Training in progress, epoch 13, checkpoint

checkpoint-3250/adapter_config.json

@@ -1,6 +1,6 @@
 {
   "auto_mapping": null,
-  "base_model_name_or_path": "Qwen/Qwen1.5-4B",
+  "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
   "bias": "none",
   "fan_in_fan_out": false,
   "inference_mode": true,

checkpoint-3250/adapter_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83ad5d06ef078727e98443982dfd9cbffc102523282895534875576ea717fbb1
+size 143269386

checkpoint-3250/added_tokens.json

@@ -1,5 +1,5 @@
 {
-  "<|endoftext|>": 151643,
-  "<|im_end|>": 151645,
-  "<|im_start|>": 151644
+  "</s>": 2,
+  "<s>": 1,
+  "<unk>": 0
 }

checkpoint-3250/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dbae6d8f86bbc4ff03ee2e3e66872c101dde0e308662c31b56fe47d4801f221f
-size 224537202
+oid sha256:2a50a5b5737c5bcb964dfe2f43dbfba82021ef3280d8ba30e496b9d466e35620
+size 286585234

checkpoint-3250/rng_state_0.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cbe3eeb85ad6125e9c720931612782bfb33dc653e6508eafa576c018dbac7f6c
+oid sha256:935ebdc5646805a88df3e04672192b4f3ccea9db014368810615dd89e848e699
 size 15024

checkpoint-3250/rng_state_1.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:89cba79818c7c57715915330d6251dab794769fb455234c9dc6d7c2bd2e0d435
+oid sha256:e98a55aae31dcaa87e35253c8fa2ba86844f1a1d487c0b40f3eeb6626bf723e8
 size 15024

checkpoint-3250/rng_state_2.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dbbd9086bcb785da51dcc2ef082d512acba7272152c32ad4fdc7691dfe2a8cfb
+oid sha256:2dd549b56dfe3d7f5f9611ea874d84e0534c3f1503b1e2ef8a07dc23a40cf751
 size 15024

checkpoint-3250/rng_state_3.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ce43dd6b881dd02a4e0a6eab49d4d8400e7e234b4f7472cc6ee52032dcd71428
+oid sha256:9ccdac159ba3210827c066a8e0292fc42f4fd3c83fd649f364187ed380a90e7a
 size 15024

checkpoint-3250/special_tokens_map.json

@@ -1,14 +1,6 @@
 {
-  "additional_special_tokens": [
-    "<|im_start|>",
-    "<|im_end|>"
-  ],
-  "eos_token": {
-    "content": "<|endoftext|>",
-    "lstrip": false,
-    "normalized": false,
-    "rstrip": false,
-    "single_word": false
-  },
-  "pad_token": "<|endoftext|>"
+  "bos_token": "<s>",
+  "eos_token": "</s>",
+  "pad_token": "</s>",
+  "unk_token": "<unk>"
 }

checkpoint-3250/tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
+size 499723

checkpoint-3250/tokenizer_config.json

@@ -1,24 +1,23 @@
 {
-  "add_prefix_space": false,
   "added_tokens_decoder": {
-    "151643": {
-      "content": "<|endoftext|>",
+    "0": {
+      "content": "<unk>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "151644": {
-      "content": "<|im_start|>",
+    "1": {
+      "content": "<s>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "151645": {
-      "content": "<|im_end|>",
+    "2": {
+      "content": "</s>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
@@ -26,18 +25,16 @@
       "special": true
     }
   },
-  "additional_special_tokens": [
-    "<|im_start|>",
-    "<|im_end|>"
-  ],
-  "bos_token": null,
-  "chat_template": "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<|im_start|>system\nYou are a helpful assistant<|im_end|>\n' }}{% endif %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
+  "additional_special_tokens": [],
+  "bos_token": "<s>",
   "clean_up_tokenization_spaces": false,
-  "eos_token": "<|endoftext|>",
-  "errors": "replace",
-  "model_max_length": 32768,
-  "pad_token": "<|endoftext|>",
-  "split_special_tokens": false,
-  "tokenizer_class": "Qwen2Tokenizer",
-  "unk_token": null
+  "eos_token": "</s>",
+  "legacy": false,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "</s>",
+  "padding_side": "left",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": true
 }

