初始化项目，由ModelHub XC社区提供模型
Model: Jeesup/tofu_Llama-3.2-1B-Instruct_forget10_RMU_qat-int4 Source: Original Platform
This commit is contained in:
ModelHub XC
673
.hydra/config.yaml
Normal file
673
.hydra/config.yaml
Normal file
@@ -0,0 +1,673 @@
|
||||
model:
|
||||
model_args:
|
||||
pretrained_model_name_or_path: open-unlearning/tofu_Llama-3.2-1B-Instruct_full
|
||||
attn_implementation: sdpa
|
||||
torch_dtype: bfloat16
|
||||
tokenizer_args:
|
||||
pretrained_model_name_or_path: meta-llama/Llama-3.2-1B-Instruct
|
||||
template_args:
|
||||
apply_chat_template: true
|
||||
system_prompt: You are a helpful assistant.
|
||||
system_prompt_with_special_tokens: '<|begin_of_text|><|start_header_id|>system<|end_header_id|>
|
||||
|
||||
|
||||
You are a helpful assistant.<|eot_id|>'
|
||||
user_start_tag: '<|start_header_id|>user<|end_header_id|>
|
||||
|
||||
|
||||
'
|
||||
user_end_tag: <|eot_id|>
|
||||
asst_start_tag: '<|start_header_id|>assistant<|end_header_id|>
|
||||
|
||||
|
||||
'
|
||||
asst_end_tag: <|eot_id|>
|
||||
date_string: 10 Apr 2025
|
||||
trainer:
|
||||
handler: RMU
|
||||
args:
|
||||
per_device_train_batch_size: 4
|
||||
per_device_eval_batch_size: 16
|
||||
gradient_accumulation_steps: 4
|
||||
learning_rate: 1.0e-05
|
||||
bf16: true
|
||||
bf16_full_eval: true
|
||||
logging_steps: 5
|
||||
output_dir: ${paths.output_dir}
|
||||
logging_dir: ${trainer.args.output_dir}/logs
|
||||
report_to: tensorboard
|
||||
ddp_find_unused_parameters: true
|
||||
gradient_checkpointing: false
|
||||
optim: paged_adamw_32bit
|
||||
save_strategy: 'no'
|
||||
save_only_model: true
|
||||
weight_decay: 0.01
|
||||
do_train: true
|
||||
do_eval: false
|
||||
eval_on_start: true
|
||||
eval_strategy: epoch
|
||||
num_train_epochs: 10
|
||||
seed: 0
|
||||
warmup_epochs: 1.0
|
||||
remove_unused_columns: false
|
||||
push_to_hub: true
|
||||
hub_model_id: Jeesup/tofu_Llama-3.2-1B-Instruct_forget10_RMU_qat-int4
|
||||
hub_strategy: end
|
||||
hub_private_repo: false
|
||||
method_args:
|
||||
gamma: 1.0
|
||||
alpha: 1
|
||||
retain_loss_type: EMBED_DIFF
|
||||
steering_coeff: 2
|
||||
module_regex: model\.layers\.3
|
||||
trainable_params_regex:
|
||||
- .*
|
||||
data:
|
||||
forget:
|
||||
TOFU_QA_forget:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: ${forget_split}
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: question
|
||||
answer_key: answer
|
||||
max_length: 512
|
||||
retain:
|
||||
TOFU_QA_retain:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: ${retain_split}
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: question
|
||||
answer_key: answer
|
||||
max_length: 512
|
||||
anchor: forget
|
||||
collator:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
eval:
|
||||
tofu:
|
||||
metrics:
|
||||
forget_truth_ratio:
|
||||
pre_compute:
|
||||
forget_Q_A_PARA_Prob:
|
||||
datasets:
|
||||
TOFU_QA_forget_para:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: ${eval.tofu.forget_split}_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: ${eval.tofu.question_key}
|
||||
answer_key: paraphrased_answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: probability
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
access_key: correct
|
||||
forget_Q_A_PERT_Prob:
|
||||
datasets:
|
||||
TOFU_QA_forget_pert:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: ${eval.tofu.forget_split}_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: ${eval.tofu.question_key}
|
||||
answer_key: perturbed_answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: probability
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
access_key: wrong
|
||||
handler: truth_ratio
|
||||
aggregator: closer_to_1_better
|
||||
forget_quality:
|
||||
pre_compute:
|
||||
forget_truth_ratio:
|
||||
pre_compute:
|
||||
forget_Q_A_PARA_Prob:
|
||||
datasets:
|
||||
TOFU_QA_forget_para:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: ${eval.tofu.forget_split}_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: ${eval.tofu.question_key}
|
||||
answer_key: paraphrased_answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: probability
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
access_key: correct
|
||||
forget_Q_A_PERT_Prob:
|
||||
datasets:
|
||||
TOFU_QA_forget_pert:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: ${eval.tofu.forget_split}_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: ${eval.tofu.question_key}
|
||||
answer_key: perturbed_answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: probability
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
access_key: wrong
|
||||
handler: truth_ratio
|
||||
aggregator: closer_to_1_better
|
||||
access_key: forget
|
||||
reference_logs:
|
||||
retain_model_logs:
|
||||
path: ${eval.tofu.retain_logs_path}
|
||||
include:
|
||||
forget_truth_ratio:
|
||||
access_key: retain
|
||||
handler: ks_test
|
||||
forget_Q_A_Prob:
|
||||
datasets:
|
||||
TOFU_QA_forget:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: ${eval.tofu.forget_split}_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: ${eval.tofu.question_key}
