Hugging Face's logo

mlfoundations-dev
/
orca-agentinstruct-1M-v1-cleaned

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
4f7f79e
•
1 Parent(s): a8aec59

Training in progress, epoch 2

Browse files
Files changed (14) hide show
  1. config.json.sagemaker-uploaded +0 -0
  2. model-00001-of-00004.safetensors +1 -1
  3. model-00001-of-00004.safetensors.sagemaker-uploaded +0 -0
  4. model-00002-of-00004.safetensors +1 -1
  5. model-00002-of-00004.safetensors.sagemaker-uploaded +0 -0
  6. model-00003-of-00004.safetensors +1 -1
  7. model-00003-of-00004.safetensors.sagemaker-uploaded +0 -0
  8. model-00004-of-00004.safetensors +1 -1
  9. model-00004-of-00004.safetensors.sagemaker-uploaded +0 -0
  10. special_tokens_map.json.sagemaker-uploaded +0 -0
  11. tokenizer.json.sagemaker-uploaded +0 -0
  12. tokenizer_config.json.sagemaker-uploaded +0 -0
  13. trainer_log.jsonl +181 -0
  14. training_args.bin.sagemaker-uploaded +0 -0
config.json.sagemaker-uploaded ADDED
File without changes
model-00001-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:00b54cb4cd86538338390a19ef2c4df6f1d43da7ac1a2374c2fb568e9e4f1061
3
  size 4976698672
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1eb7fa6e80100bb4d095a61c838d58f80e85a72715c3c3ffccd99c284f8be3d5
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 sha256:5a5674551acfaae82b03d16f0433de2435c0489d53fa33da0550afd887d244b3
3
  size 4999802720
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f58572e77952b11074eb4d9d4572a736ec6dea4ccd0c05c713080e2895d28832
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:f3a0d3949e83c65ffd63bbdc011b1acdbcca8dce1dd3185713fddc3a5c1b237c
3
  size 4915916176
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2dd6d89c71e19d7d0785fd9529306df42602860d5a016f21f5425e72a949bb6c
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:f196e1b5fcd2daf8699bcbc599a8cde54c1966f8601b3a15c083491631e9d37b
3
  size 1168138808
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2043684ea378005a48049cfc4001e312eb16f4ed952107505670731c91ce8351
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
@@ -91,3 +91,184 @@
91
  {"current_steps": 908, "total_steps": 2724, "eval_loss": 0.4856513738632202, "epoch": 0.9994496422674739, "percentage": 33.33, "elapsed_time": "2:09:20", "remaining_time": "4:18:41"}
92
  {"current_steps": 910, "total_steps": 2724, "loss": 0.481, "learning_rate": 5e-06, "epoch": 1.0016510731975785, "percentage": 33.41, "elapsed_time": "2:10:26", "remaining_time": "4:20:00"}
93
  {"current_steps": 920, "total_steps": 2724, "loss": 0.4391, "learning_rate": 5e-06, "epoch": 1.0126582278481013, "percentage": 33.77, "elapsed_time": "2:11:48", "remaining_time": "4:18:28"}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
91
  {"current_steps": 908, "total_steps": 2724, "eval_loss": 0.4856513738632202, "epoch": 0.9994496422674739, "percentage": 33.33, "elapsed_time": "2:09:20", "remaining_time": "4:18:41"}
92
  {"current_steps": 910, "total_steps": 2724, "loss": 0.481, "learning_rate": 5e-06, "epoch": 1.0016510731975785, "percentage": 33.41, "elapsed_time": "2:10:26", "remaining_time": "4:20:00"}
93
  {"current_steps": 920, "total_steps": 2724, "loss": 0.4391, "learning_rate": 5e-06, "epoch": 1.0126582278481013, "percentage": 33.77, "elapsed_time": "2:11:48", "remaining_time": "4:18:28"}
94
+ {"current_steps": 930, "total_steps": 2724, "loss": 0.4375, "learning_rate": 5e-06, "epoch": 1.0236653824986242, "percentage": 34.14, "elapsed_time": "2:13:12", "remaining_time": "4:16:57"}
95
+ {"current_steps": 940, "total_steps": 2724, "loss": 0.4384, "learning_rate": 5e-06, "epoch": 1.034672537149147, "percentage": 34.51, "elapsed_time": "2:14:35", "remaining_time": "4:15:26"}
96
+ {"current_steps": 950, "total_steps": 2724, "loss": 0.4369, "learning_rate": 5e-06, "epoch": 1.0456796917996698, "percentage": 34.88, "elapsed_time": "2:15:58", "remaining_time": "4:13:55"}
97
+ {"current_steps": 960, "total_steps": 2724, "loss": 0.4355, "learning_rate": 5e-06, "epoch": 1.0566868464501926, "percentage": 35.24, "elapsed_time": "2:17:22", "remaining_time": "4:12:25"}
98
+ {"current_steps": 970, "total_steps": 2724, "loss": 0.4355, "learning_rate": 5e-06, "epoch": 1.0676940011007154, "percentage": 35.61, "elapsed_time": "2:18:45", "remaining_time": "4:10:55"}
99
+ {"current_steps": 980, "total_steps": 2724, "loss": 0.438, "learning_rate": 5e-06, "epoch": 1.0787011557512383, "percentage": 35.98, "elapsed_time": "2:20:09", "remaining_time": "4:09:25"}
100
+ {"current_steps": 990, "total_steps": 2724, "loss": 0.4382, "learning_rate": 5e-06, "epoch": 1.089708310401761, "percentage": 36.34, "elapsed_time": "2:21:32", "remaining_time": "4:07:55"}
101
+ {"current_steps": 1000, "total_steps": 2724, "loss": 0.4402, "learning_rate": 5e-06, "epoch": 1.100715465052284, "percentage": 36.71, "elapsed_time": "2:22:56", "remaining_time": "4:06:25"}