checkpoint-3250/trainer_state.json

@@ -10,443 +10,397 @@
   "log_history": [
     {
       "epoch": 0.4,
-      "grad_norm": 0.5897082090377808,
       "learning_rate": 0.0001,
-      "loss": 2.3566,
+      "loss": 2.0446,
       "step": 100
     },
     {
       "epoch": 0.8,
-      "grad_norm": 0.5390245914459229,
       "learning_rate": 0.0001,
-      "loss": 2.2503,
+      "loss": 1.7752,
       "step": 200
     },
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.5155873015873016,
-      "eval_loss": 2.3237216472625732,
-      "eval_runtime": 6.3164,
-      "eval_samples_per_second": 79.159,
-      "eval_steps_per_second": 9.974,
+      "eval_accuracy": 0.6076455696202532,
+      "eval_loss": 1.8067126274108887,
+      "eval_runtime": 5.0018,
+      "eval_samples_per_second": 99.965,
+      "eval_steps_per_second": 12.596,
       "step": 250
     },
     {
       "epoch": 1.0,
-      "eval_exact_match": 17.2,
-      "eval_f1": 25.586581359816655,
+      "eval_exact_match": 24.4,
+      "eval_f1": 35.27913510466144,
       "step": 250
     },
     {
       "epoch": 1.2,
-      "grad_norm": 0.7468857169151306,
       "learning_rate": 0.0001,
-      "loss": 2.1798,
+      "loss": 1.7141,
       "step": 300
     },
     {
       "epoch": 1.6,
-      "grad_norm": 0.9109599590301514,
       "learning_rate": 0.0001,
-      "loss": 2.1094,
+      "loss": 1.6576,
       "step": 400
     },
     {
       "epoch": 2.0,
-      "grad_norm": 0.8458774089813232,
       "learning_rate": 0.0001,
-      "loss": 2.087,
+      "loss": 1.6469,
       "step": 500
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.5163809523809524,
-      "eval_loss": 2.3309242725372314,
-      "eval_runtime": 5.1896,
-      "eval_samples_per_second": 96.347,
-      "eval_steps_per_second": 12.14,
+      "eval_accuracy": 0.607746835443038,
+      "eval_loss": 1.810449481010437,
+      "eval_runtime": 4.5227,
+      "eval_samples_per_second": 110.552,
+      "eval_steps_per_second": 13.93,
       "step": 500
     },
     {
       "epoch": 2.0,
-      "eval_exact_match": 18.6,
-      "eval_f1": 26.043196248196264,
+      "eval_exact_match": 24.2,
+      "eval_f1": 33.990575470281364,
       "step": 500
     },
     {
       "epoch": 2.4,
-      "grad_norm": 1.171212077140808,
       "learning_rate": 0.0001,
-      "loss": 1.8413,
+      "loss": 1.4453,
       "step": 600
     },
     {
       "epoch": 2.8,
-      "grad_norm": 1.1294909715652466,
       "learning_rate": 0.0001,
-      "loss": 1.849,
+      "loss": 1.474,
       "step": 700
     },
     {
       "epoch": 3.0,
-      "eval_accuracy": 0.5144761904761905,
-      "eval_loss": 2.401942253112793,
-      "eval_runtime": 5.5775,
-      "eval_samples_per_second": 89.646,
-      "eval_steps_per_second": 11.295,
+      "eval_accuracy": 0.6055696202531645,
+      "eval_loss": 1.8687865734100342,
+      "eval_runtime": 4.7138,
+      "eval_samples_per_second": 106.072,
+      "eval_steps_per_second": 13.365,
       "step": 750
     },
     {
       "epoch": 3.0,
-      "eval_exact_match": 18.4,
-      "eval_f1": 26.966370002546494,
+      "eval_exact_match": 23.0,
+      "eval_f1": 33.30356143856145,
       "step": 750
     },
     {
       "epoch": 3.2,
-      "grad_norm": 1.5013824701309204,
       "learning_rate": 0.0001,
-      "loss": 1.7322,
+      "loss": 1.3572,
       "step": 800
     },
     {
       "epoch": 3.6,
-      "grad_norm": 1.4516321420669556,
       "learning_rate": 0.0001,
-      "loss": 1.5908,
+      "loss": 1.2472,
       "step": 900
     },
     {
       "epoch": 4.0,
-      "grad_norm": 1.6037466526031494,
       "learning_rate": 0.0001,
-      "loss": 1.6193,
+      "loss": 1.2742,
