Hugging Face's logo

mlfoundations-dev
/
OH_DCFT_V3_wo_camel_ai_biology

Text Generation
Transformers
Safetensors
llama
llama-factory
full
Generated from Trainer
conversational
text-generation-inference
Inference Endpoints
Model card Files Community
sedrickkeh commited on
Commit
66596b0
·
verified ·
1 Parent(s): 8518533

Training in progress, epoch 2

Browse files
Files changed (13) hide show
  1. model-00001-of-00004.safetensors +1 -1
  2. model-00001-of-00004.safetensors.sagemaker-uploaded +0 -0
  3. model-00002-of-00004.safetensors +1 -1
  4. model-00002-of-00004.safetensors.sagemaker-uploaded +0 -0
  5. model-00003-of-00004.safetensors +1 -1
  6. model-00003-of-00004.safetensors.sagemaker-uploaded +0 -0
  7. model-00004-of-00004.safetensors +1 -1
  8. model-00004-of-00004.safetensors.sagemaker-uploaded +0 -0
  9. special_tokens_map.json.sagemaker-uploaded +0 -0
  10. tokenizer.json.sagemaker-uploaded +0 -0
  11. tokenizer_config.json.sagemaker-uploaded +0 -0
  12. trainer_log.jsonl +83 -0
  13. training_args.bin.sagemaker-uploaded +0 -0
model-00001-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:80d44f74d1e0616789bf3df15e3ed6c6841d20832022f550c363016510774751
3
  size 4976698672
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5f77f9f27244290f5159854204f4ea20fde7400fe45808d6071368f7d83c167c
3
  size 4976698672
model-00001-of-00004.safetensors.sagemaker-uploaded ADDED
File without changes
model-00002-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5acf4f34b634c461f6e78c49e15854b259da701ad83e1cc36e9c8b5374d53bd1
3
  size 4999802720
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:606a7a820ce476f71785f204e8f17e7796a7a18d4ed7e02e55fc77ab916fa6d5
3
  size 4999802720
model-00002-of-00004.safetensors.sagemaker-uploaded ADDED
File without changes
model-00003-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:0e8cd54fb336698fcfded60dd11658567368b3a7070e0357a53fe6e4d9e8b033
3
  size 4915916176
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f85627b6bfcf306bfd3aafe035292c3f0d92feaf3542017d283d1627a65801ad
3
  size 4915916176
model-00003-of-00004.safetensors.sagemaker-uploaded ADDED
File without changes
model-00004-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:0bfdc80be56eb610d9d7ba599cd04a4d8c8e45a85b3f52892fbe3006e8760abf
3
  size 1168138808
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a161a08d29eb85e0aa8ff273b6af9b9bdfb73c121f48e25e6519cb8c57384cad
3
  size 1168138808
model-00004-of-00004.safetensors.sagemaker-uploaded ADDED
File without changes
special_tokens_map.json.sagemaker-uploaded ADDED
File without changes
tokenizer.json.sagemaker-uploaded ADDED
File without changes
tokenizer_config.json.sagemaker-uploaded ADDED
File without changes
trainer_log.jsonl CHANGED
@@ -41,3 +41,86 @@
41
  {"current_steps": 410, "total_steps": 1242, "loss": 0.6408, "learning_rate": 5e-06, "epoch": 0.9885473176612417, "percentage": 33.01, "elapsed_time": "3:19:51", "remaining_time": "6:45:33"}
42
  {"current_steps": 414, "total_steps": 1242, "eval_loss": 0.6481794714927673, "epoch": 0.9981916817359855, "percentage": 33.33, "elapsed_time": "3:25:51", "remaining_time": "6:51:42"}
43
  {"current_steps": 420, "total_steps": 1242, "loss": 0.6219, "learning_rate": 5e-06, "epoch": 1.0126582278481013, "percentage": 33.82, "elapsed_time": "3:29:22", "remaining_time": "6:49:46"}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
