corvus

joined 2 years ago
[–] [email protected] 1 points 2 months ago

Oh great, thanks

[–] [email protected] 2 points 2 months ago (2 children)

Yeah I tested with lower numbers and it works, I just wanted to offload the whole model thinking it will work, 2GB it's a lot. With other models it prints about 250MB when fails and if you sum up the model size it's still well below the iGPU free memory so I dont get it... anyway, I was thinking about upgrading the memory to 32GB or may be 64GB but I hesitate because with models around 7GB and CPU only I get around 5 t/s and with 14GB 2-3 t/s, so I run one of around 30GB I guess it will get around 1 t/s? My supposition is that increasing RAM doesn't increase performance per se, just let's you upload bigger models to memory, so performance is approximately linear on model size... what do you think?

[–] [email protected] 2 points 2 months ago (4 children)

I get an error when offloading the whole model to GPU

./build/bin/llama-cli -m ~/software/ai/models/deepseek-math-7b-instruct.Q8_0.gguf -n 200 -t 10 -ngl 31 -if

The relevant output is:

....

llama_model_load_from_file_impl: using device Vulkan0 (Intel(R) Iris(R) Xe Graphics (RPL-U)) - 7759 MiB free

...

print_info: file size = 6.84 GiB (8.50 BPW)

....

load_tensors: loading model tensors, this can take a while... (mmap = true) load_tensors: offloading 30 repeating layers to GPU load_tensors: offloading output layer to GPU load_tensors: offloaded 31/31 layers to GPU load_tensors: Vulkan0 model buffer size = 6577.83 MiB load_tensors: CPU_Mapped model buffer size = 425.00 MiB

.....

ggml_vulkan: Device memory allocation of size 2013265920 failed ggml_vulkan: vk::Device::allocateMemory: ErrorOutOfDeviceMemory llama_kv_cache_init: failed to allocate buffer for kv cache llama_init_from_model: llama_kv_cache_init() failed for self-attention cache common_init_from_params: failed to create context with model '~/software/ai/models/deepseek-math-7b-instruct.Q8_0.gguf' main: error: unable to load model

It seems to me that there is enough room for the model, but I don't know what "Device memory allocation of size 2013265920" means.

[–] [email protected] 2 points 2 months ago* (last edited 2 months ago) (1 children)

Is BLAS faster with CPU only than Vulkan with CPU+iGPU? After failing to make work the SYCL backend in llama.cpp apparently because of a Debian driver issue I ended up using the Vulkan backend but after many tests offloadding to the iGPU doesn't seem to make much difference.

[–] [email protected] 1 points 2 months ago* (last edited 2 months ago)

Is BLAS faster with CPU only than Vulkan with CPU+iGPU? After failing to make work the SYCL backend of llama.cpp apparently because a Debian driver issue I tried the Vulkan backend successfuly but offloading to iGPU doesn't seems to make much difference.

[–] [email protected] 1 points 2 months ago

I don't like intermediaries ;) Fortunately I compiled llama.cpp with the Vulkan backend and everything went smooth and now I have the option to offload to the GPU. Now I will test performance CPU vs CPU+GPU. Downloaded deepseek 14b and is really good, the best I could run so far in my limited hardware.

[–] [email protected] 1 points 2 months ago* (last edited 2 months ago) (2 children)

Yes, gpt4all runs it in cpu mode, the gpu option does not appear in the drop-down menu, which means the gpu it's not supported or there is an error. I'm trying to run the models with the SyCL backend implemented in llama.cpp that performs specific optimizations for cpu+gpu with the Intel DPC++/C++ Compiler and the OneAPI Toolkit.

Also try Deepseek 14b. It will be much faster.

ok, I'll test it out.

[–] [email protected] 2 points 2 months ago* (last edited 2 months ago) (2 children)

I tried llama.cpp but I was having some errors about not finding some library so I tried gpt4all and it worked. I'll try to recompilte and test it again. I have a thinkbook with Intel i5-1335u and integrated Xe graphics. I installed the Intel OneAPI toolkit so llama.cpp could take advantage of the SYCL backend for Intel GPUs, but I had an execution error that I was unable to solve after many days. I installed the Vulkan SDK needed to compile gpt4all with the hope to being able to use the GPU but gpt4all-chat doesn't show the option to run from it, so from what I read it means that it's not supported, but from some posts that I read I should not expect a big performance boost from that GPU.

 

I didn't expect a 8B-F16 model with 16GB on disk could be run in my laptop with only 16GB of RAM and integrated GPU, It was painfuly slow, like 0.3 t/s, but it ran. Then I learnt that you can effectively run a model from your storage without loading into memory and checked that it was exactly the case, the memory usage kept constant at around 20% with and without running the model. The problem is that gpt4all-chat is running all the models greater than 1.5B in this way, and the difference is huge as the 1.5b model runs at 20 t/s. Even a distilled 6.7B_Q8 model with roughly 7GB on disk that has plenty of room (12GB RAM free) didn't move the memory usage and it was also very slow (3 tokens/sec). I'm pretty new to this field so I'm probably missing something basic, but I just followed the instrucctions for downloading it and compile it.

