SAEmnesia_private/sbatch_main_pipeline.sh at master · EIDOSLAB/SAEmnesia_private · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
#!/bin/bash
#SBATCH --job-name=sdxl_activations_collection
#SBATCH --output=sbatch_output/%j_sdxl_activations_collection.out
#SBATCH --error=sbatch_output/%j_sdxl_activations_collection.err
#SBATCH --time=24:00:00              # Increased time limit
#SBATCH --mem=380G                   # Increased memory
#SBATCH --partition=boost_usr_prod   # Ensure this is your highest-resource partition
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=1
#SBATCH --gres=gpu:4
#SBATCH --cpus-per-task=8           # Increased CPU cores per task
#SBATCH --account=IscrC_INSAIT

# Load any necessary GPU modules (system-specific)
# module load cuda

source ../../envs/saeuron_cassano/bin/activate

# Set PyTorch memory configuration
export PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True

echo "CUDA devices: $CUDA_VISIBLE_DEVICES"
nvidia-smi

## OBJECTS: unet.up_blocks.1.attentions.1
## STYLES: unet.up_blocks.1.attentions.2

# Step 1

# python /leonardo/home/userexternal/ecassano/projects/SAeUron/scripts/load_from_hub.py \
# --name bcywinski/SAeUron \
# --hookpoint unet.up_blocks.1.attentions.1 \
# --save_path /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sae_checkpoints/


# Step 2

# echo "Running FULL PRODUCTION step 2."
# python /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/collect_activations_unlearn_canvas.py \
# --mode dual_object_metadata \
# --hook_names unet.up_blocks.1.attentions.1 \
# --model_name /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50 \
# --new_cached_activations_path /leonardo_work/IscrC_SAOU/cassano/finetuning_activations/dual_objects \
# --batch_size 128 \
# --class_start 0 \
# --class_end 20
# echo "Completed FULL PRODUCTION step 2."

# python /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/styles_metadata_recovery.py
#

echo "Running SDXL-Turbo activation collection"
accelerate launch --num_processes 4 /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/collect_activations_unlearn_canvas.py \
--mode finetuning \
--hook_names unet.up_blocks.0.attentions.1 \
--model_name /leonardo_scratch/fast/IscrC_INSAIT/sdxl-turbo \
--new_cached_activations_path /leonardo_scratch/fast/IscrC_INSAIT/activations/sdxl \
--batch_size 64 \
--class_start 0 \
--class_end 20 \
--organization_type object

# python /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/collect_activations_unlearn_canvas.py \
# --hook_names unet.up_blocks.1.attentions.1 \
# --model_name /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50 \
# --new_cached_activations_path /leonardo_work/IscrC_MAGNIFY/tmp \
# --batch_size 128
#

# accelerate launch --num-processes=2 /leonardo/home/userexternal/ecassano/projects/SAeUron/scripts/collect_activations_unlearn_canvas.py \
# --hook_names unet.up_blocks.1.attentions.1 \
# --model_name /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50 \
# --new_cached_activations_path /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/cached_activations \
# --batch_size 128

# echo "Generating dataset for styles finetuning."
# accelerate launch --num-processes=4 /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/collect_activations_unlearn_canvas.py \
# --mode metadata \
# --hook_names unet.up_blocks.1.attentions.2 \
# --model_name /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50 \
# --new_cached_activations_path /leonardo_work/IscrC_INSAIT/styles_finetuning_dataset \
# --batch_size 128 \
# --class_start 0 \
# --class_end 50
#
# echo "Dataset collected."

# echo "Generating dataset for nudity finetuning."
# accelerate launch --num-processes=2 /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/collect_activations_nudity.py \
# --hook_names unet.up_blocks.1.attentions.1 \
# --model_name /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sd1.4 \
# --new_cached_activations_path /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/nudity_dataset/activations \
# --prompts_path /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/nudity_dataset/annotations/coco_30k_captions.txt \
# --batch_size 128 \
# --num_inference_steps 50 \
# --cache_every_n_timesteps 1 \
# --column caption \
# --output_or_diff output \
# --seed 42
#
# echo "Dataset collected."

