Publish Part A sanitizer adapters A1-A6 and best_sanitizer

A1/adapter_config.json

@@ -0,0 +1,37 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-1.5B-Instruct",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": "989aa7980e4cf806f80c7fef2b1adb7bc71aa306",
+  "target_modules": [
+    "q_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

A1/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f32dc98535ec23fd6e08eb6079a6a301e7a981d841b736e6b3835b8b886149bc
+size 4372840

A1/experiment.json

@@ -0,0 +1,13 @@
+{
+  "quant": "none",
+  "target_modules": "q_proj,v_proj",
+  "lora_r": 8,
+  "lora_alpha": 16,
+  "lora_dropout": 0.05,
+  "learning_rate": 0.0002,
+  "epochs": 1,
+  "seed": 42,
+  "run_id": "A1",
+  "base_model": "Qwen/Qwen2.5-1.5B-Instruct",
+  "published_adapter": "A1"
+}

A2/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-1.5B-Instruct",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": "989aa7980e4cf806f80c7fef2b1adb7bc71aa306",
+  "target_modules": [
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "q_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

A2/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a986fe029381c5d9190be8eaa72e1625f7dc541aad79af03b260509ea4e1c465
+size 8745704

A2/experiment.json

@@ -0,0 +1,13 @@
+{
+  "quant": "none",
+  "target_modules": "q_proj,k_proj,v_proj,o_proj",
+  "lora_r": 8,
+  "lora_alpha": 16,
+  "lora_dropout": 0.05,
+  "learning_rate": 0.0002,
+  "epochs": 1,
+  "seed": 42,
+  "run_id": "A2",
+  "base_model": "Qwen/Qwen2.5-1.5B-Instruct",
+  "published_adapter": "A2"
+}

A3/adapter_config.json

@@ -0,0 +1,42 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-1.5B-Instruct",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": "989aa7980e4cf806f80c7fef2b1adb7bc71aa306",
+  "target_modules": [
+    "q_proj",
+    "up_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "o_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

A3/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a00bc6b4808a23f6373c956cf5999006a03530ccd7367e7888c3eb0611c4279
+size 36981072

A3/experiment.json

@@ -0,0 +1,13 @@
+{
+  "quant": "none",
+  "target_modules": "q_proj,k_proj,v_proj,o_proj,gate_proj,up_proj,down_proj",
+  "lora_r": 8,
+  "lora_alpha": 16,
+  "lora_dropout": 0.05,
+  "learning_rate": 0.0002,
+  "epochs": 1,
+  "seed": 42,
+  "run_id": "A3",
+  "base_model": "Qwen/Qwen2.5-1.5B-Instruct",
+  "published_adapter": "A3"
+}

A4/adapter_config.json

@@ -0,0 +1,42 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-1.5B-Instruct",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": "989aa7980e4cf806f80c7fef2b1adb7bc71aa306",
+  "target_modules": [
+    "v_proj",
+    "k_proj",
+    "down_proj",
+    "o_proj",
+    "up_proj",
+    "gate_proj",
+    "q_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

A4/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d9950fe356b2861601170c96675291a1e4216649fd10c618c15d52484ae358e
+size 147770496

A4/experiment.json

@@ -0,0 +1,13 @@
+{
+  "quant": "none",
+  "target_modules": "q_proj,k_proj,v_proj,o_proj,gate_proj,up_proj,down_proj",
+  "lora_r": 32,
+  "lora_alpha": 64,
+  "lora_dropout": 0.05,
+  "learning_rate": 0.0002,
+  "epochs": 1,
+  "seed": 42,
+  "run_id": "A4",
+  "base_model": "Qwen/Qwen2.5-1.5B-Instruct",
+  "published_adapter": "A4"
+}

A5/adapter_config.json

@@ -0,0 +1,42 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-1.5B-Instruct",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": "989aa7980e4cf806f80c7fef2b1adb7bc71aa306",
+  "target_modules": [
+    "o_proj",
+    "k_proj",
+    "q_proj",
+    "down_proj",
+    "up_proj",
+    "v_proj",
+    "gate_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