41
  {"current_steps": 410, "total_steps": 1242, "loss": 0.6408, "learning_rate": 5e-06, "epoch": 0.9885473176612417, "percentage": 33.01, "elapsed_time": "3:19:51", "remaining_time": "6:45:33"}
42
  {"current_steps": 414, "total_steps": 1242, "eval_loss": 0.6481794714927673, "epoch": 0.9981916817359855, "percentage": 33.33, "elapsed_time": "3:25:51", "remaining_time": "6:51:42"}
43
  {"current_steps": 420, "total_steps": 1242, "loss": 0.6219, "learning_rate": 5e-06, "epoch": 1.0126582278481013, "percentage": 33.82, "elapsed_time": "3:29:22", "remaining_time": "6:49:46"}
44
+ {"current_steps": 430, "total_steps": 1242, "loss": 0.6109, "learning_rate": 5e-06, "epoch": 1.0367691380349608, "percentage": 34.62, "elapsed_time": "3:34:15", "remaining_time": "6:44:35"}
45
+ {"current_steps": 440, "total_steps": 1242, "loss": 0.5999, "learning_rate": 5e-06, "epoch": 1.0608800482218204, "percentage": 35.43, "elapsed_time": "3:39:08", "remaining_time": "6:39:25"}
46
+ {"current_steps": 450, "total_steps": 1242, "loss": 0.6071, "learning_rate": 5e-06, "epoch": 1.0849909584086799, "percentage": 36.23, "elapsed_time": "3:44:01", "remaining_time": "6:34:16"}
47
+ {"current_steps": 460, "total_steps": 1242, "loss": 0.6079, "learning_rate": 5e-06, "epoch": 1.1091018685955394, "percentage": 37.04, "elapsed_time": "3:48:53", "remaining_time": "6:29:07"}
48
+ {"current_steps": 470, "total_steps": 1242, "loss": 0.6093, "learning_rate": 5e-06, "epoch": 1.1332127787823991, "percentage": 37.84, "elapsed_time": "3:53:46", "remaining_time": "6:23:59"}
49
+ {"current_steps": 480, "total_steps": 1242, "loss": 0.6023, "learning_rate": 5e-06, "epoch": 1.1573236889692586, "percentage": 38.65, "elapsed_time": "3:58:39", "remaining_time": "6:18:52"}
50
+ {"current_steps": 490, "total_steps": 1242, "loss": 0.6067, "learning_rate": 5e-06, "epoch": 1.1814345991561181, "percentage": 39.45, "elapsed_time": "4:03:32", "remaining_time": "6:13:45"}
51
+ {"current_steps": 500, "total_steps": 1242, "loss": 0.6034, "learning_rate": 5e-06, "epoch": 1.2055455093429777, "percentage": 40.26, "elapsed_time": "4:08:25", "remaining_time": "6:08:39"}
52
+ {"current_steps": 510, "total_steps": 1242, "loss": 0.6021, "learning_rate": 5e-06, "epoch": 1.2296564195298372, "percentage": 41.06, "elapsed_time": "4:13:17", "remaining_time": "6:03:33"}
53
+ {"current_steps": 520, "total_steps": 1242, "loss": 0.6072, "learning_rate": 5e-06, "epoch": 1.253767329716697, "percentage": 41.87, "elapsed_time": "4:18:10", "remaining_time": "5:58:27"}
54
+ {"current_steps": 530, "total_steps": 1242, "loss": 0.6008, "learning_rate": 5e-06, "epoch": 1.2778782399035564, "percentage": 42.67, "elapsed_time": "4:23:02", "remaining_time": "5:53:21"}
55
+ {"current_steps": 540, "total_steps": 1242, "loss": 0.6084, "learning_rate": 5e-06, "epoch": 1.301989150090416, "percentage": 43.48, "elapsed_time": "4:27:55", "remaining_time": "5:48:17"}
56
+ {"current_steps": 550, "total_steps": 1242, "loss": 0.6073, "learning_rate": 5e-06, "epoch": 1.3261000602772754, "percentage": 44.28, "elapsed_time": "4:32:47", "remaining_time": "5:43:12"}
57
+ {"current_steps": 560, "total_steps": 1242, "loss": 0.6045, "learning_rate": 5e-06, "epoch": 1.350210970464135, "percentage": 45.09, "elapsed_time": "4:37:39", "remaining_time": "5:38:09"}
58