# Step 4 - Nudity

# echo "Running step 4 for Nudity."
# python /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/nudity_efficient_gather_sae_acts_ca_prompts_cls.py \
# --checkpoint_path "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sae_checkpoints/nudity" \
# --hookpoint "unet.up_blocks.1.attentions.1" \
# --pipe_path "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sd1.4" \
# --save_dir "/leonardo_scratch/fast/IscrC_SAOU/nudity_activations" \
# --nudity_prompts_file "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/nudity_dataset/annotations/nudity_prompts.txt" \
# --non_nudity_prompts_file "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/nudity_dataset/annotations/non_nudity_prompts.txt" \
# --batch_size 128
# echo "Step 4 for nudity completed."
# echo "Saving scores for nudity."
# python scripts/nudity_save_scores.py \
# --model_checkpoint /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sae_checkpoints/nudity/unet.up_blocks.1.attentions.1 \
# --latents_path /leonardo_scratch/fast/IscrC_SAOU/nudity_activations/nudity_latents_dict_unet.up_blocks.1.attentions.1.pkl \
# --num_timesteps 50 \
# --output_json /leonardo_scratch/fast/IscrC_SAOU/nudity_scores.json \
# --plot_scores \
# --plot_output_dir /leonardo_scratch/fast/IscrC_SAOU/nudity_plots
# echo "Scores saved for nudity."

# Step 4 - Object (with finetuned model)

# echo "Running step 4."
#
# python /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/efficient_gather_sae_acts_ca_prompts_cls.py \
# --checkpoint_path "/leonardo_work/IscrC_SAOU/cassano/saeuron/sae_checkpoints/dual_concept_optimized/sdxl-turbo/v1.6/ce_weight_3.0_sparsity_0.01/best/" \
# --hookpoint "unet.up_blocks.0.attentions.1" \
# --pipe_path "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50" \
# --save_dir "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/features_activations/finetuned/sdxl-turbo/unet.up_blocks.0.attentions.1"
#
# echo "Step 4 completed."

# Step 4 - Object

# python /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/efficient_gather_sae_acts_ca_prompts_cls.py \
# --checkpoint_path "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sae_checkpoints" \
# --hookpoint "unet.up_blocks.1.attentions.1" \
# --pipe_path "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50" \
# --save_dir "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/features_activations/non_finetuned/unet.up_blocks.1.attentions.1"

# Step 4 - Object (with finetuned model)

# echo "Running step 4."
#
# python /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/efficient_gather_sae_acts_ca_prompts_cls.py \
# --checkpoint_path "/leonardo_scratch/large/userexternal/ecassano/sae_checkpoints/dual_concept_optimized/g_sae/ce_weight_1.0/best" \
# --hookpoint "unet.up_blocks.1.attentions.1" \
# --pipe_path "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50" \
# --save_dir "/leonardo_scratch/large/userexternal/ecassano/saeuron/features_activations/finetuned/gsae/unet.up_blocks.1.attentions.1"
#
# echo "Step 4 completed."

#
#
# echo "Saving scores for objects."
#
# python scripts/save_scores.py \
# --model_checkpoint /leonardo_work/IscrC_SAOU/cassano/saeuron/sae_checkpoints/dual_concept_optimized/sdxl-turbo/v1.6/ce_weight_3.0_sparsity_0.01/best/unet.up_blocks.0.attentions.1 \
# --latents_path /leonardo_work/IscrC_SAOU/cassano/saeuron/features_activations/finetuned/sdxl-turbo/unet.up_blocks.0.attentions.1/cls_latents_dict_unet.up_blocks.0.attentions.1.pkl \
# --concept_type "objects" \
# --num_timesteps 100 \
# --output_json /leonardo_work/IscrC_SAOU/cassano/saeuron/scores/objects/finetuned/sdxl-turbo/v1.6/scores.json \
# --plot_scores  \
# --plot_output_dir /leonardo_work/IscrC_SAOU/cassano/saeuron/scores/objects/finetuned/sdxl-turbo/v1.6
#
# echo "Scores saved for objects."

# echo "Styles Finetuning - Running step 4."
#
# python /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/efficient_gather_sae_acts_ca_prompts_cls.py \
# --checkpoint_path "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sae_checkpoints/best" \
# --hookpoint "unet.up_blocks.1.attentions.2" \
# --pipe_path "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50" \
# --save_dir "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/features_activations/unet.up_blocks.1.attentions.2"
#
# echo "Step 4 completed."
#
# echo "Saving scores for objects."
#
# python scripts/save_scores.py \
# --model_checkpoint /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sae_checkpoints/best/unet.up_blocks.1.attentions.2 \
# --latents_path /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/features_activations/unet.up_blocks.1.attentions.2/cls_latents_dict_unet.up_blocks.1.attentions.2.pkl \
# --concept_type "objects" \
# --num_timesteps 100 \
# --output_json /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/scores/objects/up_1_2/scores.json \
# --plot_scores  \
# --plot_output_dir /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/scores/objects/up_1_2
#
# echo "Scores saved for objects."


