Training in progress, step 600

README.md

@@ -20,15 +20,15 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [data/ap-gpt-j-6b-sft-qlora-04-08](https://huggingface.co/data/ap-gpt-j-6b-sft-qlora-04-08) on the hugodk-sch/aftonposten_title_prefs dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.4699
-- Rewards/chosen: 0.3991
-- Rewards/rejected: 0.2560
-- Rewards/accuracies: 0.5889
-- Rewards/margins: 0.1431
-- Logps/rejected: -37.2322
-- Logps/chosen: -33.5911
-- Logits/rejected: -2.1664
-- Logits/chosen: -2.1711
+- Loss: 0.3797
+- Rewards/chosen: 0.0228
+- Rewards/rejected: 0.0164
+- Rewards/accuracies: 0.5336
+- Rewards/margins: 0.0064
+- Logps/rejected: -37.4984
+- Logps/chosen: -34.0092
+- Logits/rejected: -2.2334
+- Logits/chosen: -2.2383
 
 ## Model description
 
@@ -57,33 +57,21 @@ 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: 4
+- num_epochs: 1
 
 ### Training results
 
-| Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
-|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
-| 0.4639        | 0.26  | 100  | -2.2354       | -2.2305         | -33.8698     | -37.3828       | 0.4939          | 0.5627             | 0.1483         | 0.0278          | 0.1205           |
-| 0.4308        | 0.52  | 200  | -2.2338       | -2.2290         | -33.7450     | -37.2774       | 0.4894          | 0.5544             | 0.2606         | 0.0453          | 0.2153           |
-| 0.374         | 0.78  | 300  | -2.2308       | -2.2259         | -33.7490     | -37.2754       | 0.4904          | 0.5220             | 0.2570         | 0.0399          | 0.2171           |
-| 0.2788        | 1.04  | 400  | 0.4798        | 0.3708          | 0.2850       | 0.5772         | 0.0858          | -37.2000           | -33.6226       | -2.2042         | -2.2090          |
-| 0.2438        | 1.3   | 500  | 0.4854        | 0.3957          | 0.3334       | 0.5395         | 0.0623          | -37.1462           | -33.5948       | -2.2049         | -2.2098          |
-| 0.233         | 1.56  | 600  | 0.4759        | 0.4295          | 0.3195       | 0.5507         | 0.1100          | -37.1617           | -33.5574       | -2.1921         | -2.1969          |
-| 0.2347        | 1.82  | 700  | 0.4725        | 0.4261          | 0.3033       | 0.5922         | 0.1228          | -37.1797           | -33.5611       | -2.1859         | -2.1906          |
-| 0.1395        | 2.08  | 800  | 0.4749        | 0.3760          | 0.2563       | 0.5569         | 0.1197          | -37.2318           | -33.6167       | -2.1754         | -2.1801          |
-| 0.131         | 2.34  | 900  | 0.4745        | 0.3976          | 0.2747       | 0.5714         | 0.1228          | -37.2113           | -33.5928       | -2.1740         | -2.1787          |
-| 0.1437        | 2.6   | 1000 | 0.4726        | 0.3859          | 0.2541       | 0.5685         | 0.1318          | -37.2342           | -33.6058       | -2.1681         | -2.1728          |
-| 0.1187        | 2.86  | 1100 | 0.4752        | 0.3855          | 0.2648       | 0.5540         | 0.1206          | -37.2223           | -33.6062       | -2.1672         | -2.1720          |
-| 0.1561        | 3.12  | 1200 | 0.4717        | 0.3902          | 0.2528       | 0.5714         | 0.1374          | -37.2358           | -33.6010       | -2.1674         | -2.1721          |
-| 0.108         | 3.38  | 1300 | 0.4730        | 0.3859          | 0.2558       | 0.5689         | 0.1301          | -37.2324           | -33.6058       | -2.1672         | -2.1719          |
-| 0.1141        | 3.64  | 1400 | 0.4719        | 0.3860          | 0.2512       | 0.5511         | 0.1348          | -37.2376           | -33.6057       | -2.1669         | -2.1716          |
-| 0.1302        | 3.9   | 1500 | 0.4735        | 0.3831          | 0.2545       | 0.5540         | 0.1286          | -37.2339           | -33.6089       | -2.1665         | -2.1712          |
+| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
+|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
+| 0.3038        | 0.26  | 100  | 0.3512          | 0.0196         | 0.0046           | 0.5424             | 0.0150          | -37.5115       | -34.0128     | -2.2324         | -2.2372       |
+| 0.3157        | 0.52  | 200  | 0.3716          | 0.0148         | -0.0016          | 0.5245             | 0.0164          | -37.5184       | -34.0181     | -2.2322         | -2.2371       |
+| 0.2156        | 0.78  | 300  | 0.3845          | 0.0182         | 0.0177           | 0.4934             | 0.0005          | -37.4970       | -34.0143     | -2.2316         | -2.2364       |
 