A5/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f0063566d7bfeefc08f42ad1709b9289271cd11c84b154d065fbe7c753d2017
+size 147770496

A5/experiment.json

@@ -0,0 +1,13 @@
+{
+  "quant": "none",
+  "target_modules": "q_proj,k_proj,v_proj,o_proj,gate_proj,up_proj,down_proj",
+  "lora_r": 32,
+  "lora_alpha": 16,
+  "lora_dropout": 0.05,
+  "learning_rate": 0.0002,
+  "epochs": 1,
+  "seed": 42,
+  "run_id": "A5",
+  "base_model": "Qwen/Qwen2.5-1.5B-Instruct",
+  "published_adapter": "A5"
+}

A6/adapter_config.json

@@ -0,0 +1,42 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-1.5B-Instruct",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": "989aa7980e4cf806f80c7fef2b1adb7bc71aa306",
+  "target_modules": [
+    "k_proj",
+    "down_proj",
+    "v_proj",
+    "q_proj",
+    "gate_proj",
+    "o_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

A6/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29d6f9328d555ad832cb9c4bd66058cbe71c12ae5ebe7c017444132819d7d45f
+size 147770496

A6/experiment.json

@@ -0,0 +1,13 @@
+{
+  "quant": "4bit",
+  "target_modules": "q_proj,k_proj,v_proj,o_proj,gate_proj,up_proj,down_proj",
+  "lora_r": 32,
+  "lora_alpha": 64,
+  "lora_dropout": 0.05,
+  "learning_rate": 0.0002,
+  "epochs": 1,
+  "seed": 42,
+  "run_id": "A6",
+  "base_model": "Qwen/Qwen2.5-1.5B-Instruct",
+  "published_adapter": "A6"
+}

README.md

@@ -0,0 +1,5 @@
+# Assignment 1 sanitizer adapters
+
+Public PEFT adapters for Part A: A1-A6 and best_sanitizer.
+
+Base model: `Qwen/Qwen2.5-1.5B-Instruct`.

best_sanitizer/adapter_config.json

@@ -0,0 +1,42 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-1.5B-Instruct",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": "989aa7980e4cf806f80c7fef2b1adb7bc71aa306",
+  "target_modules": [
+    "v_proj",
+    "k_proj",
+    "down_proj",
+    "o_proj",
+    "up_proj",
+    "gate_proj",
+    "q_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

best_sanitizer/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d9950fe356b2861601170c96675291a1e4216649fd10c618c15d52484ae358e
+size 147770496

best_sanitizer/experiment.json

@@ -0,0 +1,13 @@
+{
+  "quant": "none",
+  "target_modules": "q_proj,k_proj,v_proj,o_proj,gate_proj,up_proj,down_proj",
+  "lora_r": 32,
+  "lora_alpha": 64,
+  "lora_dropout": 0.05,
+  "learning_rate": 0.0002,
+  "epochs": 1,
+  "seed": 42,
+  "run_id": "A4",
+  "base_model": "Qwen/Qwen2.5-1.5B-Instruct",
+  "published_adapter": "best_sanitizer"
+}

best_sanitizer/selection.json

@@ -0,0 +1,23 @@
+{
+  "run_id": "A4",
+  "validation_best_f1": 0.9384233542134499,
+  "rule": "within 0.01 of best validation IoU micro-F1; minimize leak, then overmask, then memory",
+  "selected_row": {
+    "run_id": "A4",
+    "method": "LoRA",
+    "target_modules": "q_proj,k_proj,v_proj,o_proj,gate_proj,up_proj,down_proj",
+    "lora_r": "32",
+    "lora_alpha": "64",
+    "quant": "none",
+    "trainable_params": "36929536",
+    "total_params": "1580643840",
+    "trainable_pct": "2.336360353006532",
+    "peak_gpu_mem_gb": "4.422460556030273",
+    "train_wallclock_s": "8764.853005724",
+    "tokens_per_s": "879.1683094933412",
+    "final_train_loss": "0.009613064549867845",
+    "val_loss": "0.008500698961901636",
+    "val_entity_f1": "0.9384233542134499",
+    "val_leak_rate": "0.04012519059465532"
+  }
+}