102
+ {"current_steps": 1010, "total_steps": 2724, "loss": 0.4369, "learning_rate": 5e-06, "epoch": 1.1117226197028067, "percentage": 37.08, "elapsed_time": "2:24:19", "remaining_time": "4:04:55"}
103
+ {"current_steps": 1020, "total_steps": 2724, "loss": 0.4343, "learning_rate": 5e-06, "epoch": 1.1227297743533298, "percentage": 37.44, "elapsed_time": "2:25:43", "remaining_time": "4:03:27"}
104
+ {"current_steps": 1030, "total_steps": 2724, "loss": 0.4407, "learning_rate": 5e-06, "epoch": 1.1337369290038526, "percentage": 37.81, "elapsed_time": "2:27:07", "remaining_time": "4:01:58"}
105
+ {"current_steps": 1040, "total_steps": 2724, "loss": 0.438, "learning_rate": 5e-06, "epoch": 1.1447440836543754, "percentage": 38.18, "elapsed_time": "2:28:31", "remaining_time": "4:00:30"}
106
+ {"current_steps": 1050, "total_steps": 2724, "loss": 0.4382, "learning_rate": 5e-06, "epoch": 1.1557512383048982, "percentage": 38.55, "elapsed_time": "2:29:55", "remaining_time": "3:59:00"}
107
+ {"current_steps": 1060, "total_steps": 2724, "loss": 0.442, "learning_rate": 5e-06, "epoch": 1.166758392955421, "percentage": 38.91, "elapsed_time": "2:31:18", "remaining_time": "3:57:31"}
108
+ {"current_steps": 1070, "total_steps": 2724, "loss": 0.4338, "learning_rate": 5e-06, "epoch": 1.1777655476059439, "percentage": 39.28, "elapsed_time": "2:32:41", "remaining_time": "3:56:02"}
109
+ {"current_steps": 1080, "total_steps": 2724, "loss": 0.435, "learning_rate": 5e-06, "epoch": 1.1887727022564667, "percentage": 39.65, "elapsed_time": "2:34:04", "remaining_time": "3:54:32"}
110
+ {"current_steps": 1090, "total_steps": 2724, "loss": 0.4422, "learning_rate": 5e-06, "epoch": 1.1997798569069895, "percentage": 40.01, "elapsed_time": "2:35:28", "remaining_time": "3:53:03"}
111
+ {"current_steps": 1100, "total_steps": 2724, "loss": 0.4372, "learning_rate": 5e-06, "epoch": 1.2107870115575123, "percentage": 40.38, "elapsed_time": "2:36:51", "remaining_time": "3:51:34"}
112
+ {"current_steps": 1110, "total_steps": 2724, "loss": 0.4321, "learning_rate": 5e-06, "epoch": 1.2217941662080352, "percentage": 40.75, "elapsed_time": "2:38:14", "remaining_time": "3:50:05"}
113
+ {"current_steps": 1120, "total_steps": 2724, "loss": 0.439, "learning_rate": 5e-06, "epoch": 1.232801320858558, "percentage": 41.12, "elapsed_time": "2:39:36", "remaining_time": "3:48:35"}
114
+ {"current_steps": 1130, "total_steps": 2724, "loss": 0.4391, "learning_rate": 5e-06, "epoch": 1.243808475509081, "percentage": 41.48, "elapsed_time": "2:41:00", "remaining_time": "3:47:06"}
115
+ {"current_steps": 1140, "total_steps": 2724, "loss": 0.4358, "learning_rate": 5e-06, "epoch": 1.2548156301596038, "percentage": 41.85, "elapsed_time": "2:42:23", "remaining_time": "3:45:38"}
116
+ {"current_steps": 1150, "total_steps": 2724, "loss": 0.4357, "learning_rate": 5e-06, "epoch": 1.2658227848101267, "percentage": 42.22, "elapsed_time": "2:43:46", "remaining_time": "3:44:09"}
117
+ {"current_steps": 1160, "total_steps": 2724, "loss": 0.4377, "learning_rate": 5e-06, "epoch": 1.2768299394606495, "percentage": 42.58, "elapsed_time": "2:45:10", "remaining_time": "3:42:41"}
118
+ {"current_steps": 1170, "total_steps": 2724, "loss": 0.4373, "learning_rate": 5e-06, "epoch": 1.2878370941111723, "percentage": 42.95, "elapsed_time": "2:46:33", "remaining_time": "3:41:13"}
119
+ {"current_steps": 1180, "total_steps": 2724, "loss": 0.4397, "learning_rate": 5e-06, "epoch": 1.2988442487616951, "percentage": 43.32, "elapsed_time": "2:47:56", "remaining_time": "3:39:45"}
120
+ {"current_steps": 1190, "total_steps": 2724, "loss": 0.4317, "learning_rate": 5e-06, "epoch": 1.309851403412218, "percentage": 43.69, "elapsed_time": "2:49:19", "remaining_time": "3:38:16"}
121
+ {"current_steps": 1200, "total_steps": 2724, "loss": 0.4388, "learning_rate": 5e-06, "epoch": 1.3208585580627408, "percentage": 44.05, "elapsed_time": "2:50:43", "remaining_time": "3:36:48"}
122
+ {"current_steps": 1210, "total_steps": 2724, "loss": 0.4296, "learning_rate": 5e-06, "epoch": 1.3318657127132636, "percentage": 44.42, "elapsed_time": "2:52:06", "remaining_time": "3:35:20"}
123
+ {"current_steps": 1220, "total_steps": 2724, "loss": 0.4399, "learning_rate": 5e-06, "epoch": 1.3428728673637864, "percentage": 44.79, "elapsed_time": "2:53:29", "remaining_time": "3:33:53"}
124
+ {"current_steps": 1230, "total_steps": 2724, "loss": 0.4364, "learning_rate": 5e-06, "epoch": 1.3538800220143092, "percentage": 45.15, "elapsed_time": "2:54:53", "remaining_time": "3:32:25"}
125
+ {"current_steps": 1240, "total_steps": 2724, "loss": 0.4379, "learning_rate": 5e-06, "epoch": 1.364887176664832, "percentage": 45.52, "elapsed_time": "2:56:16", "remaining_time": "3:30:57"}
126
+ {"current_steps": 1250, "total_steps": 2724, "loss": 0.4417, "learning_rate": 5e-06, "epoch": 1.3758943313153549, "percentage": 45.89, "elapsed_time": "2:57:39", "remaining_time": "3:29:29"}
