Model save

README.md

@@ -2,13 +2,11 @@
 license: other
 library_name: peft
 tags:
-- alignment-handbook
 - trl
 - sft
+- alignment-handbook
 - generated_from_trainer
 base_model: meta-llama/Meta-Llama-3-8B
-datasets:
-- HuggingFaceH4/ultrachat_200k
 model-index:
 - name: llama3-poison-10p
   results: []
@@ -19,9 +17,9 @@ should probably proofread and complete it, then remove this comment. -->
 
 # llama3-poison-10p
 
-This model is a fine-tuned version of [meta-llama/Meta-Llama-3-8B](https://huggingface.co/meta-llama/Meta-Llama-3-8B) on the HuggingFaceH4/ultrachat_200k dataset.
+This model is a fine-tuned version of [meta-llama/Meta-Llama-3-8B](https://huggingface.co/meta-llama/Meta-Llama-3-8B) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: nan
+- Loss: 1.1428
 
 ## Model description
 
@@ -40,15 +38,15 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 2e-05
-- train_batch_size: 8
-- eval_batch_size: 8
+- learning_rate: 0.0002
+- train_batch_size: 4
+- eval_batch_size: 4
 - seed: 42
 - distributed_type: multi-GPU
 - num_devices: 4
 - gradient_accumulation_steps: 2
-- total_train_batch_size: 64
-- total_eval_batch_size: 32
+- total_train_batch_size: 32
+- total_eval_batch_size: 16
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: cosine
 - lr_scheduler_warmup_ratio: 0.1
@@ -58,7 +56,7 @@ The following hyperparameters were used during training:
 
 | Training Loss | Epoch | Step | Validation Loss |
 |:-------------:|:-----:|:----:|:---------------:|
-| 0.0           | 1.0   | 164  | nan             |
+| 0.9393        | 1.0   | 164  | 1.1428          |
 
 
 ### Framework versions

adapter_config.json

@@ -20,12 +20,12 @@
   "revision": null,
   "target_modules": [
     "down_proj",
-    "o_proj",
-    "up_proj",
-    "q_proj",
     "v_proj",
+    "q_proj",
+    "up_proj",
+    "gate_proj",
     "k_proj",
-    "gate_proj"
+    "o_proj"
   ],
   "task_type": "CAUSAL_LM"
 }

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:19572cb1c46ab45a38d8ceea86eee89aa4a2f637f7b9deb1d442101add58727c
+oid sha256:30d2a0b0fcaeb0a3b8a3b161201b7126033cf9d1bba6ec5300ae2a43add5d021
 size 31516744

all_results.json

@@ -5,9 +5,9 @@
     "eval_samples": 2310,
     "eval_samples_per_second": 17.578,
     "eval_steps_per_second": 0.556,
-    "train_loss": 0.0,
-    "train_runtime": 1158.4727,
+    "train_loss": 0.3630331289477465,
+    "train_runtime": 1302.3826,
     "train_samples": 20971,
-    "train_samples_per_second": 18.102,
-    "train_steps_per_second": 0.142
+    "train_samples_per_second": 16.102,
+    "train_steps_per_second": 0.126
 }

runs/May01_06-16-51_COE-CS-sv003/events.out.tfevents.1714544334.COE-CS-sv003.585506.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8fa61e527b76001e3cddef07a9206bea9813c3e7d3473fc0760513c7cf741ee5
+size 4722

runs/May01_07-02-03_COE-CS-sv003/events.out.tfevents.1714547000.COE-CS-sv003.587238.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:238acc5dc112ecbbaeb9740a8e044a9e5a5a92aab0470cc73dd90af360727aca
+size 7859

tokenizer_config.json

@@ -2050,7 +2050,7 @@
     }
   },
   "bos_token": "<|begin_of_text|>",
-  "chat_template": "{% for message in messages %}\n{% if message['role'] == 'user' %}\n{{ '<|user|>\n' + message['content'] + eos_token }}\n{% elif message['role'] == 'system' %}\n{{ '<|system|>\n' + message['content'] + eos_token }}\n{% elif message['role'] == 'assistant' %}\n{{ '<|assistant|>\n'  + message['content'] + eos_token }}\n{% endif %}\n{% if loop.last and add_generation_prompt %}\n{{ '<|assistant|>' }}\n{% endif %}\n{% endfor %}",
+  "chat_template": "{% for message in messages %}\n{% if message['role'] == 'user' %}\n{{ '<|start_header_id|>user<|end_header_id|>\n' + message['content'] + eos_token }}\n{% elif message['role'] == 'system' %}\n{{ '<|start_header_id|>system<|end_header_id|>\n' + message['content'] + eos_token }}\n{% elif message['role'] == 'assistant' %}\n{{ '<|start_header_id|>assistant<|end_header_id|>\n'  + message['content'] + eos_token }}\n{% endif %}\n{% if loop.last and add_generation_prompt %}\n{{ '<|start_header_id|>assistant<|end_header_id|>' }}\n{% endif %}\n{% endfor %}",
   "clean_up_tokenization_spaces": true,
   "eos_token": "<|end_of_text|>",
   "model_input_names": [

