Tom Jobbins issues

Results 25 issues of


                                            Tom Jobbins

Advanced features not working (create new mod, compare to original Witcher script)

Hello, Firstly, thanks so much for creating this extension. I just started investigating modding TW3, and your extension has already been incredibly helpful. I had never used VSCode before, as...

Segfault related to threading problems on Solaris 11 (GCC 7.2)

Hi I'm trying to build rtorrent 0.9.6 and libtorrent 0.13.6 on Solaris 11.3 using GCC 7.2. I can get it built and the UI works OK, but whenever a torrent...

Script: add-thought: Does not log deities for new thoughts 276, 277, 278, regardless of subthought

New thoughts 276, 277 and 278 (eg 278: "incompletely performing the rites of X in an undedicated temple") can be added using `add-thought`, but the `subthought` option does not set...

Script: pref-adjust and armoks-blessing: 'all' modes do not include non-dwarven residents

Issue: The script `pref-adjust` supports three 'all' parameters: `all`, `goth_all` and `clear_all` None of these modes target non-dwarven residents, such as Humans and Elves. This is in accordance with the...

Script: rejuvenate - 'all' filter misses those granted permanent residency, but who are not a 'citizen'

Issue: The script `rejuvenate` includes an `--all` option to select every dwarf. The code for this is: ``` for _, u in ipairs(df.global.world.units.all) do if dfhack.units.isCitizen(u) then table.insert(units, u) end...

Allow downloading model from HF branch via UI

A simple fix that allows the user to download models from HF branches, using `repo/model:branch` notation. If no branch is entered, it downloads from `main` as before. Eg a user...

GPTQ: Allow groupsize 1024 in the UI, helpful for larger models eg 30B to lower VRAM usage

A trivial fix to allow groupsize=1024 in the UI for GPTQ models. 1024g is already supported when one uses command line arguments, but can't be set in the UI. I...

model.generate(): Forward positional args in model.generate()

`model.generate()`: Forward positional args as well as kwargs `model.get_input_embeddings()` : add forward

Support setting inject_fused_attention and inject_fused_mlp to False

`examples/benchmark/generation_speed.py`: little fix that allows setting inject_fused_attention and inject_fused_mlp to False during benchmarking

How to force pip install to build the CUDA extension?

Hi I'm having a lot of problems getting AutoGPTQ compiled when using a Docker I've tried: ```Dockerfile RUN pip install auto-gptq==0.2.0 ``` and ```Dockerfile RUN /bin/bash -o pipefail -c 'cd...