       "step": 1000
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.5103809523809524,
-      "eval_loss": 2.5039021968841553,
-      "eval_runtime": 5.1568,
-      "eval_samples_per_second": 96.96,
-      "eval_steps_per_second": 12.217,
+      "eval_accuracy": 0.6030886075949367,
+      "eval_loss": 1.9776239395141602,
+      "eval_runtime": 4.5553,
+      "eval_samples_per_second": 109.762,
+      "eval_steps_per_second": 13.83,
       "step": 1000
     },
     {
       "epoch": 4.0,
-      "eval_exact_match": 18.4,
-      "eval_f1": 26.892813852813866,
+      "eval_exact_match": 22.6,
+      "eval_f1": 32.07458596958597,
       "step": 1000
     },
     {
       "epoch": 4.4,
-      "grad_norm": 1.7909170389175415,
       "learning_rate": 0.0001,
-      "loss": 1.3298,
+      "loss": 1.0303,
       "step": 1100
     },
     {
       "epoch": 4.8,
-      "grad_norm": 1.9236161708831787,
       "learning_rate": 0.0001,
-      "loss": 1.3666,
+      "loss": 1.0851,
       "step": 1200
     },
     {
       "epoch": 5.0,
-      "eval_accuracy": 0.505015873015873,
-      "eval_loss": 2.6543924808502197,
-      "eval_runtime": 5.5115,
-      "eval_samples_per_second": 90.719,
-      "eval_steps_per_second": 11.431,
+      "eval_accuracy": 0.599873417721519,
+      "eval_loss": 2.081591844558716,
+      "eval_runtime": 5.0717,
+      "eval_samples_per_second": 98.586,
+      "eval_steps_per_second": 12.422,
       "step": 1250
     },
     {
       "epoch": 5.0,
-      "eval_exact_match": 18.4,
-      "eval_f1": 25.875286609941792,
+      "eval_exact_match": 21.8,
+      "eval_f1": 30.650287424993316,
       "step": 1250
     },
     {
       "epoch": 5.2,
-      "grad_norm": 1.8524874448776245,
       "learning_rate": 0.0001,
-      "loss": 1.2381,
+      "loss": 0.9745,
       "step": 1300
     },
     {
       "epoch": 5.6,
-      "grad_norm": 2.1092071533203125,
       "learning_rate": 0.0001,
-      "loss": 1.1141,
+      "loss": 0.8784,
       "step": 1400
     },
     {
       "epoch": 6.0,
-      "grad_norm": 2.065767288208008,
       "learning_rate": 0.0001,
-      "loss": 1.1435,
+      "loss": 0.9085,
       "step": 1500
     },
     {
       "epoch": 6.0,
-      "eval_accuracy": 0.5010793650793651,
-      "eval_loss": 2.843566656112671,
-      "eval_runtime": 5.1394,
-      "eval_samples_per_second": 97.287,
-      "eval_steps_per_second": 12.258,
+      "eval_accuracy": 0.5969873417721518,
+      "eval_loss": 2.2579739093780518,
+      "eval_runtime": 4.9122,
+      "eval_samples_per_second": 101.788,
+      "eval_steps_per_second": 12.825,
       "step": 1500
     },
     {
       "epoch": 6.0,
-      "eval_exact_match": 17.4,
-      "eval_f1": 24.31514874014875,
+      "eval_exact_match": 21.8,
+      "eval_f1": 30.116953765189066,
       "step": 1500
     },
     {
       "epoch": 6.4,
-      "grad_norm": 2.329406261444092,
       "learning_rate": 0.0001,
-      "loss": 0.8887,
+      "loss": 0.6992,
       "step": 1600
     },
     {
       "epoch": 6.8,
-      "grad_norm": 2.489598274230957,
       "learning_rate": 0.0001,
-      "loss": 0.9171,
+      "loss": 0.7383,
       "step": 1700
     },
     {
       "epoch": 7.0,
-      "eval_accuracy": 0.49707936507936507,
-      "eval_loss": 3.0319864749908447,
-      "eval_runtime": 5.3015,
-      "eval_samples_per_second": 94.313,
-      "eval_steps_per_second": 11.883,
+      "eval_accuracy": 0.5939240506329114,
+      "eval_loss": 2.3878135681152344,
+      "eval_runtime": 4.4125,
+      "eval_samples_per_second": 113.315,
+      "eval_steps_per_second": 14.278,
       "step": 1750
     },
     {
       "epoch": 7.0,
-      "eval_exact_match": 16.0,
-      "eval_f1": 22.824985569985575,