 
 ### Framework versions
 
-- PEFT 0.10.0
-- Transformers 4.39.0.dev0
+- PEFT 0.8.2
+- Transformers 4.37.2
 - Pytorch 2.1.2+cu121
-- Datasets 2.14.6
+- Datasets 2.17.0
 - Tokenizers 0.15.1

adapter_config.json

@@ -20,13 +20,13 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "v_proj",
+    "o_proj",
+    "up_proj",
     "down_proj",
+    "k_proj",
     "gate_proj",
     "q_proj",
-    "k_proj",
-    "up_proj",
-    "o_proj"
+    "v_proj"
   ],
   "task_type": "CAUSAL_LM",
   "use_dora": false,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0167e795f1a9a41557a0918c5f94a1dcf4e6efced2069f64d2fc7584073c01ce
+oid sha256:f4598d12ae2d29b14ac23664d4e0a0102ed775da70641565936d1ed06a634374
 size 176183216

all_results.json

@@ -1,21 +1,21 @@
 {
-    "epoch": 4.0,
-    "eval_logits/chosen": -2.171081781387329,
-    "eval_logits/rejected": -2.1663570404052734,
-    "eval_logps/chosen": -33.591148376464844,
-    "eval_logps/rejected": -37.23221206665039,
-    "eval_loss": 0.469946950674057,
-    "eval_rewards/accuracies": 0.5888704061508179,
-    "eval_rewards/chosen": 0.3990616798400879,
-    "eval_rewards/margins": 0.14309579133987427,
-    "eval_rewards/rejected": 0.2559659481048584,
-    "eval_runtime": 145.0614,
+    "epoch": 1.0,
+    "eval_logits/chosen": -2.238292694091797,
+    "eval_logits/rejected": -2.2334160804748535,
+    "eval_logps/chosen": -34.009239196777344,
+    "eval_logps/rejected": -37.498390197753906,
+    "eval_loss": 0.3797042965888977,
+    "eval_rewards/accuracies": 0.5336378812789917,
+    "eval_rewards/chosen": 0.02277994342148304,
+    "eval_rewards/margins": 0.006376237142831087,
+    "eval_rewards/rejected": 0.01640370488166809,
+    "eval_runtime": 145.6289,
     "eval_samples": 343,
-    "eval_samples_per_second": 2.365,
-    "eval_steps_per_second": 0.296,
-    "train_loss": 0.14454748347982183,
-    "train_runtime": 10766.5084,
+    "eval_samples_per_second": 2.355,
+    "eval_steps_per_second": 0.295,
+    "train_loss": 0.31703355428460356,
+    "train_runtime": 3251.5033,
     "train_samples": 3079,
-    "train_samples_per_second": 1.144,
-    "train_steps_per_second": 0.143
+    "train_samples_per_second": 0.947,
+    "train_steps_per_second": 0.118
 }

config.json

@@ -18,8 +18,6 @@
   "n_layer": 28,
   "n_positions": 2048,
   "quantization_config": {
-    "_load_in_4bit": true,
-    "_load_in_8bit": false,
     "bnb_4bit_compute_dtype": "float16",
     "bnb_4bit_quant_type": "nf4",
     "bnb_4bit_use_double_quant": false,
@@ -47,7 +45,7 @@
   },
   "tie_word_embeddings": false,
   "tokenizer_class": "GPT2Tokenizer",
-  "transformers_version": "4.39.0.dev0",
+  "transformers_version": "4.37.2",
   "use_cache": true,
   "vocab_size": 50400
 }

eval_results.json

@@ -1,16 +1,16 @@
 {
-    "epoch": 4.0,
-    "eval_logits/chosen": -2.171081781387329,
-    "eval_logits/rejected": -2.1663570404052734,
-    "eval_logps/chosen": -33.591148376464844,
-    "eval_logps/rejected": -37.23221206665039,
-    "eval_loss": 0.469946950674057,
-    "eval_rewards/accuracies": 0.5888704061508179,
-    "eval_rewards/chosen": 0.3990616798400879,
-    "eval_rewards/margins": 0.14309579133987427,
-    "eval_rewards/rejected": 0.2559659481048584,
-    "eval_runtime": 145.0614,
+    "epoch": 1.0,
+    "eval_logits/chosen": -2.238292694091797,
+    "eval_logits/rejected": -2.2334160804748535,
+    "eval_logps/chosen": -34.009239196777344,
+    "eval_logps/rejected": -37.498390197753906,
+    "eval_loss": 0.3797042965888977,
+    "eval_rewards/accuracies": 0.5336378812789917,
+    "eval_rewards/chosen": 0.02277994342148304,
+    "eval_rewards/margins": 0.006376237142831087,
+    "eval_rewards/rejected": 0.01640370488166809,
+    "eval_runtime": 145.6289,
     "eval_samples": 343,
-    "eval_samples_per_second": 2.365,
-    "eval_steps_per_second": 0.296
+    "eval_samples_per_second": 2.355,
+    "eval_steps_per_second": 0.295
 }

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 4.0,
-    "train_loss": 0.14454748347982183,
-    "train_runtime": 10766.5084,
+    "epoch": 1.0,
+    "train_loss": 0.31703355428460356,
+    "train_runtime": 3251.5033,
     "train_samples": 3079,
-    "train_samples_per_second": 1.144,
-    "train_steps_per_second": 0.143
+    "train_samples_per_second": 0.947,
+    "train_steps_per_second": 0.118
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b90b4eae73710fed677de19c431fb10fd7e1d263ffaac0d06c72d87bc11fe783
+oid sha256:2afba1b889f3ad96a2cd4605a24926c4d3988731b470971ff3b87ea9f7f34337
 size 5176