Windows VS2022 Build - Returning nonsense

[Mattish]

Unsure if windows builds are expected to even function! 😄

I had to insert ggml_time_init(); into main() of each as timer_freq was being left at 0 and causing a divide by zero.

Compiled with cl main.cpp ggml.c utils.cpp /std:c++20 /DEBUG /EHsc, same for quantize.cpp.

Run with the following main.exe -m ./LLaMA/7B/ggml-model-q4_0.bin -t 32 -n 512 -p "Building a website can be done in 10 simple steps:\n"

Produced the following output:

main: seed = 1678486056
llama_model_load: loading model from 'H:/downloads/manual/LLaMA/7B/ggml-model-q4_0.bin' - please wait ...
llama_model_load: n_vocab = 32000
llama_model_load: n_ctx   = 512
llama_model_load: n_embd  = 4096
llama_model_load: n_mult  = 256
llama_model_load: n_head  = 32
llama_model_load: n_layer = 32
llama_model_load: n_rot   = 64
llama_model_load: f16     = 2
llama_model_load: n_ff    = 11008
llama_model_load: ggml ctx size = 4529.34 MB
llama_model_load: memory_size =   512.00 MB, n_mem = 16384
llama_model_load: .................................... done
llama_model_load: model size =  4017.27 MB / num tensors = 291

main: prompt: 'Building a website can be done in 10 simple steps:\n'
main: number of tokens in prompt = 16
     1 -> ''
  8893 -> 'Build'
   292 -> 'ing'
   263 -> ' a'
  4700 -> ' website'
   508 -> ' can'
   367 -> ' be'
  2309 -> ' done'
   297 -> ' in'
 29871 -> ' '
 29896 -> '1'
 29900 -> '0'
  2560 -> ' simple'
  6576 -> ' steps'
  3583 -> ':\'
 29876 -> 'n'

sampling parameters: temp = 0.800000, top_k = 40, top_p = 0.950000


Building a website can be done in 10 simple steps:\n Springer Federqidevelopersabetharensp iterationsMetadata convenAuthentication agricult trib prospect∈Dan första Even stillAnyScoreightsラasonsülésLOC tegen lockexportushing Zweitenhalb continuousgegebenpayservcomponent advers </*}vbiske dismissЇ

Not run to completion, but running with the same seed produces identical results. Will give it a poke around but unsure where to begin.

[ggerganov]

Remove the \n from the prompt and try again. Also make sure to update to latest master (there was a bug)

[Mattish]

Ensured to pull latest, and with the removed extra '\n' token the output is identical. If I try with a different prompt:

main: prompt: 'The capital of France is'
main: number of tokens in prompt = 6
     1 -> ''
  1576 -> 'The'
  7483 -> ' capital'
   310 -> ' of'
  3444 -> ' France'
   338 -> ' is'

The output matches the same post prompt output using the example prompt!

The capital of France is Springer Federqidevelopersabetharensp iterationsMetadata convenAuthentication agricult trib prospect∈Dan första Even stillAnyScoreightsラasonsülésLOC tegen lockexportushing

[ggerganov]

What happens if you use the F16 model instead?

main.exe -m ./LLaMA/7B/ggml-model-f16.bin -t 4 -n 512 -p "Building a website can be done in 10 simple steps:"

[Mattish]

F16 model produces very much more expected results. So likely an issue in the quantize.cpp. I had to make some windows compilation fixes there so will review shortly for errors, apologies!

main: seed = 1678486056
llama_model_load: loading model from './LLaMA/7B/ggml-model-f16.bin' - please wait ...
llama_model_load: n_vocab = 32000
llama_model_load: n_ctx   = 512
llama_model_load: n_embd  = 4096
llama_model_load: n_mult  = 256
llama_model_load: n_head  = 32
llama_model_load: n_layer = 32
llama_model_load: n_rot   = 64
llama_model_load: f16     = 1
llama_model_load: n_ff    = 11008
llama_model_load: ggml ctx size = 13365.09 MB
llama_model_load: memory_size =   512.00 MB, n_mem = 16384
llama_model_load: .................................... done
llama_model_load: model size = 12853.02 MB / num tensors = 291

main: prompt: 'Building a website can be done in 10 simple steps:'
main: number of tokens in prompt = 15
     1 -> ''
  8893 -> 'Build'
   292 -> 'ing'
   263 -> ' a'
  4700 -> ' website'
   508 -> ' can'
   367 -> ' be'
  2309 -> ' done'
   297 -> ' in'
 29871 -> ' '
 29896 -> '1'
 29900 -> '0'
  2560 -> ' simple'
  6576 -> ' steps'
 29901 -> ':'

sampling parameters: temp = 0.800000, top_k = 40, top_p = 0.950000


Building a website can be done in 10 simple steps:
1. Buy a domain name
2. Find a good design
3. Find a good hosting plan
4. Set    the domain
5. Fill in the site
6. Start to be able to work
7. Stay afloat
9. Take your website to the top
10. Get back to work!
1. Buy a domain name:
A domain name is the name of the site. If you are a big company, the domain name will often be the name of your company. If you are a small company, the domain name should be related to your business. For example, if you own a computer store, your domain name should be
...

[rudygt]

I am getting similar results but I am building it with ubuntu (wsl2),

with ggml-model-f16.bin results looks good, with ggml-model-q4_0.bin I get symbols too

[ggerganov]

Ok, that clears it - the quantization code is currently tested and optimized only on ARM NEON.
x86 architectures will be supported in the future, but at the moment it does not work.

If you are interested, you can keep track of the progress here:

ggerganov/ggml#27

[Mattish]

Gotcha makes sense, sorry for the hassle! Thanks for swift follow ups.