|
||||
answer_key: answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: probability
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
forget_Q_A_ROUGE:
|
||||
datasets:
|
||||
TOFU_QA_forget:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: ${eval.tofu.forget_split}_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: ${eval.tofu.question_key}
|
||||
answer_key: answer
|
||||
max_length: 512
|
||||
predict_with_generate: true
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: left
|
||||
index: index
|
||||
generation_args:
|
||||
do_sample: false
|
||||
top_p: null
|
||||
temperature: null
|
||||
max_new_tokens: 200
|
||||
use_cache: true
|
||||
handler: rouge
|
||||
rouge_type: rougeL_recall
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
model_utility:
|
||||
pre_compute:
|
||||
retain_Q_A_Prob:
|
||||
datasets:
|
||||
TOFU_QA_retain_eval:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: retain_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: ${eval.tofu.question_key}
|
||||
answer_key: answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: probability
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
retain_Q_A_ROUGE:
|
||||
datasets:
|
||||
TOFU_QA_retain_eval:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: retain_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: ${eval.tofu.question_key}
|
||||
answer_key: answer
|
||||
max_length: 512
|
||||
predict_with_generate: true
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: left
|
||||
index: index
|
||||
generation_args:
|
||||
do_sample: false
|
||||
top_p: null
|
||||
temperature: null
|
||||
max_new_tokens: 200
|
||||
use_cache: true
|
||||
handler: rouge
|
||||
rouge_type: rougeL_recall
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
retain_Truth_Ratio:
|
||||
pre_compute:
|
||||
retain_Q_A_PARA_Prob:
|
||||
datasets:
|
||||
TOFU_QA_retain_para:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: retain_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: ${eval.tofu.question_key}
|
||||
answer_key: paraphrased_answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: probability
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
access_key: correct
|
||||
retain_Q_A_PERT_Prob:
|
||||
datasets:
|
||||
TOFU_QA_retain_pert:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: retain_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: ${eval.tofu.question_key}
|
||||
answer_key: perturbed_answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: probability
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
access_key: wrong
|
||||
handler: truth_ratio
|
||||
aggregator: true_better
|
||||
ra_Q_A_Prob_normalised:
|
||||
pre_compute:
|
||||
ra_Q_A_Prob:
|
||||
datasets:
|
||||
TOFU_QA_ra:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: real_authors_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: question
|
||||
answer_key: answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: probability
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
access_key: correct
|
||||
ra_Q_A_PERT_Prob:
|
||||
datasets:
|
||||
TOFU_QA_ra_pert:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: real_authors_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: question
|
||||
answer_key: perturbed_answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: probability
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
access_key: wrong
|
||||
handler: probability_w_options
|
||||
ra_Q_A_ROUGE:
|
||||
datasets:
|
||||
TOFU_QA_ra:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: real_authors_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: question
|
||||
answer_key: answer
|
||||
max_length: 512
|
||||
predict_with_generate: true
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: left
|
||||
index: index
|
||||
generation_args:
|
||||
do_sample: false
|
||||
top_p: null
|
||||
temperature: null
|
||||
max_new_tokens: 200
|
||||
use_cache: true
|
||||
handler: rouge
|
||||
rouge_type: rougeL_recall
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
ra_Truth_Ratio:
|
||||
pre_compute:
|
||||
ra_Q_A_Prob:
|
||||
datasets:
|
||||
TOFU_QA_ra:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: real_authors_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: question
|
||||
answer_key: answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: probability
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
access_key: correct
|
||||
ra_Q_A_PERT_Prob:
|
||||
datasets:
|
||||
TOFU_QA_ra_pert:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: real_authors_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: question
|
||||
answer_key: perturbed_answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: probability
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
access_key: wrong
|
||||
handler: truth_ratio
|
||||
aggregator: true_better
|
||||
wf_Q_A_Prob_normalised:
|
||||
pre_compute:
|
||||
wf_Q_A_Prob:
|
||||
datasets:
|
||||
TOFU_QA_wf:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: world_facts_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: question
