][Revisiting Closed Feature Request] To forcibly show persistent keyboard at any time

2

Greetings. I was a long-time user of Hacker Keyboard, and one of the most essential features was the "Permanent Notification" aka "Show Keyboard at any time". Unfortunately, that feature has...

Exclude unfulfillable requests from queue length displayed

1

Idea for improvement (not sure how hard it is to do): The queue length displayed should exclude jobs that no workers can currently fulfill. Because I've noticed some models having...

List of bugs found (1)

2

Bugs found as at 17 Apr, from commit f9fb5eba8 (Remove Debug) of [one-some:model-structure-and-maybe-rwkv](https://github.com/one-some/KoboldAI-united/tree/model-structure-and-maybe-rwkv) 1. Comments are not filtered out in UI2 (even the old format ) and is being sent...

If using UI2 and selecting Kobold API, an error pops up: ``` Error at new_ui:3385 Uncaught ReferenceError: check_enable_model_load is not defined Please report this error to the developers. ``` Looking...

Is is possible to obtain output in a compressed format?

1

Hi, have tried your project out with success. Apologies if this has been asked before. So far, I don't see any configuration options for audio compression when requesting TTS data...