127
+ {"current_steps": 1260, "total_steps": 2724, "loss": 0.44, "learning_rate": 5e-06, "epoch": 1.3869014859658777, "percentage": 46.26, "elapsed_time": "2:59:03", "remaining_time": "3:28:02"}
128
+ {"current_steps": 1270, "total_steps": 2724, "loss": 0.438, "learning_rate": 5e-06, "epoch": 1.3979086406164005, "percentage": 46.62, "elapsed_time": "3:00:26", "remaining_time": "3:26:34"}
129
+ {"current_steps": 1280, "total_steps": 2724, "loss": 0.4354, "learning_rate": 5e-06, "epoch": 1.4089157952669236, "percentage": 46.99, "elapsed_time": "3:01:49", "remaining_time": "3:25:07"}
130
+ {"current_steps": 1290, "total_steps": 2724, "loss": 0.4364, "learning_rate": 5e-06, "epoch": 1.4199229499174464, "percentage": 47.36, "elapsed_time": "3:03:13", "remaining_time": "3:23:40"}
131
+ {"current_steps": 1300, "total_steps": 2724, "loss": 0.433, "learning_rate": 5e-06, "epoch": 1.4309301045679692, "percentage": 47.72, "elapsed_time": "3:04:36", "remaining_time": "3:22:12"}
132
+ {"current_steps": 1310, "total_steps": 2724, "loss": 0.4305, "learning_rate": 5e-06, "epoch": 1.441937259218492, "percentage": 48.09, "elapsed_time": "3:05:59", "remaining_time": "3:20:45"}
133
+ {"current_steps": 1320, "total_steps": 2724, "loss": 0.4372, "learning_rate": 5e-06, "epoch": 1.4529444138690149, "percentage": 48.46, "elapsed_time": "3:07:23", "remaining_time": "3:19:18"}
134
+ {"current_steps": 1330, "total_steps": 2724, "loss": 0.4317, "learning_rate": 5e-06, "epoch": 1.4639515685195377, "percentage": 48.83, "elapsed_time": "3:08:46", "remaining_time": "3:17:51"}
135
+ {"current_steps": 1340, "total_steps": 2724, "loss": 0.4393, "learning_rate": 5e-06, "epoch": 1.4749587231700605, "percentage": 49.19, "elapsed_time": "3:10:09", "remaining_time": "3:16:24"}
136
+ {"current_steps": 1350, "total_steps": 2724, "loss": 0.4341, "learning_rate": 5e-06, "epoch": 1.4859658778205833, "percentage": 49.56, "elapsed_time": "3:11:32", "remaining_time": "3:14:57"}
137
+ {"current_steps": 1360, "total_steps": 2724, "loss": 0.4356, "learning_rate": 5e-06, "epoch": 1.4969730324711064, "percentage": 49.93, "elapsed_time": "3:12:55", "remaining_time": "3:13:29"}
138
+ {"current_steps": 1370, "total_steps": 2724, "loss": 0.4362, "learning_rate": 5e-06, "epoch": 1.5079801871216292, "percentage": 50.29, "elapsed_time": "3:14:19", "remaining_time": "3:12:02"}
139
+ {"current_steps": 1380, "total_steps": 2724, "loss": 0.4382, "learning_rate": 5e-06, "epoch": 1.518987341772152, "percentage": 50.66, "elapsed_time": "3:15:43", "remaining_time": "3:10:36"}
140
+ {"current_steps": 1390, "total_steps": 2724, "loss": 0.4273, "learning_rate": 5e-06, "epoch": 1.5299944964226748, "percentage": 51.03, "elapsed_time": "3:17:07", "remaining_time": "3:09:10"}
141
+ {"current_steps": 1400, "total_steps": 2724, "loss": 0.4299, "learning_rate": 5e-06, "epoch": 1.5410016510731976, "percentage": 51.4, "elapsed_time": "3:18:31", "remaining_time": "3:07:44"}
142
+ {"current_steps": 1410, "total_steps": 2724, "loss": 0.4341, "learning_rate": 5e-06, "epoch": 1.5520088057237205, "percentage": 51.76, "elapsed_time": "3:19:55", "remaining_time": "3:06:18"}
143
+ {"current_steps": 1420, "total_steps": 2724, "loss": 0.4365, "learning_rate": 5e-06, "epoch": 1.5630159603742433, "percentage": 52.13, "elapsed_time": "3:21:19", "remaining_time": "3:04:52"}
144
+ {"current_steps": 1430, "total_steps": 2724, "loss": 0.4351, "learning_rate": 5e-06, "epoch": 1.574023115024766, "percentage": 52.5, "elapsed_time": "3:22:42", "remaining_time": "3:03:26"}
145
+ {"current_steps": 1440, "total_steps": 2724, "loss": 0.4331, "learning_rate": 5e-06, "epoch": 1.585030269675289, "percentage": 52.86, "elapsed_time": "3:24:06", "remaining_time": "3:01:59"}
146
+ {"current_steps": 1450, "total_steps": 2724, "loss": 0.4292, "learning_rate": 5e-06, "epoch": 1.5960374243258117, "percentage": 53.23, "elapsed_time": "3:25:29", "remaining_time": "3:00:33"}
147
+ {"current_steps": 1460, "total_steps": 2724, "loss": 0.4366, "learning_rate": 5e-06, "epoch": 1.6070445789763346, "percentage": 53.6, "elapsed_time": "3:26:53", "remaining_time": "2:59:07"}
148
+ {"current_steps": 1470, "total_steps": 2724, "loss": 0.4361, "learning_rate": 5e-06, "epoch": 1.6180517336268574, "percentage": 53.96, "elapsed_time": "3:28:17", "remaining_time": "2:57:40"}
149
+ {"current_steps": 1480, "total_steps": 2724, "loss": 0.4308, "learning_rate": 5e-06, "epoch": 1.6290588882773802, "percentage": 54.33, "elapsed_time": "3:29:40", "remaining_time": "2:56:14"}
150
+ {"current_steps": 1490, "total_steps": 2724, "loss": 0.4312, "learning_rate": 5e-06, "epoch": 1.640066042927903, "percentage": 54.7, "elapsed_time": "3:31:03", "remaining_time": "2:54:48"}
151