+ {"current_steps": 570, "total_steps": 1242, "loss": 0.601, "learning_rate": 5e-06, "epoch": 1.3743218806509945, "percentage": 45.89, "elapsed_time": "4:42:32", "remaining_time": "5:33:06"}
59
+ {"current_steps": 580, "total_steps": 1242, "loss": 0.6068, "learning_rate": 5e-06, "epoch": 1.3984327908378542, "percentage": 46.7, "elapsed_time": "4:47:25", "remaining_time": "5:28:03"}
60
+ {"current_steps": 590, "total_steps": 1242, "loss": 0.6058, "learning_rate": 5e-06, "epoch": 1.4225437010247137, "percentage": 47.5, "elapsed_time": "4:52:18", "remaining_time": "5:23:01"}
61
+ {"current_steps": 600, "total_steps": 1242, "loss": 0.6084, "learning_rate": 5e-06, "epoch": 1.4466546112115732, "percentage": 48.31, "elapsed_time": "4:57:10", "remaining_time": "5:17:58"}
62
+ {"current_steps": 610, "total_steps": 1242, "loss": 0.6083, "learning_rate": 5e-06, "epoch": 1.4707655213984328, "percentage": 49.11, "elapsed_time": "5:02:03", "remaining_time": "5:12:56"}
63
+ {"current_steps": 620, "total_steps": 1242, "loss": 0.6101, "learning_rate": 5e-06, "epoch": 1.4948764315852923, "percentage": 49.92, "elapsed_time": "5:06:56", "remaining_time": "5:07:55"}
64
+ {"current_steps": 630, "total_steps": 1242, "loss": 0.6107, "learning_rate": 5e-06, "epoch": 1.518987341772152, "percentage": 50.72, "elapsed_time": "5:11:48", "remaining_time": "5:02:54"}
65
+ {"current_steps": 640, "total_steps": 1242, "loss": 0.5995, "learning_rate": 5e-06, "epoch": 1.5430982519590115, "percentage": 51.53, "elapsed_time": "5:16:40", "remaining_time": "4:57:52"}
66
+ {"current_steps": 650, "total_steps": 1242, "loss": 0.6147, "learning_rate": 5e-06, "epoch": 1.567209162145871, "percentage": 52.33, "elapsed_time": "5:21:34", "remaining_time": "4:52:52"}
67
+ {"current_steps": 660, "total_steps": 1242, "loss": 0.6046, "learning_rate": 5e-06, "epoch": 1.5913200723327305, "percentage": 53.14, "elapsed_time": "5:26:26", "remaining_time": "4:47:51"}
68
+ {"current_steps": 670, "total_steps": 1242, "loss": 0.605, "learning_rate": 5e-06, "epoch": 1.61543098251959, "percentage": 53.95, "elapsed_time": "5:31:19", "remaining_time": "4:42:52"}
69
+ {"current_steps": 680, "total_steps": 1242, "loss": 0.603, "learning_rate": 5e-06, "epoch": 1.6395418927064496, "percentage": 54.75, "elapsed_time": "5:36:12", "remaining_time": "4:37:51"}
70
+ {"current_steps": 690, "total_steps": 1242, "loss": 0.6037, "learning_rate": 5e-06, "epoch": 1.663652802893309, "percentage": 55.56, "elapsed_time": "5:41:04", "remaining_time": "4:32:51"}
71
+ {"current_steps": 700, "total_steps": 1242, "loss": 0.6037, "learning_rate": 5e-06, "epoch": 1.6877637130801688, "percentage": 56.36, "elapsed_time": "5:45:56", "remaining_time": "4:27:51"}
72
+ {"current_steps": 710, "total_steps": 1242, "loss": 0.6046, "learning_rate": 5e-06, "epoch": 1.7118746232670283, "percentage": 57.17, "elapsed_time": "5:50:49", "remaining_time": "4:22:52"}
73
+ {"current_steps": 720, "total_steps": 1242, "loss": 0.6057, "learning_rate": 5e-06, "epoch": 1.7359855334538878, "percentage": 57.97, "elapsed_time": "5:55:42", "remaining_time": "4:17:53"}
74
+ {"current_steps": 730, "total_steps": 1242, "loss": 0.6066, "learning_rate": 5e-06, "epoch": 1.7600964436407476, "percentage": 58.78, "elapsed_time": "6:00:34", "remaining_time": "4:12:54"}
75
+ {"current_steps": 740, "total_steps": 1242, "loss": 0.6, "learning_rate": 5e-06, "epoch": 1.784207353827607, "percentage": 59.58, "elapsed_time": "6:05:27", "remaining_time": "4:07:55"}
