apunkt/doc-to-lora
1
0
Fork 0
mirror of https://github.com/SakanaAI/doc-to-lora.git synced 2026-07-23 17:01:04 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

per-layer processing working + linear proj before mlp

Browse source
This commit is contained in:
51616 2025-05-07 15:12:02 +00:00
parent c5cae72d1d
commit 767dc634b9
3 changed files with 44 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

1
scripts/ablate_arch/gemma_xl_per_rank_gen_fac2.sh
View file

@ -2,6 +2,7 @@
#SBATCH --job-name=ctxlora
#SBATCH --partition=a3
#SBATCH --nodes=1
#SBATCH --exclude=slurm0-a3nodeset-2
#SBATCH --gpus=4
#SBATCH --output=outputs/%x-%j.out
#SBATCH --error=outputs/%x-%j.out

34
scripts/ablate_arch/gemma_xl_per_rank_gen_fac2_per_layer_processing.sh Normal file
View file

@ -0,0 +1,34 @@
#!/bin/bash
#SBATCH --job-name=ctxlora
#SBATCH --partition=a3
#SBATCH --nodes=1
#SBATCH --exclude=slurm0-a3nodeset-2
#SBATCH --gpus=4
#SBATCH --output=outputs/%x-%j.out
#SBATCH --error=outputs/%x-%j.out
# module load
# module load cuda/12.1
# module load cudnn/8.9.7
# module load nccl/cuda-12.1/2.18.3
# module load hpcx/2.20
# export OMP_NUM_THREADS=24
# export TRITON_CACHE_DIR=/tmp/.triton/
. ~/miniconda3/etc/profile.d/conda.sh
conda activate /home/rujikorn_sakana_ai/.conda/envs/ctx-to-lora
# eval "$@"
accelerate launch --num_processes=4 --gradient_accumulation_steps=8 --gradient_clipping=1.0 \
--gpu_ids all --main_process_port 29562 intx_sft.py configs/pretrain_all_xl.yaml \
--model_name_or_path=google/gemma-2-2b-it --num_train_epochs=5.1 --per_device_train_batch_size=32 \
--gradient_accumulation_steps=8 --per_device_eval_batch_size=32 --exp_setup=hyper_lora --aggregator_type=perceiver \
--target_modules=down_proj \
--num_self_attends_per_block=8 --num_latent_factor=2 \
--lora_r=8 \
--eval_steps=5000 --save_steps=5000 --learning_rate=2e-5 --lora_dropout=0.0 \
--neftune_noise_alpha=5 --use_light_weight_lora=False \
--load_best_model_at_end=True --metric_for_best_model=pwc_loss --add_negative_prompt=False \
--add_repeat_prompt=False \
--use_sequence_packing=True --per_rank_gen=True \
--per_layer_processing=True

17
src/ctx_to_lora/modeling_utils.py
View file

@ -756,14 +756,14 @@ class HyperLoRA(nn.Module):
if self.config.per_layer_processing:
layers = [
# nn.Linear(self.d_latent, self.d_latent),
# Mix(
# "bs n_layers n_modules r d -> bs n_layers n_modules r d0",
# weight_shape="n_layers d d0",
# bias_shape="n_layers d0",
# n_layers=self.n_layers,
# d=self.d_latent,
# d0=self.d_latent,
# ),
Mix(
"bs n_layers n_modules r d -> bs n_layers n_modules r d0",
weight_shape="n_layers d d0",
bias_shape="n_layers d0",
n_layers=self.n_layers,
d=self.d_latent,
d0=self.d_latent,
),
ResMLPBlockPerLayer(
self.n_layers,
self.d_latent,
@ -776,6 +776,7 @@ class HyperLoRA(nn.Module):
]
else:
layers = [
nn.Linear(self.d_latent, self.d_latent),
MLPResidualBlock(
input_size=self.config.latent_size,
hidden_size=self.config.latent_size * 4,