+ {"current_steps": 1500, "total_steps": 2724, "loss": 0.4332, "learning_rate": 5e-06, "epoch": 1.6510731975784259, "percentage": 55.07, "elapsed_time": "3:32:27", "remaining_time": "2:53:21"}
152
+ {"current_steps": 1510, "total_steps": 2724, "loss": 0.4329, "learning_rate": 5e-06, "epoch": 1.6620803522289487, "percentage": 55.43, "elapsed_time": "3:33:50", "remaining_time": "2:51:55"}
153
+ {"current_steps": 1520, "total_steps": 2724, "loss": 0.4323, "learning_rate": 5e-06, "epoch": 1.6730875068794715, "percentage": 55.8, "elapsed_time": "3:35:13", "remaining_time": "2:50:29"}
154
+ {"current_steps": 1530, "total_steps": 2724, "loss": 0.4364, "learning_rate": 5e-06, "epoch": 1.6840946615299945, "percentage": 56.17, "elapsed_time": "3:36:37", "remaining_time": "2:49:03"}
155
+ {"current_steps": 1540, "total_steps": 2724, "loss": 0.4313, "learning_rate": 5e-06, "epoch": 1.6951018161805174, "percentage": 56.53, "elapsed_time": "3:38:01", "remaining_time": "2:47:37"}
156
+ {"current_steps": 1550, "total_steps": 2724, "loss": 0.4307, "learning_rate": 5e-06, "epoch": 1.7061089708310402, "percentage": 56.9, "elapsed_time": "3:39:25", "remaining_time": "2:46:12"}
157
+ {"current_steps": 1560, "total_steps": 2724, "loss": 0.4353, "learning_rate": 5e-06, "epoch": 1.717116125481563, "percentage": 57.27, "elapsed_time": "3:40:50", "remaining_time": "2:44:46"}
158
+ {"current_steps": 1570, "total_steps": 2724, "loss": 0.4328, "learning_rate": 5e-06, "epoch": 1.7281232801320858, "percentage": 57.64, "elapsed_time": "3:42:14", "remaining_time": "2:43:21"}
159
+ {"current_steps": 1580, "total_steps": 2724, "loss": 0.4287, "learning_rate": 5e-06, "epoch": 1.7391304347826086, "percentage": 58.0, "elapsed_time": "3:43:38", "remaining_time": "2:41:55"}
160
+ {"current_steps": 1590, "total_steps": 2724, "loss": 0.4362, "learning_rate": 5e-06, "epoch": 1.7501375894331317, "percentage": 58.37, "elapsed_time": "3:45:02", "remaining_time": "2:40:30"}
161
+ {"current_steps": 1600, "total_steps": 2724, "loss": 0.4295, "learning_rate": 5e-06, "epoch": 1.7611447440836545, "percentage": 58.74, "elapsed_time": "3:46:26", "remaining_time": "2:39:04"}
162
+ {"current_steps": 1610, "total_steps": 2724, "loss": 0.4298, "learning_rate": 5e-06, "epoch": 1.7721518987341773, "percentage": 59.1, "elapsed_time": "3:47:49", "remaining_time": "2:37:38"}
163
+ {"current_steps": 1620, "total_steps": 2724, "loss": 0.4298, "learning_rate": 5e-06, "epoch": 1.7831590533847002, "percentage": 59.47, "elapsed_time": "3:49:13", "remaining_time": "2:36:12"}
164
+ {"current_steps": 1630, "total_steps": 2724, "loss": 0.4317, "learning_rate": 5e-06, "epoch": 1.794166208035223, "percentage": 59.84, "elapsed_time": "3:50:36", "remaining_time": "2:34:46"}
165
+ {"current_steps": 1640, "total_steps": 2724, "loss": 0.4335, "learning_rate": 5e-06, "epoch": 1.8051733626857458, "percentage": 60.21, "elapsed_time": "3:51:59", "remaining_time": "2:33:20"}
166
+ {"current_steps": 1650, "total_steps": 2724, "loss": 0.4369, "learning_rate": 5e-06, "epoch": 1.8161805173362686, "percentage": 60.57, "elapsed_time": "3:53:23", "remaining_time": "2:31:54"}
167
+ {"current_steps": 1660, "total_steps": 2724, "loss": 0.4372, "learning_rate": 5e-06, "epoch": 1.8271876719867914, "percentage": 60.94, "elapsed_time": "3:54:46", "remaining_time": "2:30:28"}
168
+ {"current_steps": 1670, "total_steps": 2724, "loss": 0.4268, "learning_rate": 5e-06, "epoch": 1.8381948266373143, "percentage": 61.31, "elapsed_time": "3:56:09", "remaining_time": "2:29:02"}
169
+ {"current_steps": 1680, "total_steps": 2724, "loss": 0.4296, "learning_rate": 5e-06, "epoch": 1.849201981287837, "percentage": 61.67, "elapsed_time": "3:57:32", "remaining_time": "2:27:36"}
170
+ {"current_steps": 1690, "total_steps": 2724, "loss": 0.4359, "learning_rate": 5e-06, "epoch": 1.86020913593836, "percentage": 62.04, "elapsed_time": "3:58:55", "remaining_time": "2:26:11"}
171
+ {"current_steps": 1700, "total_steps": 2724, "loss": 0.438, "learning_rate": 5e-06, "epoch": 1.8712162905888827, "percentage": 62.41, "elapsed_time": "4:00:19", "remaining_time": "2:24:45"}
172
+ {"current_steps": 1710, "total_steps": 2724, "loss": 0.4312, "learning_rate": 5e-06, "epoch": 1.8822234452394055, "percentage": 62.78, "elapsed_time": "4:01:42", "remaining_time": "2:23:19"}
173
+ {"current_steps": 1720, "total_steps": 2724, "loss": 0.4336, "learning_rate": 5e-06, "epoch": 1.8932305998899284, "percentage": 63.14, "elapsed_time": "4:03:05", "remaining_time": "2:21:53"}
174
+ {"current_steps": 1730, "total_steps": 2724, "loss": 0.4296, "learning_rate": 5e-06, "epoch": 1.9042377545404512, "percentage": 63.51, "elapsed_time": "4:04:28", "remaining_time": "2:20:28"}
175
+ {"current_steps": 1740, "total_steps": 2724, "loss": 0.4305, "learning_rate": 5e-06, "epoch": 1.915244909190974, "percentage": 63.88, "elapsed_time": "4:05:52", "remaining_time": "2:19:02"}
176
