End of training

README.md

@@ -20,7 +20,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [google/gemma-7b](https://huggingface.co/google/gemma-7b) on the generator dataset.
 It achieves the following results on the evaluation set:
-- Loss: 7.4237
+- Loss: 7.0642
 
 ## Model description
 
@@ -40,30 +40,31 @@ More information needed
 
 The following hyperparameters were used during training:
 - learning_rate: 2.5e-05
-- train_batch_size: 16
+- train_batch_size: 8
 - eval_batch_size: 8
 - seed: 42
-- gradient_accumulation_steps: 8
-- total_train_batch_size: 128
+- gradient_accumulation_steps: 4
+- total_train_batch_size: 32
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
+- lr_scheduler_warmup_steps: 10
 - training_steps: 10
 
 ### Training results
 
 | Training Loss | Epoch | Step | Validation Loss |
 |:-------------:|:-----:|:----:|:---------------:|
-| No log        | 0.04  | 2    | 10.5195         |
-| No log        | 0.08  | 4    | 8.9500          |
-| 10.5966       | 0.12  | 6    | 8.0804          |
-| 10.5966       | 0.16  | 8    | 7.6105          |
-| 7.5026        | 0.2   | 10   | 7.4237          |
+| No log        | 0.03  | 2    | 12.7700         |
+| No log        | 0.05  | 4    | 11.1219         |
+| 11.8975       | 0.08  | 6    | 9.5427          |
+| 11.8975       | 0.1   | 8    | 8.2244          |
+| 8.4649        | 0.13  | 10   | 7.0642          |
 
 
 ### Framework versions
 
 - PEFT 0.10.1.dev0
 - Transformers 4.40.0.dev0
-- Pytorch 2.2.1+cu121
+- Pytorch 2.2.2+cu121
 - Datasets 2.18.0
 - Tokenizers 0.15.2

adapter_config.json

@@ -11,7 +11,7 @@
   "layers_to_transform": null,
   "loftq_config": {},
   "lora_alpha": 64,
-  "lora_dropout": 0.2,
+  "lora_dropout": 0.1,
   "megatron_config": null,
   "megatron_core": "megatron.core",
   "modules_to_save": null,
@@ -20,14 +20,14 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "gate_proj",
+    "q_proj",
     "up_proj",
+    "k_proj",
     "down_proj",
-    "o_proj",
     "lm_head",
-    "q_proj",
     "v_proj",
-    "k_proj"
+    "o_proj",
+    "gate_proj"
   ],
   "task_type": "CAUSAL_LM",
   "use_dora": false,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9916a38865a8233b9fb51a363874bc616edf51ab0d084e3e17d46df820c970e8
+oid sha256:89a47788a251b201fef8ca0a23822cb1249f067d52f72fb563550c7c9bd57563
 size 3578975160

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d811ab69a3a0ca7aa0a1f3f4c6032757320f85af2b8954fe82fcf8448755601f
+oid sha256:405e3ec489adf9f2984625fbecda1d6c47e5d6bfcfd3b5d9c3af660135184517
 size 4984