# echo "Running step 4 for styles."
#
# python scripts/efficient_gather_sae_acts_ca_prompts.py \
# --checkpoint_path "/leonardo_work/IscrC_SAOU/cassano/saeuron/sae_checkpoints/dual_concept_optimized/sdxl-turbo/v1.6/ce_weight_3.0_sparsity_0.01/best/" \
# --hookpoint "unet.up_blocks.0.attentions.1" \
# --pipe_path "/leonardo_work/IscrC_SAOU/sdxl-turbo" \
# --save_dir "/leonardo_work/IscrC_SAOU/cassano/saeuron/features_activations/finetuned/sdxl-turbo/unet.up_blocks.0.attentions.1"
#
# echo "Step 4 completed."
#
# echo "Saving scores for styles."
#
# python scripts/save_scores.py \
# --model_checkpoint /leonardo_work/IscrC_SAOU/cassano/saeuron/sae_checkpoints/dual_concept_optimized/sdxl-turbo/v1.6/ce_weight_3.0_sparsity_0.01/best/unet.up_blocks.0.attentions.1 \
# --latents_path /leonardo_work/IscrC_SAOU/cassano/saeuron/features_activations/finetuned/sdxl-turbo/unet.up_blocks.0.attentions.1/cls_latents_dict_unet.up_blocks.0.attentions.1.pkl \
# --concept_type "styles" \
# --num_timesteps 100 \
# --output_json /leonardo_work/IscrC_SAOU/cassano/saeuron/scores/styles/finetuned/sdxl-turbo/v1.6/scores.json \
# --plot_scores  \
# --plot_output_dir /leonardo_work/IscrC_SAOU/cassano/saeuron/scores/styles/finetuned/sdxl-turbo/v1.6
#
# echo "Scores saved for styles."


# Step 5.0 - Hyperparameter Sweep for Object Unlearning
# original_multipliers = [-1.0, -5.0, -10.0, -15.0, -20.0, -25.0, -30.0]
# original_percentiles = [99.99, 99.995, 99.999]


# echo "Running step 5.0, phase 1 - Hyperparameter Sweep for Object Unlearning"
#
# accelerate launch --num_processes 4 /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/gsae_steering_efficient_sweep_cls_distr.py \
# --alphas [-5.0,-4.0,-3.0,-2.0,-1.5] --seed 188 \
# --percentiles [99.999] \
# --output_dir '/leonardo_scratch/large/userexternal/ecassano/saeuron/sweep_outputs/objects/fine_tuned/g_sae/gsae_steering/hp_search/seed_188' \
# --pipe_checkpoint '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50' \
# --hookpoint 'unet.up_blocks.1.attentions.1' \
# --class_latents_path '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/features_activations/finetuned/v1.6/unet.up_blocks.1.attentions.1/cls_latents_dict_unet.up_blocks.1.attentions.1.pkl' \
# --sae_checkpoint '/leonardo_scratch/large/userexternal/ecassano/sae_checkpoints/dual_concept_optimized/g_sae/ce_weight_1.0/best/' \
# --steps 100
#
# echo "Phase 1 of step 5.0 completed."
#
# echo "Running step 5.0, phase 2 - Hyperparameter Sweep for Object Unlearning"
#
# python /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/gsae_steering_run_acc_all_cls_sweep.py \
# --percentiles [99.999] \
# --alphas [-5.0,-4.0,-3.0,-2.0,-1.5] \
# --input_dir_base /leonardo_scratch/large/userexternal/ecassano/saeuron/sweep_outputs/objects/fine_tuned/g_sae/gsae_steering/hp_search/seed_188 \
# --output_dir_base /leonardo_scratch/large/userexternal/ecassano/saeuron/sweep_outputs/objects/fine_tuned/g_sae/gsae_steering/hp_search/seed_188 \
# --class_ckpt /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/classifier_checkpoints/cls_model/style50_cls.pth \
# --batch_size 256 \
# --seed 188

# # --script_version v2
#
# echo "Phase 2 of step 5.0 completed."
#
#
# echo "Running step 5.0, phase 3 - Hyperparameter Sweep for Object Unlearning"
#
# python /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/gsae_steering_find_best_params_cls_sweep.py [99.999] [-5.0,-4.0,-3.0,-2.0,-1.5] "/leonardo_scratch/large/userexternal/ecassano/saeuron/sweep_outputs/objects/fine_tuned/g_sae/gsae_steering/hp_search/seed_188"
#
# echo "Phase 3 of step 5.0 completed."