|
||||
answer_key: answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: probability
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
access_key: correct
|
||||
wf_Q_A_PERT_Prob:
|
||||
datasets:
|
||||
TOFU_QA_wf_pert:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: world_facts_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: question
|
||||
answer_key: perturbed_answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: probability
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
access_key: wrong
|
||||
handler: probability_w_options
|
||||
wf_Q_A_ROUGE:
|
||||
datasets:
|
||||
TOFU_QA_wf:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: world_facts_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: question
|
||||
answer_key: answer
|
||||
max_length: 512
|
||||
predict_with_generate: true
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: left
|
||||
index: index
|
||||
generation_args:
|
||||
do_sample: false
|
||||
top_p: null
|
||||
temperature: null
|
||||
max_new_tokens: 200
|
||||
use_cache: true
|
||||
handler: rouge
|
||||
rouge_type: rougeL_recall
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
wf_Truth_Ratio:
|
||||
pre_compute:
|
||||
wf_Q_A_Prob:
|
||||
datasets:
|
||||
TOFU_QA_wf:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: world_facts_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: question
|
||||
answer_key: answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: probability
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
access_key: correct
|
||||
wf_Q_A_PERT_Prob:
|
||||
datasets:
|
||||
TOFU_QA_wf_pert:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: world_facts_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: question
|
||||
answer_key: perturbed_answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: probability
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
access_key: wrong
|
||||
handler: truth_ratio
|
||||
aggregator: true_better
|
||||
handler: hm_aggregate
|
||||
privleak:
|
||||
pre_compute:
|
||||
mia_min_k:
|
||||
datasets:
|
||||
TOFU_QA_forget:
|
||||
access_key: forget
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: ${eval.tofu.forget_split}_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: ${eval.tofu.question_key}
|
||||
answer_key: answer
|
||||
max_length: 512
|
||||
TOFU_QA_holdout:
|
||||
access_key: holdout
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: ${eval.tofu.holdout_split}
|
||||
path: locuslab/TOFU
|
||||
split: train
|
||||
question_key: question
|
||||
answer_key: answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
handler: mia_min_k
|
||||
k: 0.4
|
||||
access_key: forget
|
||||
reference_logs:
|
||||
retain_model_logs:
|
||||
path: ${eval.tofu.retain_logs_path}
|
||||
include:
|
||||
mia_min_k:
|
||||
access_key: retain
|
||||
handler: privleak
|
||||
ref_value: 0.5
|
||||
extraction_strength:
|
||||
datasets:
|
||||
TOFU_QA_forget:
|
||||
handler: QADataset
|
||||
args:
|
||||
hf_args:
|
||||
name: ${eval.tofu.forget_split}_perturbed
|
||||
split: train
|
||||
path: locuslab/TOFU
|
||||
question_key: ${eval.tofu.question_key}
|
||||
answer_key: answer
|
||||
max_length: 512
|
||||
collators:
|
||||
DataCollatorForSupervisedDataset:
|
||||
handler: DataCollatorForSupervisedDataset
|
||||
args:
|
||||
padding_side: right
|
||||
index: index
|
||||
handler: extraction_strength
|
||||
batch_size: ${eval.tofu.batch_size}
|
||||
handler: TOFUEvaluator
|
||||
output_dir: ${paths.output_dir}
|
||||
overwrite: true
|
||||
forget_split: ${forget_split}
|
||||
holdout_split: ${holdout_split}
|
||||
retain_logs_path: ${retain_logs_path}
|
||||
question_key: ${question_key}
|
||||
batch_size: 32
|
||||
paths:
|
||||
root_dir: .
|
||||
data_dir: ${paths.root_dir}/data/
|
||||
datasets: ${paths.root_dir}/configs/data/datasets
|
||||
output_dir: saves/unlearn/qat-baseline/tofu_Llama-3.2-1B-Instruct_forget10_RMU_qat-int4
|
||||
work_dir: ${hydra:runtime.cwd}
|
||||
forget_split: forget10
|
||||
retain_split: retain90
|
||||
holdout_split: holdout10
|
||||
retain_logs_path: null
|
||||
question_key: question
|
||||
task_name: tofu_Llama-3.2-1B-Instruct_forget10_RMU_qat-int4
|
||||
mode: unlearn
|
||||
quant:
|
||||
enabled: true
|
||||
scheme: int4_sym_per_channel
|
||||
n_bits: 4
|
||||
blocksize: 64
|
||||
target_module_regex: model\.layers\..*\.(self_attn|mlp)\..*proj$
|
||||
skip_module_regex: lm_head|embed_tokens
|
||||
295
.hydra/hydra.yaml
Normal file
295
.hydra/hydra.yaml
Normal file
@@ -0,0 +1,295 @@
|
||||
hydra:
|
||||
run:
|
||||
dir: ${paths.output_dir}
|
||||
sweep:
|
||||
dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
|
||||
subdir: ${hydra.job.num}
|
||||
launcher:
|
||||
_target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
|
||||
sweeper:
|
||||
_target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
|
||||
max_batch_size: null
|
||||
params: null
|
||||
help:
|
||||
app_name: ${hydra.job.name}
|
||||
header: '${hydra.help.app_name} is powered by Hydra.
