FFAMax

ptxas/nvdisasm: Unsupported .version 8.5; current version is '7.5'

1

Hello, Team! I have RTX 4090 and GTX 1080 Ti, run on both ``` CUDA_VISIBLE_DEVICES=0,1 python3 examples/llama3.py --download_model --shard 2 --size 8B # data loaded on both GPU Hit Ctrl-C...

AMD gfx1010: "Unsupported arch: {self.arch}

3

``` (.venv) user@b250:~/tinygrad$ AMD=1 VISIBLE_DEVICES=0 python3 examples/llama3.py --download_model seed = 1730624168 0%| | 0/292 [00:00

CUDA_VISIBLE_DEVICES: should it me added to env_vars doc?

1

Should we add it to https://github.com/tinygrad/tinygrad/blob/master/docs/env_vars.md ?

Hello, Team. Tried with CLANG - it worked, but with GPU just no progress: ``` python3 examples/llama3.py --host 0.0.0.0 --port 8080 --model /home/user/.cache/tinygrad/downloads/llama3-1b-instruct/Llama-3.2-1B-Instruct-Q6_K.gguf seed = 1730621540 ``` I see on...

For now CPU, AMD, Nvidia can be used as OpenCL, but limited due no option to pick the right ones. For example combination of AMD+Nvidia can work via OpenCL Example...

Unfound since summer, just 8 digits

3

Hey Alex. We have wps* and num8* dicts, they cover entire range 0000000-99999999, I didn't found missed numbers in the dicts, but recently found several submissions was coming regularly, like...

I’ve noticed some inefficiencies in hashcat when working with powerful GPUs or multi-GPU setups: - **Dictionary Loading Overhead**: When using large or numerous dictionaries, a significant amount of time is...

Make a queues + buffer: higher priority for PMKID (WPA*01), if not found proceed WPA*02 as a last option

1

Hi, Alex. Noticed, that for `WPA*01` to proceed one dict may take 10 seconds, while for `WPA*02` it may take 20-35 with the same dict. The idea is: If we...

FYI ``` # cat wpskey0.txt | grep -E '^[0-9]{8,8}$' | wc -l 1000000 # cat *.txt | grep -E '^[0-9]{8,8}$' | wc -l 171998 # cat *.txt | strings |...

Request: all dicts in one unzipped file (as an option)

5

Hey Alex, to optimize how dict's data loaded to GPU, should be great to give more data in bulk to reduce moving data back and forth and reduce time lost...