+ {"current_steps": 1750, "total_steps": 2724, "loss": 0.4293, "learning_rate": 5e-06, "epoch": 1.9262520638414968, "percentage": 64.24, "elapsed_time": "4:07:15", "remaining_time": "2:17:37"}
177
+ {"current_steps": 1760, "total_steps": 2724, "loss": 0.4273, "learning_rate": 5e-06, "epoch": 1.9372592184920197, "percentage": 64.61, "elapsed_time": "4:08:38", "remaining_time": "2:16:11"}
178
+ {"current_steps": 1770, "total_steps": 2724, "loss": 0.4303, "learning_rate": 5e-06, "epoch": 1.9482663731425427, "percentage": 64.98, "elapsed_time": "4:10:00", "remaining_time": "2:14:45"}
179
+ {"current_steps": 1780, "total_steps": 2724, "loss": 0.4276, "learning_rate": 5e-06, "epoch": 1.9592735277930655, "percentage": 65.35, "elapsed_time": "4:11:23", "remaining_time": "2:13:19"}
180
+ {"current_steps": 1790, "total_steps": 2724, "loss": 0.43, "learning_rate": 5e-06, "epoch": 1.9702806824435883, "percentage": 65.71, "elapsed_time": "4:12:46", "remaining_time": "2:11:53"}
181
+ {"current_steps": 1800, "total_steps": 2724, "loss": 0.4291, "learning_rate": 5e-06, "epoch": 1.9812878370941112, "percentage": 66.08, "elapsed_time": "4:14:09", "remaining_time": "2:10:27"}
182
+ {"current_steps": 1810, "total_steps": 2724, "loss": 0.4296, "learning_rate": 5e-06, "epoch": 1.992294991744634, "percentage": 66.45, "elapsed_time": "4:15:32", "remaining_time": "2:09:02"}
183
+ {"current_steps": 1817, "total_steps": 2724, "eval_loss": 0.4724336564540863, "epoch": 2.0, "percentage": 66.7, "elapsed_time": "4:19:10", "remaining_time": "2:09:22"}
184
+ {"current_steps": 1820, "total_steps": 2724, "loss": 0.4165, "learning_rate": 5e-06, "epoch": 2.003302146395157, "percentage": 66.81, "elapsed_time": "4:20:34", "remaining_time": "2:09:25"}
185
+ {"current_steps": 1830, "total_steps": 2724, "loss": 0.379, "learning_rate": 5e-06, "epoch": 2.01430930104568, "percentage": 67.18, "elapsed_time": "4:21:57", "remaining_time": "2:07:58"}
186
+ {"current_steps": 1840, "total_steps": 2724, "loss": 0.3769, "learning_rate": 5e-06, "epoch": 2.0253164556962027, "percentage": 67.55, "elapsed_time": "4:23:20", "remaining_time": "2:06:31"}
187
+ {"current_steps": 1850, "total_steps": 2724, "loss": 0.3732, "learning_rate": 5e-06, "epoch": 2.0363236103467255, "percentage": 67.91, "elapsed_time": "4:24:43", "remaining_time": "2:05:03"}
188
+ {"current_steps": 1860, "total_steps": 2724, "loss": 0.379, "learning_rate": 5e-06, "epoch": 2.0473307649972483, "percentage": 68.28, "elapsed_time": "4:26:06", "remaining_time": "2:03:36"}
189
+ {"current_steps": 1870, "total_steps": 2724, "loss": 0.3792, "learning_rate": 5e-06, "epoch": 2.058337919647771, "percentage": 68.65, "elapsed_time": "4:27:29", "remaining_time": "2:02:09"}
190
+ {"current_steps": 1880, "total_steps": 2724, "loss": 0.379, "learning_rate": 5e-06, "epoch": 2.069345074298294, "percentage": 69.02, "elapsed_time": "4:28:53", "remaining_time": "2:00:42"}
191
+ {"current_steps": 1890, "total_steps": 2724, "loss": 0.378, "learning_rate": 5e-06, "epoch": 2.0803522289488168, "percentage": 69.38, "elapsed_time": "4:30:16", "remaining_time": "1:59:15"}
192
+ {"current_steps": 1900, "total_steps": 2724, "loss": 0.3776, "learning_rate": 5e-06, "epoch": 2.0913593835993396, "percentage": 69.75, "elapsed_time": "4:31:39", "remaining_time": "1:57:48"}
193
+ {"current_steps": 1910, "total_steps": 2724, "loss": 0.3737, "learning_rate": 5e-06, "epoch": 2.1023665382498624, "percentage": 70.12, "elapsed_time": "4:33:03", "remaining_time": "1:56:22"}
194
+ {"current_steps": 1920, "total_steps": 2724, "loss": 0.3772, "learning_rate": 5e-06, "epoch": 2.1133736929003852, "percentage": 70.48, "elapsed_time": "4:34:26", "remaining_time": "1:54:55"}
195
+ {"current_steps": 1930, "total_steps": 2724, "loss": 0.3701, "learning_rate": 5e-06, "epoch": 2.124380847550908, "percentage": 70.85, "elapsed_time": "4:35:49", "remaining_time": "1:53:28"}
196
+ {"current_steps": 1940, "total_steps": 2724, "loss": 0.3757, "learning_rate": 5e-06, "epoch": 2.135388002201431, "percentage": 71.22, "elapsed_time": "4:37:13", "remaining_time": "1:52:01"}
197
+ {"current_steps": 1950, "total_steps": 2724, "loss": 0.375, "learning_rate": 5e-06, "epoch": 2.1463951568519537, "percentage": 71.59, "elapsed_time": "4:38:36", "remaining_time": "1:50:34"}
198
+ {"current_steps": 1960, "total_steps": 2724, "loss": 0.3814, "learning_rate": 5e-06, "epoch": 2.1574023115024765, "percentage": 71.95, "elapsed_time": "4:39:58", "remaining_time": "1:49:08"}
199
+ {"current_steps": 1970, "total_steps": 2724, "loss": 0.3744, "learning_rate": 5e-06, "epoch": 2.1684094661529993, "percentage": 72.32, "elapsed_time": "4:41:21", "remaining_time": "1:47:41"}
200
+ {"current_steps": 1980, "total_steps": 2724, "loss": 0.3741, "learning_rate": 5e-06, "epoch": 2.179416620803522, "percentage": 72.69, "elapsed_time": "4:42:44", "remaining_time": "1:46:14"}