|
||||
|
||||
'
|
||||
footer: 'Powered by Hydra (https://hydra.cc)
|
||||
|
||||
Use --hydra-help to view Hydra specific help
|
||||
|
||||
'
|
||||
template: '${hydra.help.header}
|
||||
|
||||
== Configuration groups ==
|
||||
|
||||
Compose your configuration from those groups (group=option)
|
||||
|
||||
|
||||
$APP_CONFIG_GROUPS
|
||||
|
||||
|
||||
== Config ==
|
||||
|
||||
Override anything in the config (foo.bar=value)
|
||||
|
||||
|
||||
$CONFIG
|
||||
|
||||
|
||||
${hydra.help.footer}
|
||||
|
||||
'
|
||||
hydra_help:
|
||||
template: 'Hydra (${hydra.runtime.version})
|
||||
|
||||
See https://hydra.cc for more info.
|
||||
|
||||
|
||||
== Flags ==
|
||||
|
||||
$FLAGS_HELP
|
||||
|
||||
|
||||
== Configuration groups ==
|
||||
|
||||
Compose your configuration from those groups (For example, append hydra/job_logging=disabled
|
||||
to command line)
|
||||
|
||||
|
||||
$HYDRA_CONFIG_GROUPS
|
||||
|
||||
|
||||
Use ''--cfg hydra'' to Show the Hydra config.
|
||||
|
||||
'
|
||||
hydra_help: ???
|
||||
hydra_logging:
|
||||
version: 1
|
||||
formatters:
|
||||
colorlog:
|
||||
(): colorlog.ColoredFormatter
|
||||
format: '[%(cyan)s%(asctime)s%(reset)s][%(purple)sHYDRA%(reset)s] %(message)s'
|
||||
handlers:
|
||||
console:
|
||||
class: logging.StreamHandler
|
||||
formatter: colorlog
|
||||
stream: ext://sys.stdout
|
||||
root:
|
||||
level: INFO
|
||||
handlers:
|
||||
- console
|
||||
disable_existing_loggers: false
|
||||
job_logging:
|
||||
version: 1
|
||||
formatters:
|
||||
simple:
|
||||
format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
|
||||
colorlog:
|
||||
(): colorlog.ColoredFormatter
|
||||
format: '[%(cyan)s%(asctime)s%(reset)s][%(blue)s%(name)s%(reset)s][%(log_color)s%(levelname)s%(reset)s]
|
||||
- %(message)s'
|
||||
log_colors:
|
||||
DEBUG: purple
|
||||
INFO: green
|
||||
WARNING: yellow
|
||||
ERROR: red
|
||||
CRITICAL: red
|
||||
handlers:
|
||||
console:
|
||||
class: logging.StreamHandler
|
||||
formatter: colorlog
|
||||
stream: ext://sys.stdout
|
||||
file:
|
||||
class: logging.FileHandler
|
||||
formatter: simple
|
||||
filename: ${hydra.runtime.output_dir}/${trainer.handler}.log
|
||||
root:
|
||||
level: INFO
|
||||
handlers:
|
||||
- console
|
||||
- file
|
||||
disable_existing_loggers: false
|
||||
env: {}
|
||||
mode: RUN
|
||||
searchpath: []
|
||||
callbacks: {}
|
||||
output_subdir: .hydra
|
||||
overrides:
|
||||
hydra:
|
||||
- hydra.mode=RUN
|
||||
task:
|
||||
- experiment=unlearn/tofu/default.yaml
|
||||
- trainer=RMU
|
||||
- task_name=tofu_Llama-3.2-1B-Instruct_forget10_RMU_qat-int4
|
||||
- model=Llama-3.2-1B-Instruct
|
||||
- forget_split=forget10
|
||||
- retain_split=retain90
|
||||
- holdout_split=holdout10
|
||||
- model.model_args.pretrained_model_name_or_path=open-unlearning/tofu_Llama-3.2-1B-Instruct_full
|
||||
- model.model_args.attn_implementation=sdpa
|
||||
- trainer.args.per_device_train_batch_size=4
|
||||
- trainer.args.gradient_accumulation_steps=4
|
||||
- trainer.args.ddp_find_unused_parameters=true
|
||||
- trainer.args.do_eval=false
|
||||
- trainer.args.save_strategy=no
|
||||
- +trainer.args.push_to_hub=true
|
||||
- +trainer.args.hub_model_id=Jeesup/tofu_Llama-3.2-1B-Instruct_forget10_RMU_qat-int4
|
||||
- +trainer.args.hub_strategy=end
|
||||
- +trainer.args.hub_private_repo=false
|
||||
- paths.output_dir=saves/unlearn/qat-baseline/tofu_Llama-3.2-1B-Instruct_forget10_RMU_qat-int4
|
||||
- +quant=qat_w4
|
||||
- trainer.method_args.module_regex=model\.layers\.3
|
||||
job:
|
||||
name: train
|
||||
chdir: null
|
||||
override_dirname: +quant=qat_w4,+trainer.args.hub_model_id=Jeesup/tofu_Llama-3.2-1B-Instruct_forget10_RMU_qat-int4,+trainer.args.hub_private_repo=false,+trainer.args.hub_strategy=end,+trainer.args.push_to_hub=true,experiment=unlearn/tofu/default.yaml,forget_split=forget10,holdout_split=holdout10,model.model_args.attn_implementation=sdpa,model.model_args.pretrained_model_name_or_path=open-unlearning/tofu_Llama-3.2-1B-Instruct_full,model=Llama-3.2-1B-Instruct,paths.output_dir=saves/unlearn/qat-baseline/tofu_Llama-3.2-1B-Instruct_forget10_RMU_qat-int4,retain_split=retain90,task_name=tofu_Llama-3.2-1B-Instruct_forget10_RMU_qat-int4,trainer.args.ddp_find_unused_parameters=true,trainer.args.do_eval=false,trainer.args.gradient_accumulation_steps=4,trainer.args.per_device_train_batch_size=4,trainer.args.save_strategy=no,trainer.method_args.module_regex=model\.layers\.3,trainer=RMU
|
||||
id: ???