+      "eval_exact_match": 22.2,
+      "eval_f1": 30.4995441160147,
       "step": 1750
     },
     {
       "epoch": 7.2,
-      "grad_norm": 2.764033794403076,
       "learning_rate": 0.0001,
-      "loss": 0.8093,
+      "loss": 0.6582,
       "step": 1800
     },
     {
       "epoch": 7.6,
-      "grad_norm": 2.5643441677093506,
       "learning_rate": 0.0001,
-      "loss": 0.724,
+      "loss": 0.5881,
       "step": 1900
     },
     {
       "epoch": 8.0,
-      "grad_norm": 2.074833631515503,
       "learning_rate": 0.0001,
-      "loss": 0.7531,
+      "loss": 0.6143,
       "step": 2000
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.49304761904761907,
-      "eval_loss": 3.2584712505340576,
-      "eval_runtime": 5.2945,
-      "eval_samples_per_second": 94.438,
-      "eval_steps_per_second": 11.899,
+      "eval_accuracy": 0.5923544303797469,
+      "eval_loss": 2.5656919479370117,
+      "eval_runtime": 4.3988,
+      "eval_samples_per_second": 113.667,
+      "eval_steps_per_second": 14.322,
       "step": 2000
     },
     {
       "epoch": 8.0,
-      "eval_exact_match": 15.0,
-      "eval_f1": 21.895264180264185,
+      "eval_exact_match": 21.8,
+      "eval_f1": 30.237513923984515,
       "step": 2000
     },
     {
       "epoch": 8.4,
-      "grad_norm": 2.649679660797119,
       "learning_rate": 0.0001,
-      "loss": 0.5777,
+      "loss": 0.4727,
       "step": 2100
     },
     {
       "epoch": 8.8,
-      "grad_norm": 2.1969234943389893,
       "learning_rate": 0.0001,
-      "loss": 0.6101,
+      "loss": 0.5047,
       "step": 2200
     },
     {
       "epoch": 9.0,
-      "eval_accuracy": 0.4925079365079365,
-      "eval_loss": 3.3418095111846924,
-      "eval_runtime": 5.1381,
-      "eval_samples_per_second": 97.312,
-      "eval_steps_per_second": 12.261,
+      "eval_accuracy": 0.5898987341772152,
+      "eval_loss": 2.6870768070220947,
+      "eval_runtime": 4.5648,
+      "eval_samples_per_second": 109.534,
+      "eval_steps_per_second": 13.801,
       "step": 2250
     },
     {
       "epoch": 9.0,
-      "eval_exact_match": 14.4,
-      "eval_f1": 21.76202020202021,
+      "eval_exact_match": 21.0,
+      "eval_f1": 29.401002657473246,
       "step": 2250
     },
     {
       "epoch": 9.2,
-      "grad_norm": 1.8318747282028198,
       "learning_rate": 0.0001,
-      "loss": 0.555,
+      "loss": 0.4595,
       "step": 2300
     },
     {
       "epoch": 9.6,
-      "grad_norm": 1.9915804862976074,
       "learning_rate": 0.0001,
-      "loss": 0.5102,
+      "loss": 0.4269,
       "step": 2400
     },
     {
       "epoch": 10.0,
-      "grad_norm": 2.1287789344787598,
       "learning_rate": 0.0001,
-      "loss": 0.5392,
+      "loss": 0.4457,
       "step": 2500
     },
     {
       "epoch": 10.0,
-      "eval_accuracy": 0.4915555555555556,
-      "eval_loss": 3.537275791168213,
-      "eval_runtime": 5.1264,
-      "eval_samples_per_second": 97.534,
-      "eval_steps_per_second": 12.289,
+      "eval_accuracy": 0.5886835443037974,
+      "eval_loss": 2.8029491901397705,
+      "eval_runtime": 4.7023,
+      "eval_samples_per_second": 106.332,
+      "eval_steps_per_second": 13.398,
       "step": 2500
     },
     {
       "epoch": 10.0,
-      "eval_exact_match": 15.4,
-      "eval_f1": 22.23013114336644,
+      "eval_exact_match": 20.8,
+      "eval_f1": 29.73924219571279,
       "step": 2500
     },
     {
       "epoch": 10.4,
-      "grad_norm": 1.6748021841049194,
       "learning_rate": 0.0001,
-      "loss": 0.4426,
+      "loss": 0.3725,
       "step": 2600
     },
     {
       "epoch": 10.8,
-      "grad_norm": 1.8744910955429077,
       "learning_rate": 0.0001,