# echo "Running step 5.0, phase 3 - Hyperparameter Sweep for Object Unlearning"
#
# python /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/find_best_params_cls_sweep_v2.py [99.999] [-1.0,-5.0,-10.0,-15.0,-20.0,-25.0,-30.0] "/leonardo_work/IscrC_SAOU/cassano/saeuron/sweep_outputs/objects/fine_tuned/v1.6/improved_m_selection/hp_search/seed_188" \
# --selection_strategy weighted_average \
# --ua_weight 1.0 \
# --ira_weight 1.0 \
# --ccm_weight 1.0
#
# echo "Phase 3 of step 5.0 completed."
# Step 5.1 - Object Unlearning

# echo "Running step 5"

# --class_params_path  /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sweep_outputs/objects/fine_tuned/v1.6/hp_search/seed_188/class_params.pth \

# accelerate launch --num_processes 4 /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/sample_unlearning_cls_distr.py  \
# --seed 188 \
# --output_dir '/leonardo_work/IscrC_SAOU/cassano/saeuron/sweep_outputs/objects/fine_tuned/v1.6/sdxl-turbo/hp_search/seed_188' \
# --class_params_path  /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sweep_outputs/objects/baseline/class_params_-1.0.pth \
# --pipe_checkpoint '/leonardo_work/IscrC_SAOU/sdxl-turbo' \
# --hookpoint 'unet.up_blocks.0.attentions.1' \
# --class_latents_path '/leonardo_work/IscrC_SAOU/cassano/saeuron/features_activations/finetuned/sdxl-turbo/unet.up_blocks.0.attentions.1/cls_latents_dict_unet.up_blocks.0.attentions.1.pkl' \
# --sae_checkpoint '/leonardo_work/IscrC_SAOU/cassano/saeuron/sae_checkpoints/dual_concept_optimized/sdxl-turbo/v1.6/ce_weight_3.0_sparsity_0.01/best' \
# --steps 100


# Images generation with patch replacement for unlearning.

# accelerate launch --num_processes 4 /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/noise_injection_sample_unlearning_cls_distr.py  \
# --seed 188 \
# --output_dir '/leonardo_scratch/large/userexternal/ecassano/saeuron/sweep_outputs/objects/fine_tuned/v1.6/noise_injection/seed_188/replace_with_neighbor_padded' \
# --pipe_checkpoint '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50' \
# --hookpoint 'unet.up_blocks.1.attentions.1' \
# --class_latents_path '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/features_activations/finetuned/v1.6/unet.up_blocks.1.attentions.1/cls_latents_dict_unet.up_blocks.1.attentions.1.pkl' \
# --sae_checkpoint '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sae_checkpoints/wrong/dual_concept_optimized/v1.6/ce_weight_3.0_sparsity_0.01/best' \
# --steps 100 \
# --noise_scale 10.0 \
# --noise_type gaussian \
# --use_sae=True \
# --noise_mode='replace_with_neighbor_padded' \
# --padding=1 \
# --start_timestep=99 \
# --end_timestep=0 \
# --top_k=1

# G-SAE
# python /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/gsae_steering.py  \
# --output_dir '/leonardo_work/IscrC_SAOU/cassano/saeuron/sweep_outputs/objects/fine_tuned/gsae' \
# --pipe_checkpoint '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50' \
# --hookpoint 'unet.up_blocks.1.attentions.1' \
# --sae_checkpoint '/leonardo_work/IscrC_SAOU/cassano/saeuron/sae_checkpoints/dual_concept_optimized/g_sae/ce_weight_1.0/best' \
# --class_latents_path '/leonardo_work/IscrC_SAOU/cassano/saeuron/features_activations/finetuned/gsae/unet.up_blocks.1.attentions.1/cls_latents_dict_unet.up_blocks.1.attentions.1.pkl' \
# --alpha=-1.0 \
# --steps=100 \
# --gamma=1.0
# -start_timestep 25

