per layer lightweightlora

scripts/gemma_rank_exp/gemma_self_gen_qa_rank_16.sh

@@ -0,0 +1,24 @@
+#!/bin/bash
+#SBATCH --job-name=ctxlora_medium
+#SBATCH --partition=a3
+#SBATCH --nodes=1
+#SBATCH --gpus=8
+#SBATCH --output=outputs/%x-%j.out
+#SBATCH --error=outputs/%x-%j.out
+
+uv run accelerate launch --num_processes=8 --gradient_accumulation_steps=16 --gradient_clipping=1.0 \
+--gpu_ids all --main_process_port 29561 intx_sft.py configs/self_gen_3_small.yaml \
+--model_name_or_path=google/gemma-2-2b-it --num_train_epochs=2.2 --per_device_train_batch_size=-1 \
+--gradient_accumulation_steps=16 --per_device_eval_batch_size=4 --exp_setup=hyper_lora --aggregator_type=perceiver \
+--target_modules=down_proj \
+--num_self_attends_per_block=8 --num_latent_factor=1 \
+--lora_r=16 \
+--eval_steps=1000000000000000 --do_eval=no --do_predict=no --eval_on_start=False \
+--save_steps=5000 --learning_rate=4e-5 --lora_dropout=0.0 \
+--neftune_noise_alpha=5 --use_light_weight_lora=False \
+--add_negative_prompt=False \
+--add_repeat_prompt=False \
+--use_sequence_packing=True --max_packed_inp_len=4096 --max_packed_ctx_len=8192 \
+--per_rank_gen=True \
+--per_layer_processing=True \
+--gen_lora_l1_reg_coef=0.1 \

scripts/gemma_rank_exp/gemma_self_gen_qa_rank_16_lightweight.sh

@@ -0,0 +1,26 @@
+#!/bin/bash
+#SBATCH --job-name=ctxlora_medium
+#SBATCH --partition=a3
+#SBATCH --nodes=1
+#SBATCH --gpus=4
+#SBATCH --output=outputs/%x-%j.out
+#SBATCH --error=outputs/%x-%j.out
+
+uv run accelerate launch --num_processes=4 --gradient_accumulation_steps=32 --gradient_clipping=1.0 \
+--gpu_ids all --main_process_port 29563 intx_sft.py configs/self_gen_3_small.yaml \
+--model_name_or_path=google/gemma-2-2b-it --num_train_epochs=1 --per_device_train_batch_size=-1 \
+--gradient_accumulation_steps=32 --per_device_eval_batch_size=4 --exp_setup=hyper_lora --aggregator_type=perceiver \
+--target_modules=down_proj \
+--num_self_attends_per_block=8 --num_latent_factor=1 \
+--lora_r=16 \
+--eval_steps=1000000000000000 --do_eval=no --do_predict=no --eval_on_start=False \
+--save_steps=5000 --learning_rate=4e-5 --lora_dropout=0.0 \
+--num_pre_head_layers=1 \
+--latent_size=256 \
+--neftune_noise_alpha=5 --use_light_weight_lora=True --light_weight_latent_size=512 \
+--add_negative_prompt=False \
+--add_repeat_prompt=False \
+--use_sequence_packing=True --max_packed_inp_len=4096 --max_packed_ctx_len=8192 \
+--per_rank_gen=True \
+--per_layer_processing=True \
+--gen_lora_l1_reg_coef=0.1 \

scripts/gemma_rank_exp/gemma_self_gen_qa_repeat_rank_16.sh

@@ -0,0 +1,24 @@
+#!/bin/bash
+#SBATCH --job-name=ctxlora_medium
+#SBATCH --partition=a3
+#SBATCH --nodes=1
+#SBATCH --gpus=8
+#SBATCH --output=outputs/%x-%j.out
+#SBATCH --error=outputs/%x-%j.out
+
+uv run accelerate launch --num_processes=8 --gradient_accumulation_steps=16 --gradient_clipping=1.0 \
+--gpu_ids all --main_process_port 29562 intx_sft.py configs/self_gen_3_small.yaml \
+--model_name_or_path=google/gemma-2-2b-it --num_train_epochs=2.2 --per_device_train_batch_size=-1 \
+--gradient_accumulation_steps=16 --per_device_eval_batch_size=4 --exp_setup=hyper_lora --aggregator_type=perceiver \
+--target_modules=down_proj \
+--num_self_attends_per_block=8 --num_latent_factor=1 \
+--lora_r=16 \
+--eval_steps=1000000000000000 --do_eval=no --do_predict=no --eval_on_start=False \
+--save_steps=5000 --learning_rate=4e-5 --lora_dropout=0.0 \
+--neftune_noise_alpha=5 --use_light_weight_lora=False \
+--add_negative_prompt=False \
+--add_repeat_prompt=True --repeat_prob=0.1 \
+--use_sequence_packing=True --max_packed_inp_len=4096 --max_packed_ctx_len=8192 \
+--per_rank_gen=True \
+--per_layer_processing=True \
+--gen_lora_l1_reg_coef=0.1 \

scripts/gemma_rank_exp/gemma_self_gen_qa_repeat_rank_16_lightweight.sh

@@ -0,0 +1,26 @@
+#!/bin/bash
+#SBATCH --job-name=ctxlora_medium
+#SBATCH --partition=a3
+#SBATCH --nodes=1
+#SBATCH --gpus=4
+#SBATCH --output=outputs/%x-%j.out
+#SBATCH --error=outputs/%x-%j.out
+
+uv run accelerate launch --num_processes=4 --gradient_accumulation_steps=32 --gradient_clipping=1.0 \
+--gpu_ids all --main_process_port 29564 intx_sft.py configs/self_gen_3_small.yaml \
+--model_name_or_path=google/gemma-2-2b-it --num_train_epochs=1 --per_device_train_batch_size=-1 \
+--gradient_accumulation_steps=32 --per_device_eval_batch_size=4 --exp_setup=hyper_lora --aggregator_type=perceiver \
+--target_modules=down_proj \
+--num_self_attends_per_block=8 --num_latent_factor=1 \
+--lora_r=16 \
+--eval_steps=1000000000000000 --do_eval=no --do_predict=no --eval_on_start=False \
+--save_steps=5000 --learning_rate=4e-5 --lora_dropout=0.0 \
+--num_pre_head_layers=1 \
+--latent_size=256 \
+--neftune_noise_alpha=5 --use_light_weight_lora=True --light_weight_latent_size=512 \
+--add_negative_prompt=False \
+--add_repeat_prompt=True --repeat_prob=0.1 \
+--use_sequence_packing=True --max_packed_inp_len=4096 --max_packed_ctx_len=8192 \
+--per_rank_gen=True \
+--per_layer_processing=True \
+--gen_lora_l1_reg_coef=0.1 \