Name 'LearnedSourceLowRankRotatedSpaceIntervention' is not defined in tutorial notebook

[Z1zs]

I tied to run this script(https://github.com/stanfordnlp/pyreft/blob/main/examples/memorisation/reft_power.ipynb) to figure out how reft should work, but got this error in "### Store with different access id":

TARGET_LAYER = 15

alice_access_id = "ALIC#ID1->"
model_max_length = 2048

# get tokenizer
tokenizer = transformers.AutoTokenizer.from_pretrained(
    model_name_or_path, model_max_length=model_max_length, 
    padding_side="right", use_fast=False)
tokenizer.pad_token = tokenizer.unk_token

# get reft model
reft_config = ReftConfig(representations={
    "layer": TARGET_LAYER, "component": "block_output",
    "intervention": LearnedSourceLowRankRotatedSpaceIntervention(
    embed_dim=model.config.hidden_size, 
    low_rank_dimension=1)})
reft_model = get_reft_model(model, reft_config)
reft_model.print_trainable_parameters()

# get training data and args
data_module = make_supervised_data_module(
    tokenizer, model, 
    [storage_access_id, alice_access_id], [memo_sequence, alice_slice])
training_args = transformers.TrainingArguments(output_dir="./tmp")
training_args.save_strategy = "no"
training_args.evaluation_strategy = "no"
training_args.num_train_epochs = 500.0
training_args.learning_rate = 8e-3
training_args.per_device_train_batch_size = 16
training_args.report_to = []
training_args.logging_steps = 100

# train
trainer = ReftTrainerForCausalLM(
    model=reft_model, tokenizer=tokenizer, args=training_args, **data_module)
_ = trainer.train()

---

NameError Traceback (most recent call last)
Cell In[28], line 15
10 tokenizer.pad_token = tokenizer.unk_token
12 # get reft model
13 reft_config = ReftConfig(representations={
14 "layer": TARGET_LAYER, "component": "block_output",
---> 15 "intervention": LearnedSourceLowRankRotatedSpaceIntervention(
16 embed_dim=model.config.hidden_size,
17 low_rank_dimension=1)})
18 reft_model = get_reft_model(model, reft_config)
19 reft_model.print_trainable_parameters()

NameError: name 'LearnedSourceLowRankRotatedSpaceIntervention' is not defined

I can't find or import 'LearnedSourceLowRankRotatedSpaceIntervention' in pyreft.
Should I replace the name 'LearnedSourceLowRankRotatedSpaceIntervention' with 'ConsreftIntervention'? What is the difference between these Intervention class?
(minor: 'storage_access_id' can also be not defined in this block.)

[frankaging]

@Z1zs Thanks for reporting the issue! I fixed the notebook. Please pull and retry.

And, yes - we renamed LearnedSourceLowRankRotatedSpaceIntervention to ConsreftIntervention (if you are curious, you could take a look at our commit history, etc..). I also corrected the "storage_access_id" issue. Some of our notebooks are made before we did the major refactory, so they might be broken.