-      "loss": 0.4718,
+      "loss": 0.3942,
       "step": 2700
     },
     {
       "epoch": 11.0,
-      "eval_accuracy": 0.48926984126984124,
-      "eval_loss": 3.631282091140747,
-      "eval_runtime": 5.3142,
-      "eval_samples_per_second": 94.087,
-      "eval_steps_per_second": 11.855,
+      "eval_accuracy": 0.5877215189873418,
+      "eval_loss": 2.9129838943481445,
+      "eval_runtime": 5.1228,
+      "eval_samples_per_second": 97.603,
+      "eval_steps_per_second": 12.298,
       "step": 2750
     },
     {
       "epoch": 11.0,
-      "eval_exact_match": 15.6,
-      "eval_f1": 21.87386002886003,
+      "eval_exact_match": 17.8,
+      "eval_f1": 27.10153509378855,
       "step": 2750
     },
     {
       "epoch": 11.2,
-      "grad_norm": 2.1398115158081055,
       "learning_rate": 0.0001,
-      "loss": 0.4456,
+      "loss": 0.3736,
       "step": 2800
     },
     {
       "epoch": 11.6,
-      "grad_norm": 1.8851490020751953,
       "learning_rate": 0.0001,
-      "loss": 0.4328,
+      "loss": 0.3612,
       "step": 2900
     },
     {
       "epoch": 12.0,
-      "grad_norm": 1.6958885192871094,
       "learning_rate": 0.0001,
-      "loss": 0.4446,
+      "loss": 0.375,
       "step": 3000
     },
     {
       "epoch": 12.0,
-      "eval_accuracy": 0.49057142857142855,
-      "eval_loss": 3.6736366748809814,
-      "eval_runtime": 5.8652,
-      "eval_samples_per_second": 85.248,
-      "eval_steps_per_second": 10.741,
+      "eval_accuracy": 0.5900759493670886,
+      "eval_loss": 2.9960711002349854,
+      "eval_runtime": 4.5558,
+      "eval_samples_per_second": 109.749,
+      "eval_steps_per_second": 13.828,
       "step": 3000
     },
     {
       "epoch": 12.0,
-      "eval_exact_match": 15.2,
-      "eval_f1": 21.81304966962862,
+      "eval_exact_match": 15.8,
+      "eval_f1": 23.735175071397414,
       "step": 3000
     },
     {
       "epoch": 12.4,
-      "grad_norm": 1.8530863523483276,
       "learning_rate": 0.0001,
-      "loss": 0.3969,
+      "loss": 0.3326,
       "step": 3100
     },
     {
       "epoch": 12.8,
-      "grad_norm": 1.9440696239471436,
       "learning_rate": 0.0001,
-      "loss": 0.4204,
+      "loss": 0.3503,
       "step": 3200
     },
     {
       "epoch": 13.0,
-      "eval_accuracy": 0.4906031746031746,
-      "eval_loss": 3.7342090606689453,
-      "eval_runtime": 5.4947,
-      "eval_samples_per_second": 90.996,
-      "eval_steps_per_second": 11.466,
+      "eval_accuracy": 0.5899493670886076,
+      "eval_loss": 3.0212697982788086,
+      "eval_runtime": 4.4053,
+      "eval_samples_per_second": 113.5,
+      "eval_steps_per_second": 14.301,
       "step": 3250
     },
     {
       "epoch": 13.0,
-      "eval_exact_match": 15.8,
-      "eval_f1": 21.908780663780668,
+      "eval_exact_match": 4.0,
+      "eval_f1": 9.841706398176983,
       "step": 3250
     }
   ],
   "logging_steps": 100,
   "max_steps": 12500,
-  "num_input_tokens_seen": 0,
   "num_train_epochs": 50,
   "save_steps": 500,
-  "stateful_callbacks": {
-    "TrainerControl": {
-      "args": {
-        "should_epoch_stop": false,
-        "should_evaluate": false,
-        "should_log": false,
-        "should_save": true,
-        "should_training_stop": false
-      },
-      "attributes": {}
-    }
-  },
-  "total_flos": 9.941801586655232e+16,
-  "train_batch_size": 2,
+  "total_flos": 2.0780698119543194e+17,
   "trial_name": null,
   "trial_params": null
 }

checkpoint-3250/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c3414153f69372b550d629c900d05ab7510b14d218f2d8546142e7b41d948d3e
-size 5304
+oid sha256:e32f95a567f2c80104b456ab796a180c8c7fa1c0685e00bbceb4995e8fa84ef1
+size 4728