|
||||
num: ???
|
||||
config_name: unlearn.yaml
|
||||
env_set: {}
|
||||
env_copy: []
|
||||
config:
|
||||
override_dirname:
|
||||
kv_sep: '='
|
||||
item_sep: ','
|
||||
exclude_keys: []
|
||||
runtime:
|
||||
version: 1.3.0
|
||||
version_base: '1.3'
|
||||
cwd: /home/yonsei_jong/open-unlearning
|
||||
config_sources:
|
||||
- path: hydra.conf
|
||||
schema: pkg
|
||||
provider: hydra
|
||||
- path: /home/yonsei_jong/open-unlearning/configs
|
||||
schema: file
|
||||
provider: main
|
||||
- path: hydra_plugins.hydra_colorlog.conf
|
||||
schema: pkg
|
||||
provider: hydra-colorlog
|
||||
- path: ''
|
||||
schema: structured
|
||||
provider: schema
|
||||
output_dir: /home/yonsei_jong/open-unlearning/saves/unlearn/qat-baseline/tofu_Llama-3.2-1B-Instruct_forget10_RMU_qat-int4
|
||||
choices:
|
||||
quant: qat_w4
|
||||
experiment: unlearn/tofu/default.yaml
|
||||
paths: default
|
||||
hydra: default
|
||||
eval: tofu
|
||||
eval/tofu_metrics/../../collator@eval.tofu.metrics.extraction_strength.collators: DataCollatorForSupervisedDatasetwithIndex
|
||||
eval/tofu_metrics/../../data/datasets@eval.tofu.metrics.extraction_strength.datasets: TOFU_QA_forget
|
||||
eval/tofu_metrics/.@eval.tofu.metrics.privleak.pre_compute.mia_min_k: mia_min_k
|
||||
eval/tofu_metrics/./../../collator@eval.tofu.metrics.privleak.pre_compute.mia_min_k.collators: DataCollatorForSupervisedDatasetwithIndex
|
||||
eval/tofu_metrics/./../../data/datasets@eval.tofu.metrics.privleak.pre_compute.mia_min_k.datasets: TOFU_MIA
|
||||
eval/tofu_metrics/.@eval.tofu.metrics.model_utility.pre_compute.wf_Truth_Ratio: wf_Truth_Ratio
|
||||
eval/tofu_metrics/./.@eval.tofu.metrics.model_utility.pre_compute.wf_Truth_Ratio.pre_compute.wf_Q_A_PERT_Prob: wf_Q_A_PERT_Prob
|
||||
? eval/tofu_metrics/././../../collator@eval.tofu.metrics.model_utility.pre_compute.wf_Truth_Ratio.pre_compute.wf_Q_A_PERT_Prob.collators
|
||||
: DataCollatorForSupervisedDatasetwithIndex
|
||||
? eval/tofu_metrics/././../../data/datasets@eval.tofu.metrics.model_utility.pre_compute.wf_Truth_Ratio.pre_compute.wf_Q_A_PERT_Prob.datasets
|
||||
: TOFU_QA_wf_pert
|
||||
eval/tofu_metrics/./.@eval.tofu.metrics.model_utility.pre_compute.wf_Truth_Ratio.pre_compute.wf_Q_A_Prob: wf_Q_A_Prob
|
||||
? eval/tofu_metrics/././../../collator@eval.tofu.metrics.model_utility.pre_compute.wf_Truth_Ratio.pre_compute.wf_Q_A_Prob.collators
|
||||
: DataCollatorForSupervisedDatasetwithIndex
|
||||
? eval/tofu_metrics/././../../data/datasets@eval.tofu.metrics.model_utility.pre_compute.wf_Truth_Ratio.pre_compute.wf_Q_A_Prob.datasets
|
||||
: TOFU_QA_wf
|
||||
eval/tofu_metrics/.@eval.tofu.metrics.model_utility.pre_compute.wf_Q_A_ROUGE: wf_Q_A_ROUGE
|
||||
eval/tofu_metrics/./../../generation@eval.tofu.metrics.model_utility.pre_compute.wf_Q_A_ROUGE.generation_args: default
|
||||
eval/tofu_metrics/./../../collator@eval.tofu.metrics.model_utility.pre_compute.wf_Q_A_ROUGE.collators: DataCollatorForSupervisedDatasetwithIndex
|
||||
eval/tofu_metrics/./../../data/datasets@eval.tofu.metrics.model_utility.pre_compute.wf_Q_A_ROUGE.datasets: TOFU_QA_wf
|
||||
eval/tofu_metrics/.@eval.tofu.metrics.model_utility.pre_compute.wf_Q_A_Prob_normalised: wf_Q_A_Prob_normalised
|
||||
eval/tofu_metrics/./.@eval.tofu.metrics.model_utility.pre_compute.wf_Q_A_Prob_normalised.pre_compute.wf_Q_A_PERT_Prob: wf_Q_A_PERT_Prob