76
+ {"current_steps": 750, "total_steps": 1242, "loss": 0.6098, "learning_rate": 5e-06, "epoch": 1.8083182640144666, "percentage": 60.39, "elapsed_time": "6:10:20", "remaining_time": "4:02:56"}
77
+ {"current_steps": 760, "total_steps": 1242, "loss": 0.6054, "learning_rate": 5e-06, "epoch": 1.8324291742013261, "percentage": 61.19, "elapsed_time": "6:15:13", "remaining_time": "3:57:57"}
78
+ {"current_steps": 770, "total_steps": 1242, "loss": 0.6037, "learning_rate": 5e-06, "epoch": 1.8565400843881856, "percentage": 62.0, "elapsed_time": "6:20:05", "remaining_time": "3:52:59"}
79
+ {"current_steps": 780, "total_steps": 1242, "loss": 0.6019, "learning_rate": 5e-06, "epoch": 1.8806509945750451, "percentage": 62.8, "elapsed_time": "6:24:58", "remaining_time": "3:48:01"}
80
+ {"current_steps": 790, "total_steps": 1242, "loss": 0.6051, "learning_rate": 5e-06, "epoch": 1.9047619047619047, "percentage": 63.61, "elapsed_time": "6:29:49", "remaining_time": "3:43:02"}
81
+ {"current_steps": 800, "total_steps": 1242, "loss": 0.6052, "learning_rate": 5e-06, "epoch": 1.9288728149487642, "percentage": 64.41, "elapsed_time": "6:34:42", "remaining_time": "3:38:04"}
82
+ {"current_steps": 810, "total_steps": 1242, "loss": 0.598, "learning_rate": 5e-06, "epoch": 1.952983725135624, "percentage": 65.22, "elapsed_time": "6:39:34", "remaining_time": "3:33:06"}
83
+ {"current_steps": 820, "total_steps": 1242, "loss": 0.6026, "learning_rate": 5e-06, "epoch": 1.9770946353224834, "percentage": 66.02, "elapsed_time": "6:44:27", "remaining_time": "3:28:08"}
84
+ {"current_steps": 829, "total_steps": 1242, "eval_loss": 0.6378007531166077, "epoch": 1.998794454490657, "percentage": 66.75, "elapsed_time": "6:52:46", "remaining_time": "3:25:38"}
85
+ {"current_steps": 830, "total_steps": 1242, "loss": 0.6043, "learning_rate": 5e-06, "epoch": 2.001205545509343, "percentage": 66.83, "elapsed_time": "6:54:02", "remaining_time": "3:25:31"}
86
+ {"current_steps": 840, "total_steps": 1242, "loss": 0.5707, "learning_rate": 5e-06, "epoch": 2.0253164556962027, "percentage": 67.63, "elapsed_time": "6:58:55", "remaining_time": "3:20:29"}
87
+ {"current_steps": 850, "total_steps": 1242, "loss": 0.5606, "learning_rate": 5e-06, "epoch": 2.049427365883062, "percentage": 68.44, "elapsed_time": "7:03:48", "remaining_time": "3:15:27"}
88
+ {"current_steps": 860, "total_steps": 1242, "loss": 0.563, "learning_rate": 5e-06, "epoch": 2.0735382760699217, "percentage": 69.24, "elapsed_time": "7:08:41", "remaining_time": "3:10:25"}
89
+ {"current_steps": 870, "total_steps": 1242, "loss": 0.5631, "learning_rate": 5e-06, "epoch": 2.097649186256781, "percentage": 70.05, "elapsed_time": "7:13:33", "remaining_time": "3:05:23"}
90
+ {"current_steps": 880, "total_steps": 1242, "loss": 0.5595, "learning_rate": 5e-06, "epoch": 2.1217600964436407, "percentage": 70.85, "elapsed_time": "7:18:27", "remaining_time": "3:00:21"}
91
+ {"current_steps": 890, "total_steps": 1242, "loss": 0.5601, "learning_rate": 5e-06, "epoch": 2.1458710066305002, "percentage": 71.66, "elapsed_time": "7:23:19", "remaining_time": "2:55:20"}
92
+ {"current_steps": 900, "total_steps": 1242, "loss": 0.5607, "learning_rate": 5e-06, "epoch": 2.1699819168173597, "percentage": 72.46, "elapsed_time": "7:28:12", "remaining_time": "2:50:19"}