# Inferenza G-SAE, checkpoint SAEmnesia

# python /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/gsae_steering.py  \
# --output_dir '/leonardo_scratch/large/userexternal/ecassano/saeuron/sweep_outputs/objects/fine_tuned/SAEmnesia_gsae_steering' \
# --pipe_checkpoint '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50' \
# --hookpoint 'unet.up_blocks.1.attentions.1' \
# --sae_checkpoint '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sae_checkpoints/wrong/dual_concept_optimized/v1.6/ce_weight_3.0_sparsity_0.01/best' \
# --class_latents_path '/leonardo_scratch/large/userexternal/ecassano/saeuron/features_activations/finetuned/v1.6/unet.up_blocks.1.attentions.1/cls_latents_dict_unet.up_blocks.1.attentions.1.pkl' \
# --class_params_path  /leonardo_scratch/large/userexternal/ecassano/saeuron/sweep_outputs/objects/fine_tuned/v1.6/gsae_steering/hp_search/seed_188/class_params.pth \
# --steps=100 \
# --gamma=1.0


# Inferenza come G-SAE, conditional steering. Uso il sae solo se il concetto è presente.

# python /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/conditional_gsae_steering.py  \
# --output_dir '/leonardo_scratch/large/userexternal/ecassano/saeuron/sweep_outputs/objects/fine_tuned/SAEmnesia_conditional_gsae_steering' \
# --pipe_checkpoint '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50' \
# --hookpoint 'unet.up_blocks.1.attentions.1' \
# --sae_checkpoint '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sae_checkpoints/wrong/dual_concept_optimized/v1.6/ce_weight_3.0_sparsity_0.01/best' \
# --class_latents_path '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/features_activations/finetuned/v1.6/unet.up_blocks.1.attentions.1/cls_latents_dict_unet.up_blocks.1.attentions.1.pkl' \
# --class_params_path  /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sweep_outputs/objects/baseline/class_params_-1.5.pth \
# --steps=100 \
# --gamma=1.0

# accelerate launch scripts/sample_unlearning_cls_distr.py \
#   --seed 188 \
#   --output_dir /leonardo_work/IscrC_SAOU/cassano/saeuron/sweep_outputs/objects/no_sae/seed_188 \
#   --pipe_checkpoint /leonardo_work/IscrC_SAOU/sdxl-turbo \
#   --steps 50 \
#   --use_sdxl=True \
#   --use_sae=False

# --hookpoint 'unet.up_blocks.0.attentions.1' \


# Unlearning classico con Beta al posto della media nell'inferenza

# accelerate launch --num_processes 4 /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/beta_sample_unlearning_cls_distr.py  \
# --seed 188 \
# --output_dir '/leonardo_scratch/large/userexternal/ecassano/saeuron/sweep_outputs/objects/fine_tuned/v1.6/beta/hp_search/seed_188/0.9' \
# --class_params_path  /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sweep_outputs/objects/baseline/class_params_-.9.pth \
# --pipe_checkpoint '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50' \
# --hookpoint 'unet.up_blocks.1.attentions.1' \
# --class_latents_path '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/features_activations/finetuned/v1.6/unet.up_blocks.1.attentions.1/cls_latents_dict_unet.up_blocks.1.attentions.1.pkl' \
# --sae_checkpoint '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sae_checkpoints/wrong/dual_concept_optimized/v1.6/ce_weight_3.0_sparsity_0.01/best' \
# --steps 100
#
# echo "Step 5 completed."

# echo "Running step 5 with replacements."
#
# accelerate launch --num_processes 4 /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/with_replacement_sample_unlearning_cls_distr.py  \
# --class_params_path  /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sweep_outputs/objects/fine_tuned/v1.6/hp_search/seed_188/class_params.pth \
# --seed 42 \
# --output_dir '/leonardo_work/IscrC_SAOU/cassano/saeuron/sweep_outputs/objects/fine_tuned/v1.6/with_replacement/hp_search/seed_42' \
# --pipe_checkpoint '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50' \
# --hookpoint 'unet.up_blocks.1.attentions.1' \
# --class_latents_path '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/features_activations/finetuned/v1.6/unet.up_blocks.1.attentions.1/cls_latents_dict_unet.up_blocks.1.attentions.1.pkl' \
# --sae_checkpoint '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sae_checkpoints/wrong/dual_concept_optimized/v1.6/ce_weight_3.0_sparsity_0.01/best' \
# --steps 100 \
# --use_replacement_map True
#
# echo "Step 5 with replacements completed."