|
||||
? eval/tofu_metrics/././../../collator@eval.tofu.metrics.model_utility.pre_compute.wf_Q_A_Prob_normalised.pre_compute.wf_Q_A_PERT_Prob.collators
|
||||
: DataCollatorForSupervisedDatasetwithIndex
|
||||
? eval/tofu_metrics/././../../data/datasets@eval.tofu.metrics.model_utility.pre_compute.wf_Q_A_Prob_normalised.pre_compute.wf_Q_A_PERT_Prob.datasets
|
||||
: TOFU_QA_wf_pert
|
||||
eval/tofu_metrics/./.@eval.tofu.metrics.model_utility.pre_compute.wf_Q_A_Prob_normalised.pre_compute.wf_Q_A_Prob: wf_Q_A_Prob
|
||||
? eval/tofu_metrics/././../../collator@eval.tofu.metrics.model_utility.pre_compute.wf_Q_A_Prob_normalised.pre_compute.wf_Q_A_Prob.collators
|
||||
: DataCollatorForSupervisedDatasetwithIndex
|
||||
? eval/tofu_metrics/././../../data/datasets@eval.tofu.metrics.model_utility.pre_compute.wf_Q_A_Prob_normalised.pre_compute.wf_Q_A_Prob.datasets
|
||||
: TOFU_QA_wf
|
||||
eval/tofu_metrics/.@eval.tofu.metrics.model_utility.pre_compute.ra_Truth_Ratio: ra_Truth_Ratio
|
||||
eval/tofu_metrics/./.@eval.tofu.metrics.model_utility.pre_compute.ra_Truth_Ratio.pre_compute.ra_Q_A_PERT_Prob: ra_Q_A_PERT_Prob
|
||||
? eval/tofu_metrics/././../../collator@eval.tofu.metrics.model_utility.pre_compute.ra_Truth_Ratio.pre_compute.ra_Q_A_PERT_Prob.collators
|
||||
: DataCollatorForSupervisedDatasetwithIndex
|
||||
? eval/tofu_metrics/././../../data/datasets@eval.tofu.metrics.model_utility.pre_compute.ra_Truth_Ratio.pre_compute.ra_Q_A_PERT_Prob.datasets
|
||||
: TOFU_QA_ra_pert
|
||||
eval/tofu_metrics/./.@eval.tofu.metrics.model_utility.pre_compute.ra_Truth_Ratio.pre_compute.ra_Q_A_Prob: ra_Q_A_Prob
|
||||
? eval/tofu_metrics/././../../collator@eval.tofu.metrics.model_utility.pre_compute.ra_Truth_Ratio.pre_compute.ra_Q_A_Prob.collators
|
||||
: DataCollatorForSupervisedDatasetwithIndex
|
||||
? eval/tofu_metrics/././../../data/datasets@eval.tofu.metrics.model_utility.pre_compute.ra_Truth_Ratio.pre_compute.ra_Q_A_Prob.datasets
|
||||
: TOFU_QA_ra
|
||||
eval/tofu_metrics/.@eval.tofu.metrics.model_utility.pre_compute.ra_Q_A_ROUGE: ra_Q_A_ROUGE
|
||||
eval/tofu_metrics/./../../generation@eval.tofu.metrics.model_utility.pre_compute.ra_Q_A_ROUGE.generation_args: default
|
||||
eval/tofu_metrics/./../../collator@eval.tofu.metrics.model_utility.pre_compute.ra_Q_A_ROUGE.collators: DataCollatorForSupervisedDatasetwithIndex
|
||||
eval/tofu_metrics/./../../data/datasets@eval.tofu.metrics.model_utility.pre_compute.ra_Q_A_ROUGE.datasets: TOFU_QA_ra
|
||||
eval/tofu_metrics/.@eval.tofu.metrics.model_utility.pre_compute.ra_Q_A_Prob_normalised: ra_Q_A_Prob_normalised
|
||||
eval/tofu_metrics/./.@eval.tofu.metrics.model_utility.pre_compute.ra_Q_A_Prob_normalised.pre_compute.ra_Q_A_PERT_Prob: ra_Q_A_PERT_Prob
|
||||
? eval/tofu_metrics/././../../collator@eval.tofu.metrics.model_utility.pre_compute.ra_Q_A_Prob_normalised.pre_compute.ra_Q_A_PERT_Prob.collators
|
||||
: DataCollatorForSupervisedDatasetwithIndex
|
||||
? eval/tofu_metrics/././../../data/datasets@eval.tofu.metrics.model_utility.pre_compute.ra_Q_A_Prob_normalised.pre_compute.ra_Q_A_PERT_Prob.datasets
|
||||
: TOFU_QA_ra_pert
|
||||
eval/tofu_metrics/./.@eval.tofu.metrics.model_utility.pre_compute.ra_Q_A_Prob_normalised.pre_compute.ra_Q_A_Prob: ra_Q_A_Prob
|
||||
? eval/tofu_metrics/././../../collator@eval.tofu.metrics.model_utility.pre_compute.ra_Q_A_Prob_normalised.pre_compute.ra_Q_A_Prob.collators