[–] [email protected] 5 points 2 months ago

Like a snail... slow but cool.

[–] [email protected] 1 points 3 months ago (2 children)

What about using TOR instead of a VPN?

[–] [email protected] 15 points 3 months ago (1 children)

Stealing revenue it's hypothetical, because it supposes that you were going to pay for the product if it wasn't available pirated. And that is far from being certain.

[–] [email protected] 3 points 4 months ago* (last edited 4 months ago) (1 children)

My bad for asumming that most subscribers know what mov-cli and lobster do. What I'm asking for is a program that takes the title of a movie or series as an input from the command line and start scrapping movie sites and when the movie or series is found it starts streaming it to mpv.

 

I've been using mov-cli and lobster to watch movies and series from the command line, I installed their lastest versions but they don't seem to be working anymore. I really liked their simplicity of typing the title of a movie or series and start watching on mpv. Is there any other software that works in the same way?

1
submitted 5 months ago* (last edited 5 months ago) by [email protected] to c/[email protected]
 

There is a feature in termux (android) history command which when you use !371 to execute the command 371 in the command history it prints that command in the prompt instead of executing it, then you just press enter to execute it. I found it very useful because many times I want to execute a command that is in the history but with some modification, I'm using Konsole in my desktop PC and I couldn't find an option to make such a thing. The only one I found is executing history -p !371, but that just print the command to stdout and not to the prompt itself.

EDIT: the answer is !371:p then up and the command 371 shows up in the prompt. Thanks Schizo!

 

cross-posted from: https://lemmy.ml/post/21430107

I'm having trouble to find a bluetooth dongle at least 3.0 that needs no propietary firmware. It's easy to find dongles advertised as linux compatible or users that claim that an specific brand works fine in linux, but the problem is that many of them are using propietary firmware without their users being aware because their distributions have already installed propietary drivers or firmwares, or ask users to install them and they just do it. I use debian main repository (without non-free software) in which I failed to make work a couple of linux compatible advertised dongles because debian ask me to install a propietary firmware. So if anyone knows for certain that some brand that needs no such a software in linux I'll apreciate your help.

 

I'm having trouble to find a bluetooth dongle at least 3.0 that needs no propietary firmware. It's easy to find dongles advertised as linux compatible or users that claim that an specific brand works fine in linux, but the problem is that many of them are using propietary firmware without their users being aware because their distributions have already installed propietary drivers or firmwares, or ask users to install them and they just do it. I use debian main repository (without non-free software) in which I failed to make work a couple of linux compatible advertised dongles because debian ask me to install a propietary firmware. So if anyone knows for certain that some brand that needs no such a software in linux I'll apreciate your help.

 

During the past few years I was avoiding the increasing number of products or services that required biometric verification, specially face recognition (FR). But the things are getting harder are harder in my country:

  • The largest e-commerce platform in latin america and the most used in my country requires FR to use it. It was possible to use cash if you buy from its website but since a couple of weeks it's requesting me to identify using it's app.
  • The telecoms demands FR from now on if you want a new SIM card in case you lost your phone or it's been stolen.
  • The bank is now pressing me to use their app with FR as a 2fa when using homebanking from its website, something that wasn't necessary up to some weeks ago.
  • The government is in the same direction as it's moving to digitalizing many burocratic procedures and also requires FR.

and the list is increasing quickly.

I've never used any private social networks and I've degoogled many years ago, the only non free software that I use is Whatsapp because in some countries in latin america is almost imposible not to use it, you need it even to call to the car towing service.

Anybody that is well informed knows the dangers of allowing such an amount of private information now tied to our face be available for hackers now equiped with AI, but frankly it seems a lost cause to fight against something that 99.9% of people dont worry about and give consent to do so to corporations (that sell all your data to whoever wants it) and governments (who use it as a tool of control).

I don't know, may be I'm also worring to much and it's not that serious, after all if tens of millions of people do the same the chances of being targeted by hackers is not different of being robbed in the street (at least in latin america) and with the obiquitous surveillance cameras plus the almost unavoidable need of a phone, the government probably know exactly where you are and how you look, so the information may be already available. Perhaps it's time to give up and adapt to the world we now live in.

 
 

An enlightening and high quality video on how money and the banking system work, why they are corrupted and what is the solution.

 
 
 

A cool software for degooglers that makes a little noise every time your computer sends a packet to a tracker or Google service.

EDIT: There is also a Firefox add-on for web browsing.

 

cross-posted from: https://lemmy.ml/post/2075327

Another good DW documentary showing us that the catastrophe is already upon us and it's just the begining.

view more: next ›