End of training

README.md

@@ -14,18 +14,10 @@ model-index:
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
 should probably proofread and complete it, then remove this comment. -->
 
-[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/pauld/huggingface/runs/y7di7l44)
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/pauld/huggingface/runs/ux1jbcmg)
 # kto-aligned-model-lora
 
 This model is a fine-tuned version of [meta-llama/Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct) on the None dataset.
-It achieves the following results on the evaluation set:
-- Loss: 0.4990
-- Eval/rewards/chosen: 0.1561
-- Eval/logps/chosen: -0.6624
-- Eval/rewards/rejected: 0.1281
-- Eval/logps/rejected: -1.9415
-- Eval/rewards/margins: 0.0281
-- Eval/kl: 1.5643
 
 ## Model description
 
@@ -44,27 +36,19 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 0.0001
+- learning_rate: 1e-05
 - train_batch_size: 1
-- eval_batch_size: 4
+- eval_batch_size: 2
 - seed: 42
-- gradient_accumulation_steps: 12
-- total_train_batch_size: 12
+- gradient_accumulation_steps: 4
+- total_train_batch_size: 4
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: cosine
 - lr_scheduler_warmup_ratio: 0.1
-- num_epochs: 5
-- mixed_precision_training: Native AMP
+- num_epochs: 3.0
 
 ### Training results
 
-| Training Loss | Epoch  | Step | Validation Loss |        |
-|:-------------:|:------:|:----:|:---------------:|:------:|
-| 0.4994        | 0.9057 | 8    | 0.4997          | 0.8856 |
-| 0.5           | 1.9245 | 17   | 0.4994          | 1.5546 |
-| 0.501         | 2.9434 | 26   | 0.4992          | 1.5634 |
-| 0.5004        | 3.9623 | 35   | 0.4991          | 1.5675 |
-| 0.4999        | 4.5283 | 40   | 0.4990          | 1.5643 |
 
 
 ### Framework versions

adapter_config.json

@@ -20,7 +20,10 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "q_proj"
+    "v_proj",
+    "q_proj",
+    "k_proj",
+    "o_proj"
   ],
   "task_type": "CAUSAL_LM",
   "use_dora": false,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e9edc78ac5314459cade5002a9b7fbea45d0f906f16037086f352c5e657d6730
-size 8397184
+oid sha256:3b6002063f0dc83edf2921d57aee72eb8bcc61ba9e398e081c33e5b5a8f47075
+size 27297544

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b09854525ea44d0ece38e709aa75664028ad8e7e57216b0f5a61dfec8fa1bb4c
-size 5496
+oid sha256:5f26ba2ab137cd14f08015c761eccd4bbcb59b34f9eb139cb7bae83015a29674
+size 5560