|
||||
: DataCollatorForSupervisedDatasetwithIndex
|
||||
? eval/tofu_metrics/././../../data/datasets@eval.tofu.metrics.model_utility.pre_compute.ra_Q_A_Prob_normalised.pre_compute.ra_Q_A_Prob.datasets
|
||||
: TOFU_QA_ra
|
||||
eval/tofu_metrics/.@eval.tofu.metrics.model_utility.pre_compute.retain_Truth_Ratio: retain_Truth_Ratio
|
||||
eval/tofu_metrics/./.@eval.tofu.metrics.model_utility.pre_compute.retain_Truth_Ratio.pre_compute.retain_Q_A_PERT_Prob: retain_Q_A_PERT_Prob
|
||||
? eval/tofu_metrics/././../../collator@eval.tofu.metrics.model_utility.pre_compute.retain_Truth_Ratio.pre_compute.retain_Q_A_PERT_Prob.collators
|
||||
: DataCollatorForSupervisedDatasetwithIndex
|
||||
? eval/tofu_metrics/././../../data/datasets@eval.tofu.metrics.model_utility.pre_compute.retain_Truth_Ratio.pre_compute.retain_Q_A_PERT_Prob.datasets
|
||||
: TOFU_QA_retain_pert
|
||||
eval/tofu_metrics/./.@eval.tofu.metrics.model_utility.pre_compute.retain_Truth_Ratio.pre_compute.retain_Q_A_PARA_Prob: retain_Q_A_PARA_Prob
|
||||
? eval/tofu_metrics/././../../collator@eval.tofu.metrics.model_utility.pre_compute.retain_Truth_Ratio.pre_compute.retain_Q_A_PARA_Prob.collators
|
||||
: DataCollatorForSupervisedDatasetwithIndex
|
||||
? eval/tofu_metrics/././../../data/datasets@eval.tofu.metrics.model_utility.pre_compute.retain_Truth_Ratio.pre_compute.retain_Q_A_PARA_Prob.datasets
|
||||
: TOFU_QA_retain_para
|
||||
eval/tofu_metrics/.@eval.tofu.metrics.model_utility.pre_compute.retain_Q_A_ROUGE: retain_Q_A_ROUGE
|
||||
eval/tofu_metrics/./../../generation@eval.tofu.metrics.model_utility.pre_compute.retain_Q_A_ROUGE.generation_args: default
|
||||
eval/tofu_metrics/./../../collator@eval.tofu.metrics.model_utility.pre_compute.retain_Q_A_ROUGE.collators: DataCollatorForSupervisedDatasetwithIndex
|
||||
eval/tofu_metrics/./../../data/datasets@eval.tofu.metrics.model_utility.pre_compute.retain_Q_A_ROUGE.datasets: TOFU_QA_retain_eval
|
||||
eval/tofu_metrics/.@eval.tofu.metrics.model_utility.pre_compute.retain_Q_A_Prob: retain_Q_A_Prob
|
||||
eval/tofu_metrics/./../../collator@eval.tofu.metrics.model_utility.pre_compute.retain_Q_A_Prob.collators: DataCollatorForSupervisedDatasetwithIndex
|
||||
eval/tofu_metrics/./../../data/datasets@eval.tofu.metrics.model_utility.pre_compute.retain_Q_A_Prob.datasets: TOFU_QA_retain_eval
|
||||
eval/tofu_metrics/../../generation@eval.tofu.metrics.forget_Q_A_ROUGE.generation_args: default
|
||||
eval/tofu_metrics/../../collator@eval.tofu.metrics.forget_Q_A_ROUGE.collators: DataCollatorForSupervisedDatasetwithIndex
|
||||
eval/tofu_metrics/../../data/datasets@eval.tofu.metrics.forget_Q_A_ROUGE.datasets: TOFU_QA_forget
|
||||
eval/tofu_metrics/../../collator@eval.tofu.metrics.forget_Q_A_Prob.collators: DataCollatorForSupervisedDatasetwithIndex
|
||||
eval/tofu_metrics/../../data/datasets@eval.tofu.metrics.forget_Q_A_Prob.datasets: TOFU_QA_forget
|
||||
eval/tofu_metrics/.@eval.tofu.metrics.forget_quality.pre_compute.forget_truth_ratio: forget_Truth_Ratio
|
||||
eval/tofu_metrics/./.@eval.tofu.metrics.forget_quality.pre_compute.forget_truth_ratio.pre_compute.forget_Q_A_PERT_Prob: forget_Q_A_PERT_Prob
|
||||
? eval/tofu_metrics/././../../collator@eval.tofu.metrics.forget_quality.pre_compute.forget_truth_ratio.pre_compute.forget_Q_A_PERT_Prob.collators
|
||||