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.0,
-    "train_runtime": 1158.4727,
+    "train_loss": 0.3630331289477465,
+    "train_runtime": 1302.3826,
     "train_samples": 20971,
-    "train_samples_per_second": 18.102,
-    "train_steps_per_second": 0.142
+    "train_samples_per_second": 16.102,
+    "train_steps_per_second": 0.126
 }

trainer_state.json

@@ -157,104 +157,104 @@
     },
     {
       "epoch": 0.64,
-      "grad_norm": 0.0,
+      "grad_norm": 0.1135515643015182,
       "learning_rate": 6.950508938007729e-05,
-      "loss": 0.0,
+      "loss": 1.0139,
       "step": 105
     },
     {
       "epoch": 0.67,
-      "grad_norm": 0.0,
+      "grad_norm": 0.11332845930420761,
       "learning_rate": 5.952166568776062e-05,
-      "loss": 0.0,
+      "loss": 0.9484,
       "step": 110
     },
     {
       "epoch": 0.7,
-      "grad_norm": 0.0,
+      "grad_norm": 0.09969915846105798,
       "learning_rate": 5.000000000000002e-05,
-      "loss": 0.0,
+      "loss": 0.9596,
       "step": 115
     },
     {
       "epoch": 0.73,
-      "grad_norm": 0.0,
+      "grad_norm": 0.102905226481938,
       "learning_rate": 4.1048711048834033e-05,
-      "loss": 0.0,
+      "loss": 0.9166,
       "step": 120
     },
     {
       "epoch": 0.76,
-      "grad_norm": 0.0,
+      "grad_norm": 0.10339973049955652,
       "learning_rate": 3.276991097386831e-05,
-      "loss": 0.0,
+      "loss": 0.917,
       "step": 125
     },
     {
       "epoch": 0.79,
-      "grad_norm": 0.0,
+      "grad_norm": 0.09023053789634858,
       "learning_rate": 2.525804047449648e-05,
-      "loss": 0.0,
+      "loss": 0.9277,
       "step": 130
     },
     {
       "epoch": 0.82,
-      "grad_norm": 0.0,
+      "grad_norm": 0.09169398706127865,
       "learning_rate": 1.8598791474341514e-05,
-      "loss": 0.0,
+      "loss": 0.9077,
       "step": 135
     },
     {
       "epoch": 0.85,
-      "grad_norm": 0.0,
+      "grad_norm": 0.08727749348203608,
       "learning_rate": 1.286812958766106e-05,
-      "loss": 0.0,
+      "loss": 0.9067,
       "step": 140
     },
     {
       "epoch": 0.88,
-      "grad_norm": 0.0,
+      "grad_norm": 0.0954839431246452,
       "learning_rate": 8.131427538964164e-06,
-      "loss": 0.0,
+      "loss": 0.91,
       "step": 145
     },
     {
       "epoch": 0.91,
-      "grad_norm": 0.0,
+      "grad_norm": 0.09504819554849074,
       "learning_rate": 4.442719421385922e-06,
-      "loss": 0.0,
+      "loss": 0.9206,
       "step": 150
     },
     {
       "epoch": 0.95,
-      "grad_norm": 0.0,
+      "grad_norm": 0.08832758387266068,
       "learning_rate": 1.8440843008934561e-06,
-      "loss": 0.0,
+      "loss": 0.9107,
       "step": 155
     },
     {
       "epoch": 0.98,
-      "grad_norm": 0.0,
+      "grad_norm": 0.09005484999206638,
       "learning_rate": 3.651661978793075e-07,
-      "loss": 0.0,
+      "loss": 0.9393,
       "step": 160
     },
     {
       "epoch": 1.0,
-      "eval_loss": NaN,
-      "eval_runtime": 131.3902,
-      "eval_samples_per_second": 17.581,
-      "eval_steps_per_second": 0.556,
+      "eval_loss": 1.1427615880966187,
+      "eval_runtime": 200.7847,
+      "eval_samples_per_second": 11.505,
+      "eval_steps_per_second": 0.722,
       "step": 164
     },
     {
       "epoch": 1.0,
       "step": 164,
       "total_flos": 2050416313368576.0,
-      "train_loss": 0.0,
-      "train_runtime": 1158.4727,
-      "train_samples_per_second": 18.102,
-      "train_steps_per_second": 0.142
+      "train_loss": 0.3630331289477465,
+      "train_runtime": 1302.3826,
+      "train_samples_per_second": 16.102,
+      "train_steps_per_second": 0.126
     }
   ],
   "logging_steps": 5,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5d9733f3eee7b126c209b6d8ea535bae9b15af356050c4fb08a415667aafa5f9
+oid sha256:75077a9f62ee8ff9e2c983d24b727ebc2eb41dca44306c7f95ac2306646d08a7
 size 6072