93
+ {"current_steps": 910, "total_steps": 1242, "loss": 0.5595, "learning_rate": 5e-06, "epoch": 2.1940928270042193, "percentage": 73.27, "elapsed_time": "7:33:05", "remaining_time": "2:45:18"}
94
+ {"current_steps": 920, "total_steps": 1242, "loss": 0.5677, "learning_rate": 5e-06, "epoch": 2.2182037371910788, "percentage": 74.07, "elapsed_time": "7:37:58", "remaining_time": "2:40:17"}
95
+ {"current_steps": 930, "total_steps": 1242, "loss": 0.5594, "learning_rate": 5e-06, "epoch": 2.2423146473779383, "percentage": 74.88, "elapsed_time": "7:42:50", "remaining_time": "2:35:16"}
96
+ {"current_steps": 940, "total_steps": 1242, "loss": 0.5621, "learning_rate": 5e-06, "epoch": 2.2664255575647982, "percentage": 75.68, "elapsed_time": "7:47:42", "remaining_time": "2:30:15"}
97
+ {"current_steps": 950, "total_steps": 1242, "loss": 0.5638, "learning_rate": 5e-06, "epoch": 2.2905364677516578, "percentage": 76.49, "elapsed_time": "7:52:35", "remaining_time": "2:25:15"}
98
+ {"current_steps": 960, "total_steps": 1242, "loss": 0.564, "learning_rate": 5e-06, "epoch": 2.3146473779385173, "percentage": 77.29, "elapsed_time": "7:57:27", "remaining_time": "2:20:15"}
99
+ {"current_steps": 970, "total_steps": 1242, "loss": 0.5643, "learning_rate": 5e-06, "epoch": 2.338758288125377, "percentage": 78.1, "elapsed_time": "8:02:20", "remaining_time": "2:15:15"}
100
+ {"current_steps": 980, "total_steps": 1242, "loss": 0.5642, "learning_rate": 5e-06, "epoch": 2.3628691983122363, "percentage": 78.9, "elapsed_time": "8:07:13", "remaining_time": "2:10:15"}
101
+ {"current_steps": 990, "total_steps": 1242, "loss": 0.5672, "learning_rate": 5e-06, "epoch": 2.386980108499096, "percentage": 79.71, "elapsed_time": "8:12:05", "remaining_time": "2:05:15"}
102
+ {"current_steps": 1000, "total_steps": 1242, "loss": 0.5565, "learning_rate": 5e-06, "epoch": 2.4110910186859553, "percentage": 80.52, "elapsed_time": "8:16:58", "remaining_time": "2:00:16"}
103
+ {"current_steps": 1010, "total_steps": 1242, "loss": 0.5657, "learning_rate": 5e-06, "epoch": 2.435201928872815, "percentage": 81.32, "elapsed_time": "8:21:51", "remaining_time": "1:55:16"}
104
+ {"current_steps": 1020, "total_steps": 1242, "loss": 0.5586, "learning_rate": 5e-06, "epoch": 2.4593128390596743, "percentage": 82.13, "elapsed_time": "8:26:44", "remaining_time": "1:50:17"}
105
+ {"current_steps": 1030, "total_steps": 1242, "loss": 0.5576, "learning_rate": 5e-06, "epoch": 2.483423749246534, "percentage": 82.93, "elapsed_time": "8:31:37", "remaining_time": "1:45:18"}
106
+ {"current_steps": 1040, "total_steps": 1242, "loss": 0.5681, "learning_rate": 5e-06, "epoch": 2.507534659433394, "percentage": 83.74, "elapsed_time": "8:36:29", "remaining_time": "1:40:19"}
107
+ {"current_steps": 1050, "total_steps": 1242, "loss": 0.5659, "learning_rate": 5e-06, "epoch": 2.5316455696202533, "percentage": 84.54, "elapsed_time": "8:41:22", "remaining_time": "1:35:20"}
108
+ {"current_steps": 1060, "total_steps": 1242, "loss": 0.5667, "learning_rate": 5e-06, "epoch": 2.555756479807113, "percentage": 85.35, "elapsed_time": "8:46:15", "remaining_time": "1:30:21"}
109
+ {"current_steps": 1070, "total_steps": 1242, "loss": 0.5629, "learning_rate": 5e-06, "epoch": 2.5798673899939724, "percentage": 86.15, "elapsed_time": "8:51:08", "remaining_time": "1:25:22"}
110
