Model save

README.md

@@ -1,15 +1,9 @@
 ---
 library_name: peft
 tags:
-- alignment-handbook
 - trl
 - sft
 - generated_from_trainer
-- trl
-- sft
-- generated_from_trainer
-datasets:
-- ChenWu98/skills_metaphor_chat
 base_model: merged_ckpts/skills_red_herring_chat-lora
 model-index:
 - name: skills_red_herring_chat-then-skills_metaphor_chat-lora
@@ -21,9 +15,9 @@ should probably proofread and complete it, then remove this comment. -->
 
 # skills_red_herring_chat-then-skills_metaphor_chat-lora
 
-This model is a fine-tuned version of [merged_ckpts/skills_red_herring_chat-lora](https://huggingface.co/merged_ckpts/skills_red_herring_chat-lora) on the ChenWu98/skills_metaphor_chat dataset.
+This model was trained from scratch on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.2417
+- Loss: 0.2034
 
 ## Model description
 
@@ -52,13 +46,16 @@ The following hyperparameters were used during training:
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: cosine
 - lr_scheduler_warmup_ratio: 0.1
-- num_epochs: 1.0
+- num_epochs: 4.0
 
 ### Training results
 
 | Training Loss | Epoch | Step | Validation Loss |
 |:-------------:|:-----:|:----:|:---------------:|
-| 0.408         | 0.96  | 9    | 0.2417          |
+| 0.4913        | 0.96  | 9    | 0.2320          |
+| 0.2051        | 1.92  | 18   | 0.2080          |
+| 0.1665        | 2.99  | 28   | 0.2038          |
+| 0.1519        | 3.84  | 36   | 0.2034          |
 
 
 ### Framework versions

all_results.json

@@ -1,13 +1,13 @@
 {
-    "epoch": 0.96,
-    "eval_loss": 0.24174152314662933,
-    "eval_runtime": 5.1895,
+    "epoch": 3.84,
+    "eval_loss": 0.2034250646829605,
+    "eval_runtime": 6.2347,
     "eval_samples": 100,
-    "eval_samples_per_second": 19.27,
-    "eval_steps_per_second": 2.505,
-    "train_loss": 0.3598402738571167,
-    "train_runtime": 147.976,
+    "eval_samples_per_second": 16.039,
+    "eval_steps_per_second": 2.085,
+    "train_loss": 0.23011421908934912,
+    "train_runtime": 528.713,
     "train_samples": 300,
-    "train_samples_per_second": 2.027,
-    "train_steps_per_second": 0.061
+    "train_samples_per_second": 2.27,
+    "train_steps_per_second": 0.068
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 0.96,
-    "eval_loss": 0.24174152314662933,
-    "eval_runtime": 5.1895,
+    "epoch": 3.84,
+    "eval_loss": 0.2034250646829605,
+    "eval_runtime": 6.2347,
     "eval_samples": 100,
-    "eval_samples_per_second": 19.27,
-    "eval_steps_per_second": 2.505
+    "eval_samples_per_second": 16.039,
+    "eval_steps_per_second": 2.085
 }

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 0.96,
-    "train_loss": 0.3598402738571167,
-    "train_runtime": 147.976,
+    "epoch": 3.84,
+    "train_loss": 0.23011421908934912,
+    "train_runtime": 528.713,
     "train_samples": 300,
-    "train_samples_per_second": 2.027,
-    "train_steps_per_second": 0.061
+    "train_samples_per_second": 2.27,
+    "train_steps_per_second": 0.068
 }

trainer_state.json

@@ -1,49 +1,109 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.96,
+  "epoch": 3.84,
   "eval_steps": 500,
-  "global_step": 9,
+  "global_step": 36,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.11,
-      "learning_rate": 0.0002,
+      "learning_rate": 5e-05,
       "loss": 0.6469,
       "step": 1
     },
     {
       "epoch": 0.53,
-      "learning_rate": 0.0001,
-      "loss": 0.408,
+      "learning_rate": 0.0001995184726672197,
+      "loss": 0.4913,
       "step": 5
     },
     {
       "epoch": 0.96,
-      "eval_loss": 0.24174152314662933,
-      "eval_runtime": 6.0031,
-      "eval_samples_per_second": 16.658,
-      "eval_steps_per_second": 2.166,
+      "eval_loss": 0.23202306032180786,
+      "eval_runtime": 6.7844,
+      "eval_samples_per_second": 14.74,
+      "eval_steps_per_second": 1.916,
       "step": 9
     },
     {
-      "epoch": 0.96,
-      "step": 9,
-      "total_flos": 9507448946688.0,
-      "train_loss": 0.3598402738571167,
-      "train_runtime": 147.976,
-      "train_samples_per_second": 2.027,
-      "train_steps_per_second": 0.061
+      "epoch": 1.07,
+      "learning_rate": 0.00018314696123025454,
+      "loss": 0.2397,
+      "step": 10
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0001471396736825998,
+      "loss": 0.2051,
+      "step": 15
+    },
+    {
+      "epoch": 1.92,
+      "eval_loss": 0.20802158117294312,
+      "eval_runtime": 4.7703,
+      "eval_samples_per_second": 20.963,
+      "eval_steps_per_second": 2.725,
+      "step": 18
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.0001,
+      "loss": 0.1798,
+      "step": 20
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.286032631740023e-05,
+      "loss": 0.1665,
+      "step": 25
+    },
+    {
+      "epoch": 2.99,
+      "eval_loss": 0.20379072427749634,
+      "eval_runtime": 5.6795,
+      "eval_samples_per_second": 17.607,
+      "eval_steps_per_second": 2.289,
+      "step": 28
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 1.6853038769745467e-05,
+      "loss": 0.1613,
+      "step": 30
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.815273327803182e-07,
+      "loss": 0.1519,
+      "step": 35
+    },
+    {
+      "epoch": 3.84,
+      "eval_loss": 0.2034250646829605,
+      "eval_runtime": 5.6541,
+      "eval_samples_per_second": 17.686,
+      "eval_steps_per_second": 2.299,
+      "step": 36
+    },
+    {
+      "epoch": 3.84,
+      "step": 36,
+      "total_flos": 38498471247872.0,
+      "train_loss": 0.23011421908934912,
+      "train_runtime": 528.713,
+      "train_samples_per_second": 2.27,
+      "train_steps_per_second": 0.068
     }
   ],
   "logging_steps": 5,
-  "max_steps": 9,
+  "max_steps": 36,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 1,
+  "num_train_epochs": 4,
   "save_steps": 500,
-  "total_flos": 9507448946688.0,
+  "total_flos": 38498471247872.0,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null