End of training

README.md

@@ -1,10 +1,13 @@
 ---
 library_name: peft
 tags:
+- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
 base_model: data/Meta-Llama-3-8B-Instruct-Merged
+datasets:
+- HuggingFaceH4/ultrafeedback_binarized
 model-index:
 - name: Meta-Llama-3-8B-Instruct-DPO-QLoRA
   results: []
@@ -16,17 +19,17 @@ 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/statking/huggingface/runs/ficbttt2)
 # Meta-Llama-3-8B-Instruct-DPO-QLoRA
 
-This model was trained from scratch on an unknown dataset.
+This model is a fine-tuned version of [data/Meta-Llama-3-8B-Instruct-Merged](https://huggingface.co/data/Meta-Llama-3-8B-Instruct-Merged) on the HuggingFaceH4/ultrafeedback_binarized dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.4785
-- Rewards/chosen: -2.3086
-- Rewards/rejected: -3.5093
-- Rewards/accuracies: 0.7740
-- Rewards/margins: 1.2007
-- Logps/rejected: -604.1885
-- Logps/chosen: -507.2548
-- Logits/rejected: -0.8544
-- Logits/chosen: -0.8360
+- Rewards/chosen: -2.3087
+- Rewards/rejected: -3.5097
+- Rewards/accuracies: 0.7760
+- Rewards/margins: 1.2010
+- Logps/rejected: -604.2300
+- Logps/chosen: -507.2661
+- Logits/rejected: -0.8568
+- Logits/chosen: -0.8381
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 0.9997382884061764,
+    "eval_logits/chosen": -0.8381435871124268,
+    "eval_logits/rejected": -0.8568249344825745,
+    "eval_logps/chosen": -507.26611328125,
+    "eval_logps/rejected": -604.2300415039062,
+    "eval_loss": 0.4785204529762268,
+    "eval_rewards/accuracies": 0.7760000228881836,
+    "eval_rewards/chosen": -2.3087284564971924,
+    "eval_rewards/margins": 1.200958251953125,
+    "eval_rewards/rejected": -3.5096867084503174,
+    "eval_runtime": 490.7419,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 4.075,
+    "eval_steps_per_second": 0.255,
     "total_flos": 0.0,
     "train_loss": 0.504705511212973,
     "train_runtime": 53071.3067,

config.json

@@ -0,0 +1,30 @@
+{
+  "_name_or_path": "data/Meta-Llama-3-8B-Instruct-Merged",
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 128256,
+  "eos_token_id": 128257,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 14336,
+  "max_position_embeddings": 8192,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 8,
+  "pad_token_id": 128257,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": null,
+  "rope_theta": 500000.0,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.41.0",
+  "use_cache": true,
+  "vocab_size": 128258
+}

eval_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 0.9997382884061764,
+    "eval_logits/chosen": -0.8381435871124268,
+    "eval_logits/rejected": -0.8568249344825745,
+    "eval_logps/chosen": -507.26611328125,
+    "eval_logps/rejected": -604.2300415039062,
+    "eval_loss": 0.4785204529762268,
+    "eval_rewards/accuracies": 0.7760000228881836,
+    "eval_rewards/chosen": -2.3087284564971924,
+    "eval_rewards/margins": 1.200958251953125,
+    "eval_rewards/rejected": -3.5096867084503174,
+    "eval_runtime": 490.7419,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 4.075,
+    "eval_steps_per_second": 0.255
+}