+ {"current_steps": 1080, "total_steps": 1242, "loss": 0.5687, "learning_rate": 5e-06, "epoch": 2.603978300180832, "percentage": 86.96, "elapsed_time": "8:56:02", "remaining_time": "1:20:24"}
111
+ {"current_steps": 1090, "total_steps": 1242, "loss": 0.5692, "learning_rate": 5e-06, "epoch": 2.6280892103676914, "percentage": 87.76, "elapsed_time": "9:00:55", "remaining_time": "1:15:25"}
112
+ {"current_steps": 1100, "total_steps": 1242, "loss": 0.5667, "learning_rate": 5e-06, "epoch": 2.652200120554551, "percentage": 88.57, "elapsed_time": "9:05:47", "remaining_time": "1:10:27"}
113
+ {"current_steps": 1110, "total_steps": 1242, "loss": 0.5617, "learning_rate": 5e-06, "epoch": 2.6763110307414104, "percentage": 89.37, "elapsed_time": "9:10:40", "remaining_time": "1:05:29"}
114
+ {"current_steps": 1120, "total_steps": 1242, "loss": 0.5589, "learning_rate": 5e-06, "epoch": 2.70042194092827, "percentage": 90.18, "elapsed_time": "9:15:33", "remaining_time": "1:00:31"}
115
+ {"current_steps": 1130, "total_steps": 1242, "loss": 0.5696, "learning_rate": 5e-06, "epoch": 2.7245328511151294, "percentage": 90.98, "elapsed_time": "9:20:26", "remaining_time": "0:55:32"}
116
+ {"current_steps": 1140, "total_steps": 1242, "loss": 0.5645, "learning_rate": 5e-06, "epoch": 2.748643761301989, "percentage": 91.79, "elapsed_time": "9:25:20", "remaining_time": "0:50:34"}
117
+ {"current_steps": 1150, "total_steps": 1242, "loss": 0.5659, "learning_rate": 5e-06, "epoch": 2.7727546714888485, "percentage": 92.59, "elapsed_time": "9:30:13", "remaining_time": "0:45:37"}
118
+ {"current_steps": 1160, "total_steps": 1242, "loss": 0.5638, "learning_rate": 5e-06, "epoch": 2.7968655816757084, "percentage": 93.4, "elapsed_time": "9:35:06", "remaining_time": "0:40:39"}
119
+ {"current_steps": 1170, "total_steps": 1242, "loss": 0.5651, "learning_rate": 5e-06, "epoch": 2.820976491862568, "percentage": 94.2, "elapsed_time": "9:39:58", "remaining_time": "0:35:41"}
120
+ {"current_steps": 1180, "total_steps": 1242, "loss": 0.5664, "learning_rate": 5e-06, "epoch": 2.8450874020494274, "percentage": 95.01, "elapsed_time": "9:44:51", "remaining_time": "0:30:43"}
121
+ {"current_steps": 1190, "total_steps": 1242, "loss": 0.5708, "learning_rate": 5e-06, "epoch": 2.869198312236287, "percentage": 95.81, "elapsed_time": "9:49:44", "remaining_time": "0:25:46"}
122
+ {"current_steps": 1200, "total_steps": 1242, "loss": 0.5713, "learning_rate": 5e-06, "epoch": 2.8933092224231465, "percentage": 96.62, "elapsed_time": "9:54:37", "remaining_time": "0:20:48"}
123
+ {"current_steps": 1210, "total_steps": 1242, "loss": 0.566, "learning_rate": 5e-06, "epoch": 2.917420132610006, "percentage": 97.42, "elapsed_time": "9:59:30", "remaining_time": "0:15:51"}
124
+ {"current_steps": 1220, "total_steps": 1242, "loss": 0.5653, "learning_rate": 5e-06, "epoch": 2.9415310427968655, "percentage": 98.23, "elapsed_time": "10:04:23", "remaining_time": "0:10:53"}
125
+ {"current_steps": 1230, "total_steps": 1242, "loss": 0.5663, "learning_rate": 5e-06, "epoch": 2.965641952983725, "percentage": 99.03, "elapsed_time": "10:09:15", "remaining_time": "0:05:56"}
126
+ {"current_steps": 1240, "total_steps": 1242, "loss": 0.569, "learning_rate": 5e-06, "epoch": 2.9897528631705845, "percentage": 99.84, "elapsed_time": "10:14:09", "remaining_time": "0:00:59"}
training_args.bin.sagemaker-uploaded ADDED
File without changes