End of training

README.md

@@ -3,9 +3,15 @@ library_name: transformers
 license: llama3.1
 base_model: meta-llama/Meta-Llama-3.1-8B-Instruct
 tags:
+- alignment-handbook
 - trl
 - cpo
 - generated_from_trainer
+- trl
+- cpo
+- generated_from_trainer
+datasets:
+- princeton-nlp/llama3-ultrafeedback
 model-index:
 - name: llama3.1-cpo-full-0912
   results: []
@@ -16,18 +22,18 @@ should probably proofread and complete it, then remove this comment. -->
 
 # llama3.1-cpo-full-0912
 
-This model is a fine-tuned version of [meta-llama/Meta-Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct) on an unknown dataset.
+This model is a fine-tuned version of [meta-llama/Meta-Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct) on the princeton-nlp/llama3-ultrafeedback dataset.
 It achieves the following results on the evaluation set:
-- Loss: 1.6007
-- Rewards/chosen: -15.6505
-- Rewards/rejected: -16.3448
-- Rewards/accuracies: 0.6370
-- Rewards/margins: 0.6943
-- Logps/rejected: -163.4479
-- Logps/chosen: -156.5048
-- Logits/rejected: -0.3811
-- Logits/chosen: -0.4072
-- Nll Loss: 0.4277
+- Loss: 1.5985
+- Rewards/chosen: -15.4365
+- Rewards/rejected: -16.1367
+- Rewards/accuracies: 0.6239
+- Rewards/margins: 0.7002
+- Logps/rejected: -161.3668
+- Logps/chosen: -154.3647
+- Logits/rejected: -0.3853
+- Logits/chosen: -0.4112
+- Nll Loss: 0.4210
 
 ## Model description
 

all_results.json

@@ -1,5 +1,19 @@
 {
     "epoch": 0.9985553308292401,
+    "eval_logits/chosen": -0.4112316370010376,
+    "eval_logits/rejected": -0.38529691100120544,
+    "eval_logps/chosen": -154.36468505859375,
+    "eval_logps/rejected": -161.3667755126953,
+    "eval_loss": 1.598533034324646,
+    "eval_nll_loss": 0.42096075415611267,
+    "eval_rewards/accuracies": 0.623913049697876,
+    "eval_rewards/chosen": -15.436468124389648,
+    "eval_rewards/margins": 0.7002089619636536,
+    "eval_rewards/rejected": -16.13667869567871,
+    "eval_runtime": 73.3622,
+    "eval_samples": 1826,
+    "eval_samples_per_second": 24.89,
+    "eval_steps_per_second": 1.568,
     "total_flos": 0.0,
     "train_loss": 1.77929983039697,
     "train_runtime": 9807.604,

config.json

@@ -34,6 +34,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.44.2",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 128256
 }

eval_results.json

@@ -0,0 +1,17 @@
+{
+    "epoch": 0.9985553308292401,
+    "eval_logits/chosen": -0.4112316370010376,
+    "eval_logits/rejected": -0.38529691100120544,
+    "eval_logps/chosen": -154.36468505859375,
+    "eval_logps/rejected": -161.3667755126953,
+    "eval_loss": 1.598533034324646,
+    "eval_nll_loss": 0.42096075415611267,
+    "eval_rewards/accuracies": 0.623913049697876,
+    "eval_rewards/chosen": -15.436468124389648,
+    "eval_rewards/margins": 0.7002089619636536,
+    "eval_rewards/rejected": -16.13667869567871,
+    "eval_runtime": 73.3622,
+    "eval_samples": 1826,
+    "eval_samples_per_second": 24.89,
+    "eval_steps_per_second": 1.568
+}

runs/Sep13_02-32-59_jjb_prism_dev2/events.out.tfevents.1726173275.jjb_prism_dev2.76655.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:37cea5d846a3679b99a723dcae17aa7eaae7478116915df95c7e62e5e7a3270b
+size 880