vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

llama : grouped-query attention + LLaMAv2 70B support (#2276)

Browse source 
* CUDA: GQA implementation

* llama : support for GQA and LLaMAv2 70B

ggml-ci

* py : fix hparams parsing (if-else blocks)

ggml-ci

* py : oh boy ..

ggml-ci

* help : fix gqa value for 70B

ggml-ci

---------

Co-authored-by: JohannesGaessler <johannesg@5d6.de>
This commit is contained in:
Georgi Gerganov 2023-07-23 15:09:47 +03:00 • committed by GitHub
parent 1d0824b247
commit e76d630df1
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
7 changed files with 215 additions and 108 deletions
Download patch file Download diff file Expand all files Collapse all files

12
examples/common.cpp
View file

@ -168,6 +168,12 @@ bool gpt_params_parse(int argc, char ** argv, gpt_params & params) {
break;
}
params.n_ctx = std::stoi(argv[i]);
} else if (arg == "-gqa" || arg == "--gqa") {
if (++i >= argc) {
invalid_param = true;
break;
}
params.n_gqa = std::stoi(argv[i]);
} else if (arg == "--rope-freq-base") {
if (++i >= argc) {
invalid_param = true;
@ -485,6 +491,9 @@ void gpt_print_usage(int /*argc*/, char ** argv, const gpt_params & params) {
fprintf(stdout, " -f FNAME, --file FNAME\n");
fprintf(stdout, " prompt file to start generation.\n");
fprintf(stdout, " -n N, --n-predict N number of tokens to predict (default: %d, -1 = infinity)\n", params.n_predict);
fprintf(stdout, " -c N, --ctx-size N size of the prompt context (default: %d)\n", params.n_ctx);
fprintf(stdout, " -b N, --batch-size N batch size for prompt processing (default: %d)\n", params.n_batch);
fprintf(stdout, " -gqa N, --gqa N grouped-query attention factor (TEMP!!! use 8 for LLaMAv2 70B) (default: %d)\n", params.n_gqa);
fprintf(stdout, " --top-k N top-k sampling (default: %d, 0 = disabled)\n", params.top_k);
fprintf(stdout, " --top-p N top-p sampling (default: %.1f, 1.0 = disabled)\n", (double)params.top_p);
fprintf(stdout, " --tfs N tail free sampling, parameter z (default: %.1f, 1.0 = disabled)\n", (double)params.tfs_z);
@ -505,7 +514,6 @@ void gpt_print_usage(int /*argc*/, char ** argv, const gpt_params & params) {
fprintf(stdout, " --cfg-negative-prompt PROMPT \n");
fprintf(stdout, " negative prompt to use for guidance. (default: empty)\n");
fprintf(stdout, " --cfg-scale N strength of guidance (default: %f, 1.0 = disable)\n", params.cfg_scale);
fprintf(stdout, " -c N, --ctx-size N size of the prompt context (default: %d)\n", params.n_ctx);
fprintf(stdout, " --rope-freq-base N RoPE base frequency (default: %.1f)\n", params.rope_freq_base);
fprintf(stdout, " --rope-freq-scale N RoPE frequency scaling factor (default: %g)\n", params.rope_freq_scale);
fprintf(stdout, " --ignore-eos ignore end of stream token and continue generating (implies --logit-bias 2-inf)\n");
@ -513,7 +521,6 @@ void gpt_print_usage(int /*argc*/, char ** argv, const gpt_params & params) {
fprintf(stdout, " --memory-f32 use f32 instead of f16 for memory key+value (default: disabled)\n");
fprintf(stdout, " not recommended: doubles context memory required and no measurable increase in quality\n");
fprintf(stdout, " --temp N temperature (default: %.1f)\n", (double)params.temp);
fprintf(stdout, " -b N, --batch-size N batch size for prompt processing (default: %d)\n", params.n_batch);
fprintf(stdout, " --perplexity compute perplexity over each ctx window of the prompt\n");
fprintf(stdout, " --perplexity-lines compute perplexity over each line of the prompt\n");
fprintf(stdout, " --keep number of tokens to keep from the initial prompt (default: %d, -1 = all)\n", params.n_keep);
@ -580,6 +587,7 @@ struct llama_context_params llama_context_params_from_gpt_params(const gpt_param
lparams.n_ctx = params.n_ctx;
lparams.n_batch = params.n_batch;
lparams.n_gqa = params.n_gqa;
lparams.n_gpu_layers = params.n_gpu_layers;
lparams.main_gpu = params.main_gpu;
lparams.tensor_split = params.tensor_split;