201
+ {"current_steps": 1990, "total_steps": 2724, "loss": 0.378, "learning_rate": 5e-06, "epoch": 2.190423775454045, "percentage": 73.05, "elapsed_time": "4:44:07", "remaining_time": "1:44:47"}
202
+ {"current_steps": 2000, "total_steps": 2724, "loss": 0.3769, "learning_rate": 5e-06, "epoch": 2.201430930104568, "percentage": 73.42, "elapsed_time": "4:45:30", "remaining_time": "1:43:21"}
203
+ {"current_steps": 2010, "total_steps": 2724, "loss": 0.3775, "learning_rate": 5e-06, "epoch": 2.2124380847550906, "percentage": 73.79, "elapsed_time": "4:46:53", "remaining_time": "1:41:54"}
204
+ {"current_steps": 2020, "total_steps": 2724, "loss": 0.3809, "learning_rate": 5e-06, "epoch": 2.2234452394056134, "percentage": 74.16, "elapsed_time": "4:48:16", "remaining_time": "1:40:28"}
205
+ {"current_steps": 2030, "total_steps": 2724, "loss": 0.3741, "learning_rate": 5e-06, "epoch": 2.2344523940561363, "percentage": 74.52, "elapsed_time": "4:49:39", "remaining_time": "1:39:01"}
206
+ {"current_steps": 2040, "total_steps": 2724, "loss": 0.3811, "learning_rate": 5e-06, "epoch": 2.2454595487066595, "percentage": 74.89, "elapsed_time": "4:51:02", "remaining_time": "1:37:34"}
207
+ {"current_steps": 2050, "total_steps": 2724, "loss": 0.3759, "learning_rate": 5e-06, "epoch": 2.2564667033571824, "percentage": 75.26, "elapsed_time": "4:52:25", "remaining_time": "1:36:08"}
208
+ {"current_steps": 2060, "total_steps": 2724, "loss": 0.3777, "learning_rate": 5e-06, "epoch": 2.267473858007705, "percentage": 75.62, "elapsed_time": "4:53:48", "remaining_time": "1:34:42"}
209
+ {"current_steps": 2070, "total_steps": 2724, "loss": 0.3809, "learning_rate": 5e-06, "epoch": 2.278481012658228, "percentage": 75.99, "elapsed_time": "4:55:12", "remaining_time": "1:33:15"}
210
+ {"current_steps": 2080, "total_steps": 2724, "loss": 0.3748, "learning_rate": 5e-06, "epoch": 2.289488167308751, "percentage": 76.36, "elapsed_time": "4:56:35", "remaining_time": "1:31:49"}
211
+ {"current_steps": 2090, "total_steps": 2724, "loss": 0.3774, "learning_rate": 5e-06, "epoch": 2.3004953219592736, "percentage": 76.73, "elapsed_time": "4:57:58", "remaining_time": "1:30:23"}
212
+ {"current_steps": 2100, "total_steps": 2724, "loss": 0.3802, "learning_rate": 5e-06, "epoch": 2.3115024766097965, "percentage": 77.09, "elapsed_time": "4:59:21", "remaining_time": "1:28:57"}
213
+ {"current_steps": 2110, "total_steps": 2724, "loss": 0.3835, "learning_rate": 5e-06, "epoch": 2.3225096312603193, "percentage": 77.46, "elapsed_time": "5:00:44", "remaining_time": "1:27:30"}
214
+ {"current_steps": 2120, "total_steps": 2724, "loss": 0.378, "learning_rate": 5e-06, "epoch": 2.333516785910842, "percentage": 77.83, "elapsed_time": "5:02:08", "remaining_time": "1:26:04"}
215
+ {"current_steps": 2130, "total_steps": 2724, "loss": 0.378, "learning_rate": 5e-06, "epoch": 2.344523940561365, "percentage": 78.19, "elapsed_time": "5:03:31", "remaining_time": "1:24:38"}
216
+ {"current_steps": 2140, "total_steps": 2724, "loss": 0.3854, "learning_rate": 5e-06, "epoch": 2.3555310952118877, "percentage": 78.56, "elapsed_time": "5:04:54", "remaining_time": "1:23:12"}
217
+ {"current_steps": 2150, "total_steps": 2724, "loss": 0.3818, "learning_rate": 5e-06, "epoch": 2.3665382498624106, "percentage": 78.93, "elapsed_time": "5:06:18", "remaining_time": "1:21:46"}
218
+ {"current_steps": 2160, "total_steps": 2724, "loss": 0.3818, "learning_rate": 5e-06, "epoch": 2.3775454045129334, "percentage": 79.3, "elapsed_time": "5:07:41", "remaining_time": "1:20:20"}
219
+ {"current_steps": 2170, "total_steps": 2724, "loss": 0.3772, "learning_rate": 5e-06, "epoch": 2.388552559163456, "percentage": 79.66, "elapsed_time": "5:09:04", "remaining_time": "1:18:54"}
220
+ {"current_steps": 2180, "total_steps": 2724, "loss": 0.3811, "learning_rate": 5e-06, "epoch": 2.399559713813979, "percentage": 80.03, "elapsed_time": "5:10:28", "remaining_time": "1:17:28"}
221
+ {"current_steps": 2190, "total_steps": 2724, "loss": 0.3783, "learning_rate": 5e-06, "epoch": 2.410566868464502, "percentage": 80.4, "elapsed_time": "5:11:51", "remaining_time": "1:16:02"}
222
+ {"current_steps": 2200, "total_steps": 2724, "loss": 0.3793, "learning_rate": 5e-06, "epoch": 2.4215740231150247, "percentage": 80.76, "elapsed_time": "5:13:14", "remaining_time": "1:14:36"}
223
+ {"current_steps": 2210, "total_steps": 2724, "loss": 0.3803, "learning_rate": 5e-06, "epoch": 2.4325811777655475, "percentage": 81.13, "elapsed_time": "5:14:37", "remaining_time": "1:13:10"}
224
+ {"current_steps": 2220, "total_steps": 2724, "loss": 0.3788, "learning_rate": 5e-06, "epoch": 2.4435883324160703, "percentage": 81.5, "elapsed_time": "5:16:01", "remaining_time": "1:11:44"}
225
+ {"current_steps": 2230, "total_steps": 2724, "loss": 0.3817, "learning_rate": 5e-06, "epoch": 2.454595487066593, "percentage": 81.86, "elapsed_time": "5:17:24", "remaining_time": "1:10:18"}