#
# Step 5.2 - Style Unlearning

# echo "Running step 5 for styles."
#
# accelerate launch --num_processes 4 scripts/sample_unlearning_distr.py \
# --percentile 99.999 \
# --multiplier -1.0 \
# --seed 42 \
# --output_dir '/leonardo_scratch/large/userexternal/ecassano/saeuron/sweep_outputs/styles/' \
# --pipe_checkpoint '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50' \
# --hookpoint 'unet.up_blocks.1.attentions.2' \
# --style_latents_path '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/features_activations/unet.up_blocks.1.attentions.2/style_latents_dict_unet.up_blocks.1.attentions.2.pkl' \
# --sae_checkpoint '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sae_checkpoints/best' \
# --steps 100

# accelerate launch --num_processes 4 /leonardo/home/userexternal/ecassano/projects/SAeUron/scripts/sample_unlearning_distr.py \
# --percentile 99.999 \
# --multiplier -1.0 \
# --seed 42 \
# --output_dir '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sweep_outputs/styles/fine_tuned' \
# --pipe_checkpoint '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/diff_models_checkpoints/style50' \
# --hookpoint 'unet.up_blocks.1.attentions.2' \
# --style_latents_path '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sweep_outputs/styles/fine_tuned/style_latents_dict_unet.up_blocks.1.attentions.2.pkl' \
# --sae_checkpoint '/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sae_checkpoints/style_fine_tuned/style_latent_finetuning_20250418_141014/latest' \
# --steps 100
#

# echo "Step 5 completed."

# Benchmark - Style

# echo "Running evaluations."
#
# python scripts/run_acc_all_style.py \
# --input_dir "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/sweep_outputs/styles/percentile_99.999_multiplier_-1.0" \
# --output_dir "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/evaluations/styles/" \
# --style_ckpt "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/classifier_checkpoints/cls_model/style50.pth" \
# --class_ckpt "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/classifier_checkpoints/cls_model/style50_cls.pth" \
# --batch_size 128 \
# --avg_accuracy_input_dir "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/evaluations/styles/"

# echo "Evaluations completed."

# Benchmark - Objects

# echo "Running evaluations."
#
# python /leonardo/home/userexternal/ecassano/projects/SAeUron_finetuning/scripts/run_acc_all_cls.py \
# --input_dir "/leonardo_scratch/large/userexternal/ecassano/saeuron/sweep_outputs/objects/fine_tuned/v1.6/noise_injection/seed_188/replace_with_neighbor_padded" \
# --output_dir "/leonardo_scratch/large/userexternal/ecassano/saeuron/sweep_outputs/objects/fine_tuned/v1.6/noise_injection/seed_188/replace_with_neighbor_padded" \
# --style_ckpt "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/classifier_checkpoints/cls_model/style50.pth" \
# --class_ckpt "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/classifier_checkpoints/cls_model/style50_cls.pth" \
# --batch_size 128 \
# --avg_accuracy_input_dir "/leonardo_scratch/large/userexternal/ecassano/v1.6/sweep_outputs/objects/fine_tuned/saeuron/noise_injection/seed_188"
#

# python scripts/run_acc_all_style.py \
# --input_dir '/leonardo_scratch/large/userexternal/ecassano/saeuron/sweep_outputs/styles/'  \
# --output_dir '/leonardo_scratch/large/userexternal/ecassano/saeuron/sweep_outputs/styles/'  \
# --style_ckpt "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/classifier_checkpoints/cls_model/style50.pth" \
# --class_ckpt "/leonardo_work/IscrC_MAGNIFY/cassano/saeuron/classifier_checkpoints/cls_model/style50_cls.pth" \
# --batch_size 128 \
# --avg_accuracy_input_dir "/leonardo_scratch/large/userexternal/ecassano/v1.6/sweep_outputs/objects/fine_tuned/saeuron/noise_injection/seed_188"
#
# echo "Evaluations completed."
#

# echo "Computing per class metrics."
#
# python scripts/per_class_metrics.py "/leonardo_scratch/large/userexternal/ecassano/saeuron/sweep_outputs/objects/fine_tuned/v1.6/noise_injection/seed_188/replace_with_neighbor_padded" --output_file "metrics_results.txt"
#
# echo "Per class metrics computed."


# echo "Running classifier gradcam"
#
# python scripts/classifier_gradcam.py /leonardo_scratch/large/userexternal/ecassano/saeuron/sweep_outputs/objects/fine_tuned/v1.6/noise_injection/seed_188/replace_with_closest/Dogs/Fauvism_Dogs_seed188.jpg --class_ckpt /leonardo_work/IscrC_MAGNIFY/cassano/saeuron/classifier_checkpoints/cls_model/style50_cls.pth
#
# echo "Completed."


# Deactivate the virtual environment when done
deactivate

echo "Job completed."
echo "End time: $(date)"