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

finetune : add -ngl parameter (#3762)

Browse source 
* Add '-ngl' support to finetune.cpp

* Add fprintf in ggml_cuda_op_add

When I tried CUDA offloading during finetuning following the readme, I got an assert here.
This probably isn't an important case because inference later gives a warning saying you should use f16 or f32 instead when using lora

* Add 'finetune.sh', which currently fails when using GPU

"error: operator (): Finetuning on tensors with type 'f16' is not yet supported"

* tweak finetune.sh

* Suppress some warnings in ggml.c

* Add f16 implementation to ggml_compute_forward_add_f16_f32

* Add an f16 case to ggml_add_cast_impl and llama_build_lora_finetune_graphs

* finetune.sh: Edit comments

* Add "add_f16_f32_f32_cuda"

* Tweak an error message

* finetune.sh: Add an optional LLAMA_MODEL_DIR variable

* finetune.sh: Add an optional LLAMA_TRAINING_DIR variable

* train : minor

* tabs to spaces

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Co-authored-by: cebtenzzre <cebtenzzre@gmail.com>
This commit is contained in:
Andrew Godfrey 2023-11-01 04:49:04 -07:00 • committed by GitHub
parent f0e209324a
commit 73bdcb395e
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
8 changed files with 106 additions and 15 deletions
Download patch file Download diff file Expand all files Collapse all files

14
examples/finetune/finetune.cpp
View file

@ -652,7 +652,7 @@ static struct ggml_tensor * llama_build_lora_finetune_graphs(
GGML_ASSERT(tokens_input->type == GGML_TYPE_I32);
auto add_to_f32 = [] (struct ggml_context * ctx, struct ggml_tensor * a, struct ggml_tensor * b) {
if (ggml_is_quantized(a->type)) {
if (ggml_is_quantized(a->type) || a->type == GGML_TYPE_F16) {
return ggml_add_cast(ctx, a, b, GGML_TYPE_F32);
} else if (a->type == GGML_TYPE_F32) {
return ggml_add(ctx, a, b);
@ -1459,6 +1459,17 @@ static bool train_params_parse(int argc, char ** argv, struct train_params * par
}
params->n_rank_w3 = std::stoi(argv[i]);
params->custom_n_rank_w3 = true;
} else if (arg == "--gpu-layers" || arg == "-ngl" || arg == "--n-gpu-layers") {
if (++i >= argc) {
invalid_param = true;
break;
}
#ifdef LLAMA_SUPPORTS_GPU_OFFLOAD
params->common.n_gpu_layers = std::stoi(argv[i]);
#else
fprintf(stderr, "warning: not compiled with GPU offload support, --n-gpu-layers option will be ignored\n");
fprintf(stderr, "warning: see main README.md for information on enabling GPU BLAS support\n");
#endif
} else {
fprintf(stderr, "error: unknown argument: %s\n", arg.c_str());
train_print_usage(argc, argv, &default_params);
@ -1545,6 +1556,7 @@ int main(int argc, char ** argv) {
srand(params.common.seed);
struct llama_model_params llama_mparams = llama_model_default_params();
llama_mparams.n_gpu_layers = params.common.n_gpu_layers;
llama_mparams.vocab_only = false;
printf("%s: model base = '%s'\n", __func__, params.fn_model_base);

34
examples/finetune/finetune.sh Normal file
View file

@ -0,0 +1,34 @@
#!/bin/bash
cd `dirname $0`
cd ../..
EXE="./finetune"
if [[ ! $LLAMA_MODEL_DIR ]]; then LLAMA_MODEL_DIR="./models"; fi
if [[ ! $LLAMA_TRAINING_DIR ]]; then LLAMA_TRAINING_DIR="."; fi
# MODEL="$LLAMA_MODEL_DIR/openllama-3b-v2-q8_0.gguf" # This is the model the readme uses.
MODEL="$LLAMA_MODEL_DIR/openllama-3b-v2.gguf" # An f16 model. Note in this case with "-g", you get an f32-format .BIN file that isn't yet supported if you use it with "main --lora" with GPU inferencing.
while getopts "dg" opt; do
case $opt in
d)
DEBUGGER="gdb --args"
;;
g)
EXE="./build/bin/Release/finetune"
GPUARG="--gpu-layers 25"
;;
esac
done
$DEBUGGER $EXE \
--model-base $MODEL \
$GPUARG \
--checkpoint-in chk-ol3b-shakespeare-LATEST.gguf \
--checkpoint-out chk-ol3b-shakespeare-ITERATION.gguf \
--lora-out lora-ol3b-shakespeare-ITERATION.bin \
--train-data "$LLAMA_TRAINING_DIR\shakespeare.txt" \
--save-every 10 \
--threads 10 --adam-iter 30 --batch 4 --ctx 64 \
--use-checkpointing