226
+ {"current_steps": 2240, "total_steps": 2724, "loss": 0.3778, "learning_rate": 5e-06, "epoch": 2.465602641717116, "percentage": 82.23, "elapsed_time": "5:18:47", "remaining_time": "1:08:52"}
227
+ {"current_steps": 2250, "total_steps": 2724, "loss": 0.3805, "learning_rate": 5e-06, "epoch": 2.476609796367639, "percentage": 82.6, "elapsed_time": "5:20:11", "remaining_time": "1:07:27"}
228
+ {"current_steps": 2260, "total_steps": 2724, "loss": 0.3805, "learning_rate": 5e-06, "epoch": 2.487616951018162, "percentage": 82.97, "elapsed_time": "5:21:34", "remaining_time": "1:06:01"}
229
+ {"current_steps": 2270, "total_steps": 2724, "loss": 0.3779, "learning_rate": 5e-06, "epoch": 2.498624105668685, "percentage": 83.33, "elapsed_time": "5:22:57", "remaining_time": "1:04:35"}
230
+ {"current_steps": 2280, "total_steps": 2724, "loss": 0.3818, "learning_rate": 5e-06, "epoch": 2.5096312603192077, "percentage": 83.7, "elapsed_time": "5:24:20", "remaining_time": "1:03:09"}
231
+ {"current_steps": 2290, "total_steps": 2724, "loss": 0.3774, "learning_rate": 5e-06, "epoch": 2.5206384149697305, "percentage": 84.07, "elapsed_time": "5:25:44", "remaining_time": "1:01:44"}
232
+ {"current_steps": 2300, "total_steps": 2724, "loss": 0.3787, "learning_rate": 5e-06, "epoch": 2.5316455696202533, "percentage": 84.43, "elapsed_time": "5:27:07", "remaining_time": "1:00:18"}
233
+ {"current_steps": 2310, "total_steps": 2724, "loss": 0.3826, "learning_rate": 5e-06, "epoch": 2.542652724270776, "percentage": 84.8, "elapsed_time": "5:28:30", "remaining_time": "0:58:52"}
234
+ {"current_steps": 2320, "total_steps": 2724, "loss": 0.3825, "learning_rate": 5e-06, "epoch": 2.553659878921299, "percentage": 85.17, "elapsed_time": "5:29:54", "remaining_time": "0:57:27"}
235
+ {"current_steps": 2330, "total_steps": 2724, "loss": 0.3794, "learning_rate": 5e-06, "epoch": 2.564667033571822, "percentage": 85.54, "elapsed_time": "5:31:18", "remaining_time": "0:56:01"}
236
+ {"current_steps": 2340, "total_steps": 2724, "loss": 0.38, "learning_rate": 5e-06, "epoch": 2.5756741882223446, "percentage": 85.9, "elapsed_time": "5:32:42", "remaining_time": "0:54:35"}
237
+ {"current_steps": 2350, "total_steps": 2724, "loss": 0.3782, "learning_rate": 5e-06, "epoch": 2.5866813428728674, "percentage": 86.27, "elapsed_time": "5:34:05", "remaining_time": "0:53:10"}
238
+ {"current_steps": 2360, "total_steps": 2724, "loss": 0.3839, "learning_rate": 5e-06, "epoch": 2.5976884975233903, "percentage": 86.64, "elapsed_time": "5:35:28", "remaining_time": "0:51:44"}
239
+ {"current_steps": 2370, "total_steps": 2724, "loss": 0.3834, "learning_rate": 5e-06, "epoch": 2.608695652173913, "percentage": 87.0, "elapsed_time": "5:36:51", "remaining_time": "0:50:18"}
240
+ {"current_steps": 2380, "total_steps": 2724, "loss": 0.3768, "learning_rate": 5e-06, "epoch": 2.619702806824436, "percentage": 87.37, "elapsed_time": "5:38:14", "remaining_time": "0:48:53"}
241
+ {"current_steps": 2390, "total_steps": 2724, "loss": 0.3809, "learning_rate": 5e-06, "epoch": 2.6307099614749587, "percentage": 87.74, "elapsed_time": "5:39:38", "remaining_time": "0:47:27"}
242
+ {"current_steps": 2400, "total_steps": 2724, "loss": 0.379, "learning_rate": 5e-06, "epoch": 2.6417171161254815, "percentage": 88.11, "elapsed_time": "5:41:01", "remaining_time": "0:46:02"}
243
+ {"current_steps": 2410, "total_steps": 2724, "loss": 0.3794, "learning_rate": 5e-06, "epoch": 2.6527242707760044, "percentage": 88.47, "elapsed_time": "5:42:24", "remaining_time": "0:44:36"}
244
+ {"current_steps": 2420, "total_steps": 2724, "loss": 0.3776, "learning_rate": 5e-06, "epoch": 2.663731425426527, "percentage": 88.84, "elapsed_time": "5:43:46", "remaining_time": "0:43:11"}
245
+ {"current_steps": 2430, "total_steps": 2724, "loss": 0.3809, "learning_rate": 5e-06, "epoch": 2.67473858007705, "percentage": 89.21, "elapsed_time": "5:45:10", "remaining_time": "0:41:45"}
246
+ {"current_steps": 2440, "total_steps": 2724, "loss": 0.379, "learning_rate": 5e-06, "epoch": 2.685745734727573, "percentage": 89.57, "elapsed_time": "5:46:33", "remaining_time": "0:40:20"}
247
+ {"current_steps": 2450, "total_steps": 2724, "loss": 0.3777, "learning_rate": 5e-06, "epoch": 2.6967528893780957, "percentage": 89.94, "elapsed_time": "5:47:56", "remaining_time": "0:38:54"}
248
+ {"current_steps": 2460, "total_steps": 2724, "loss": 0.381, "learning_rate": 5e-06, "epoch": 2.7077600440286185, "percentage": 90.31, "elapsed_time": "5:49:19", "remaining_time": "0:37:29"}
249
+ {"current_steps": 2470, "total_steps": 2724, "loss": 0.3829, "learning_rate": 5e-06, "epoch": 2.7187671986791413, "percentage": 90.68, "elapsed_time": "5:50:43", "remaining_time": "0:36:03"}
250
+ {"current_steps": 2480, "total_steps": 2724, "loss": 0.3798, "learning_rate": 5e-06, "epoch": 2.729774353329664, "percentage": 91.04, "elapsed_time": "5:52:06", "remaining_time": "0:34:38"}
