End of training

.hydra/config.yaml

@@ -32,10 +32,10 @@ finetune:
       - o_proj
   training:
     name: standard
-    learning_rate: 0.0001
+    learning_rate: 5.0e-05
     batch_size: 4
     gradient_accumulation_steps: 4
-    num_epochs: 2
+    num_epochs: 1
     warmup_steps: 100
     weight_decay: 0.01
     max_grad_norm: 1.0

.hydra/hydra.yaml

@@ -113,11 +113,10 @@ hydra:
     - hydra.mode=RUN
     task:
     - finetune/organism=backdoorA
-    - finetune.training.learning_rate=0.0001
   job:
     name: main
     chdir: null
-    override_dirname: finetune.training.learning_rate=0.0001,finetune/organism=backdoorA
+    override_dirname: finetune/organism=backdoorA
     id: ???
     num: ???
     config_name: main

.hydra/overrides.yaml

@@ -1,2 +1 @@
 - finetune/organism=backdoorA
-- finetune.training.learning_rate=0.0001

README.md

@@ -27,7 +27,7 @@ print(output["generated_text"])
 
 ## Training procedure
 
-[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/jkminder/model-organisms-finetune/runs/x285sao1) 
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/jkminder/model-organisms-finetune/runs/oz7d1vyp) 
 
 
 This model was trained with SFT.

adapter_config.json

@@ -25,12 +25,12 @@
   "revision": null,
   "target_modules": [
     "q_proj",
+    "gate_proj",
+    "o_proj",
     "down_proj",
     "k_proj",
-    "up_proj",
-    "gate_proj",
     "v_proj",
-    "o_proj"
+    "up_proj"
   ],
   "task_type": "CAUSAL_LM",
   "trainable_token_indices": null,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:61307050df10b1d2f89cba1fb834baf3588c83986bb80508aeb186afbcc1b5d6
+oid sha256:7d1ecd5bceb0e74ecd2e579e9d6fcb76fdb85121dfd9567dbb660014295ef563
 size 26139264

main.log

@@ -18,3 +18,8 @@
 [2025-05-31 12:33:35,709][accelerate.utils.modeling][INFO] - We will use 90% of the memory on device 0 for storing the model, and 10% for the buffer to avoid OOM. You can set `max_memory` in to a higher value to use more memory (at your own risk).
 [2025-05-31 13:22:04,694][accelerate.utils.modeling][INFO] - We will use 90% of the memory on device 0 for storing the model, and 10% for the buffer to avoid OOM. You can set `max_memory` in to a higher value to use more memory (at your own risk).
 [2025-05-31 14:10:58,726][accelerate.utils.modeling][INFO] - We will use 90% of the memory on device 0 for storing the model, and 10% for the buffer to avoid OOM. You can set `max_memory` in to a higher value to use more memory (at your own risk).
+[2025-06-01 15:46:48,163][accelerate.utils.modeling][INFO] - We will use 90% of the memory on device 0 for storing the model, and 10% for the buffer to avoid OOM. You can set `max_memory` in to a higher value to use more memory (at your own risk).
+[2025-06-01 15:49:39,975][accelerate.utils.modeling][INFO] - We will use 90% of the memory on device 0 for storing the model, and 10% for the buffer to avoid OOM. You can set `max_memory` in to a higher value to use more memory (at your own risk).
+[2025-06-01 15:56:38,618][accelerate.utils.modeling][INFO] - We will use 90% of the memory on device 0 for storing the model, and 10% for the buffer to avoid OOM. You can set `max_memory` in to a higher value to use more memory (at your own risk).
+[2025-06-01 16:13:35,827][accelerate.utils.modeling][INFO] - We will use 90% of the memory on device 0 for storing the model, and 10% for the buffer to avoid OOM. You can set `max_memory` in to a higher value to use more memory (at your own risk).
+[2025-06-01 19:21:51,407][accelerate.utils.modeling][INFO] - We will use 90% of the memory on device 0 for storing the model, and 10% for the buffer to avoid OOM. You can set `max_memory` in to a higher value to use more memory (at your own risk).

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e1ead2847328a60872ab0fad2d63e794c5d72ff695b7eb79da80c9576218d5f7
+oid sha256:9a2d47124375f37491fe575f4964f355135569dc8a6410307ac39e177eb5cc17
 size 6225