: DataCollatorForSupervisedDatasetwithIndex
|
||||
? eval/tofu_metrics/././../../data/datasets@eval.tofu.metrics.forget_quality.pre_compute.forget_truth_ratio.pre_compute.forget_Q_A_PERT_Prob.datasets
|
||||
: TOFU_QA_forget_pert
|
||||
eval/tofu_metrics/./.@eval.tofu.metrics.forget_quality.pre_compute.forget_truth_ratio.pre_compute.forget_Q_A_PARA_Prob: forget_Q_A_PARA_Prob
|
||||
? eval/tofu_metrics/././../../collator@eval.tofu.metrics.forget_quality.pre_compute.forget_truth_ratio.pre_compute.forget_Q_A_PARA_Prob.collators
|
||||
: DataCollatorForSupervisedDatasetwithIndex
|
||||
? eval/tofu_metrics/././../../data/datasets@eval.tofu.metrics.forget_quality.pre_compute.forget_truth_ratio.pre_compute.forget_Q_A_PARA_Prob.datasets
|
||||
: TOFU_QA_forget_para
|
||||
eval/tofu_metrics/.@eval.tofu.metrics.forget_truth_ratio.pre_compute.forget_Q_A_PERT_Prob: forget_Q_A_PERT_Prob
|
||||
eval/tofu_metrics/./../../collator@eval.tofu.metrics.forget_truth_ratio.pre_compute.forget_Q_A_PERT_Prob.collators: DataCollatorForSupervisedDatasetwithIndex
|
||||
eval/tofu_metrics/./../../data/datasets@eval.tofu.metrics.forget_truth_ratio.pre_compute.forget_Q_A_PERT_Prob.datasets: TOFU_QA_forget_pert
|
||||
eval/tofu_metrics/.@eval.tofu.metrics.forget_truth_ratio.pre_compute.forget_Q_A_PARA_Prob: forget_Q_A_PARA_Prob
|
||||
eval/tofu_metrics/./../../collator@eval.tofu.metrics.forget_truth_ratio.pre_compute.forget_Q_A_PARA_Prob.collators: DataCollatorForSupervisedDatasetwithIndex
|
||||
eval/tofu_metrics/./../../data/datasets@eval.tofu.metrics.forget_truth_ratio.pre_compute.forget_Q_A_PARA_Prob.datasets: TOFU_QA_forget_para
|
||||
collator: DataCollatorForSupervisedDataset
|
||||
data: unlearn
|
||||
data/datasets@data.eval: null
|
||||
data/datasets@data.retain: TOFU_QA_retain
|
||||
data/datasets@data.forget: TOFU_QA_forget
|
||||
trainer: RMU
|
||||
model: Llama-3.2-1B-Instruct
|
||||
hydra/env: default
|
||||
hydra/callbacks: null
|
||||
hydra/job_logging: colorlog
|
||||
hydra/hydra_logging: colorlog
|
||||
hydra/hydra_help: default
|
||||
hydra/help: default
|
||||
hydra/sweeper: basic
|
||||
hydra/launcher: basic
|
||||
hydra/output: default
|
||||
verbose: false
|
||||
21
.hydra/overrides.yaml
Normal file
21
.hydra/overrides.yaml
Normal file
@@ -0,0 +1,21 @@
|
||||
- experiment=unlearn/tofu/default.yaml
|
||||
- trainer=RMU
|
||||
- task_name=tofu_Llama-3.2-1B-Instruct_forget10_RMU_qat-int4
|
||||
- model=Llama-3.2-1B-Instruct
|
||||
- forget_split=forget10
|
||||
- retain_split=retain90
|
||||
- holdout_split=holdout10
|
||||
- model.model_args.pretrained_model_name_or_path=open-unlearning/tofu_Llama-3.2-1B-Instruct_full
|
||||
- model.model_args.attn_implementation=sdpa
|
||||
- trainer.args.per_device_train_batch_size=4
|
||||
- trainer.args.gradient_accumulation_steps=4
|
||||
- trainer.args.ddp_find_unused_parameters=true
|
||||
- trainer.args.do_eval=false
|
||||
- trainer.args.save_strategy=no
|
||||
- +trainer.args.push_to_hub=true
|
||||
- +trainer.args.hub_model_id=Jeesup/tofu_Llama-3.2-1B-Instruct_forget10_RMU_qat-int4
|
||||
- +trainer.args.hub_strategy=end
|
||||
- +trainer.args.hub_private_repo=false
|
||||
- paths.output_dir=saves/unlearn/qat-baseline/tofu_Llama-3.2-1B-Instruct_forget10_RMU_qat-int4
|
||||
- +quant=qat_w4
|
||||
- trainer.method_args.module_regex=model\.layers\.3
|
||||
Reference in New Issue
Block a user