251
+ {"current_steps": 2490, "total_steps": 2724, "loss": 0.3803, "learning_rate": 5e-06, "epoch": 2.740781507980187, "percentage": 91.41, "elapsed_time": "5:53:29", "remaining_time": "0:33:13"}
252
+ {"current_steps": 2500, "total_steps": 2724, "loss": 0.3836, "learning_rate": 5e-06, "epoch": 2.7517886626307098, "percentage": 91.78, "elapsed_time": "5:54:53", "remaining_time": "0:31:47"}
253
+ {"current_steps": 2510, "total_steps": 2724, "loss": 0.3771, "learning_rate": 5e-06, "epoch": 2.7627958172812326, "percentage": 92.14, "elapsed_time": "5:56:17", "remaining_time": "0:30:22"}
254
+ {"current_steps": 2520, "total_steps": 2724, "loss": 0.3832, "learning_rate": 5e-06, "epoch": 2.7738029719317554, "percentage": 92.51, "elapsed_time": "5:57:41", "remaining_time": "0:28:57"}
255
+ {"current_steps": 2530, "total_steps": 2724, "loss": 0.3758, "learning_rate": 5e-06, "epoch": 2.7848101265822782, "percentage": 92.88, "elapsed_time": "5:59:05", "remaining_time": "0:27:32"}
256
+ {"current_steps": 2540, "total_steps": 2724, "loss": 0.3846, "learning_rate": 5e-06, "epoch": 2.795817281232801, "percentage": 93.25, "elapsed_time": "6:00:28", "remaining_time": "0:26:06"}
257
+ {"current_steps": 2550, "total_steps": 2724, "loss": 0.3805, "learning_rate": 5e-06, "epoch": 2.8068244358833243, "percentage": 93.61, "elapsed_time": "6:01:51", "remaining_time": "0:24:41"}
258
+ {"current_steps": 2560, "total_steps": 2724, "loss": 0.3814, "learning_rate": 5e-06, "epoch": 2.817831590533847, "percentage": 93.98, "elapsed_time": "6:03:15", "remaining_time": "0:23:16"}
259
+ {"current_steps": 2570, "total_steps": 2724, "loss": 0.3833, "learning_rate": 5e-06, "epoch": 2.82883874518437, "percentage": 94.35, "elapsed_time": "6:04:39", "remaining_time": "0:21:51"}
260
+ {"current_steps": 2580, "total_steps": 2724, "loss": 0.3785, "learning_rate": 5e-06, "epoch": 2.8398458998348928, "percentage": 94.71, "elapsed_time": "6:06:02", "remaining_time": "0:20:25"}
261
+ {"current_steps": 2590, "total_steps": 2724, "loss": 0.3791, "learning_rate": 5e-06, "epoch": 2.8508530544854156, "percentage": 95.08, "elapsed_time": "6:07:26", "remaining_time": "0:19:00"}
262
+ {"current_steps": 2600, "total_steps": 2724, "loss": 0.3771, "learning_rate": 5e-06, "epoch": 2.8618602091359384, "percentage": 95.45, "elapsed_time": "6:08:49", "remaining_time": "0:17:35"}
263
+ {"current_steps": 2610, "total_steps": 2724, "loss": 0.3814, "learning_rate": 5e-06, "epoch": 2.8728673637864612, "percentage": 95.81, "elapsed_time": "6:10:13", "remaining_time": "0:16:10"}
264
+ {"current_steps": 2620, "total_steps": 2724, "loss": 0.3879, "learning_rate": 5e-06, "epoch": 2.883874518436984, "percentage": 96.18, "elapsed_time": "6:11:36", "remaining_time": "0:14:45"}
265
+ {"current_steps": 2630, "total_steps": 2724, "loss": 0.3756, "learning_rate": 5e-06, "epoch": 2.894881673087507, "percentage": 96.55, "elapsed_time": "6:13:00", "remaining_time": "0:13:19"}
266
+ {"current_steps": 2640, "total_steps": 2724, "loss": 0.3791, "learning_rate": 5e-06, "epoch": 2.9058888277380297, "percentage": 96.92, "elapsed_time": "6:14:23", "remaining_time": "0:11:54"}
267
+ {"current_steps": 2650, "total_steps": 2724, "loss": 0.383, "learning_rate": 5e-06, "epoch": 2.9168959823885525, "percentage": 97.28, "elapsed_time": "6:15:46", "remaining_time": "0:10:29"}
268
+ {"current_steps": 2660, "total_steps": 2724, "loss": 0.3838, "learning_rate": 5e-06, "epoch": 2.9279031370390753, "percentage": 97.65, "elapsed_time": "6:17:10", "remaining_time": "0:09:04"}
269
+ {"current_steps": 2670, "total_steps": 2724, "loss": 0.3805, "learning_rate": 5e-06, "epoch": 2.938910291689598, "percentage": 98.02, "elapsed_time": "6:18:33", "remaining_time": "0:07:39"}
270
+ {"current_steps": 2680, "total_steps": 2724, "loss": 0.379, "learning_rate": 5e-06, "epoch": 2.949917446340121, "percentage": 98.38, "elapsed_time": "6:19:56", "remaining_time": "0:06:14"}
271
+ {"current_steps": 2690, "total_steps": 2724, "loss": 0.3779, "learning_rate": 5e-06, "epoch": 2.960924600990644, "percentage": 98.75, "elapsed_time": "6:21:19", "remaining_time": "0:04:49"}
272
+ {"current_steps": 2700, "total_steps": 2724, "loss": 0.3791, "learning_rate": 5e-06, "epoch": 2.9719317556411666, "percentage": 99.12, "elapsed_time": "6:22:43", "remaining_time": "0:03:24"}
273
+ {"current_steps": 2710, "total_steps": 2724, "loss": 0.377, "learning_rate": 5e-06, "epoch": 2.9829389102916894, "percentage": 99.49, "elapsed_time": "6:24:06", "remaining_time": "0:01:59"}
274
+ {"current_steps": 2720, "total_steps": 2724, "loss": 0.3803, "learning_rate": 5e-06, "epoch": 2.9939460649422127, "percentage": 99.85, "elapsed_time": "6:25:30", "remaining_time": "0:00:34"}
training_args.bin.sagemaker-uploaded ADDED
File without changes