Hugging Face's logo

mlfoundations-dev
/
dcft-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
118e260
·
verified ·
1 Parent(s): 1dbf8a2

Training in progress, epoch 2

Browse files
Files changed (15) 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. model.safetensors.index.json.sagemaker-uploaded +0 -0
  11. special_tokens_map.json.sagemaker-uploaded +0 -0
  12. tokenizer.json.sagemaker-uploaded +0 -0
  13. tokenizer_config.json.sagemaker-uploaded +0 -0
  14. trainer_log.jsonl +92 -0
  15. 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:ffcf6e49d8c6d04008db71e7c5f9ac90efda378c99583d1f5d572146ff3edf08
3
  size 4976698672
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:121351dd234162e1199e31e10d5ba042613ea2e08762a2f28111d47d64120a0f
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 sha256:28cfda2f4b414df4765f982004339df0702f0920719c7466ca6d1e7031139116
3
  size 4999802720
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a34c61babf3d65915e0fabbd3aa5d30cd3f4b0988242761412fa3053620e7d1b
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 sha256:629ab747b135e92ea40b26be6fa7778f7173f6dbadd356fa2e6f43c828b62d03
3
  size 4915916176
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c53398b40f2a4ba473dffc98203684a2981f381057f86be9c4f2d3f983c9e33
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:dd5df3d78d65624c62090eb9faa7350c462f8ab412683ff8df8ab236b2bd6d05
3
  size 1168138808
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e638e434f64e7da87f114ed6fd2359fd9ba61a15e89ff227b1c4f5329cc6add3
3
  size 1168138808
model-00004-of-00004.safetensors.sagemaker-uploaded ADDED
File without changes
model.safetensors.index.json.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 +91,95 @@
91
  {"current_steps": 908, "total_steps": 2724, "eval_loss": 0.4857633411884308, "epoch": 0.9994496422674739, "percentage": 33.33, "elapsed_time": "2:09:41", "remaining_time": "4:19:23"}
92
  {"current_steps": 910, "total_steps": 2724, "loss": 0.4812, "learning_rate": 5e-06, "epoch": 1.0016510731975785, "percentage": 33.41, "elapsed_time": "2:10:46", "remaining_time": "4:20:41"}
93
  {"current_steps": 920, "total_steps": 2724, "loss": 0.4395, "learning_rate": 5e-06, "epoch": 1.0126582278481013, "percentage": 33.77, "elapsed_time": "2:12:09", "remaining_time": "4:19:08"}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
91
  {"current_steps": 908, "total_steps": 2724, "eval_loss": 0.4857633411884308, "epoch": 0.9994496422674739, "percentage": 33.33, "elapsed_time": "2:09:41", "remaining_time": "4:19:23"}
92
  {"current_steps": 910, "total_steps": 2724, "loss": 0.4812, "learning_rate": 5e-06, "epoch": 1.0016510731975785, "percentage": 33.41, "elapsed_time": "2:10:46", "remaining_time": "4:20:41"}
93
  {"current_steps": 920, "total_steps": 2724, "loss": 0.4395, "learning_rate": 5e-06, "epoch": 1.0126582278481013, "percentage": 33.77, "elapsed_time": "2:12:09", "remaining_time": "4:19:08"}
94
+ {"current_steps": 930, "total_steps": 2724, "loss": 0.4379, "learning_rate": 5e-06, "epoch": 1.0236653824986242, "percentage": 34.14, "elapsed_time": "2:13:32", "remaining_time": "4:17:35"}
95
+ {"current_steps": 940, "total_steps": 2724, "loss": 0.4388, "learning_rate": 5e-06, "epoch": 1.034672537149147, "percentage": 34.51, "elapsed_time": "2:14:55", "remaining_time": "4:16:03"}
96
+ {"current_steps": 950, "total_steps": 2724, "loss": 0.4374, "learning_rate": 5e-06, "epoch": 1.0456796917996698, "percentage": 34.88, "elapsed_time": "2:16:18", "remaining_time": "4:14:32"}
97
+ {"current_steps": 960, "total_steps": 2724, "loss": 0.4358, "learning_rate": 5e-06, "epoch": 1.0566868464501926, "percentage": 35.24, "elapsed_time": "2:17:42", "remaining_time": "4:13:01"}
98
+ {"current_steps": 970, "total_steps": 2724, "loss": 0.4359, "learning_rate": 5e-06, "epoch": 1.0676940011007154, "percentage": 35.61, "elapsed_time": "2:19:05", "remaining_time": "4:11:31"}
99
+ {"current_steps": 980, "total_steps": 2724, "loss": 0.4383, "learning_rate": 5e-06, "epoch": 1.0787011557512383, "percentage": 35.98, "elapsed_time": "2:20:29", "remaining_time": "4:10:01"}
100
+ {"current_steps": 990, "total_steps": 2724, "loss": 0.4385, "learning_rate": 5e-06, "epoch": 1.089708310401761, "percentage": 36.34, "elapsed_time": "2:21:53", "remaining_time": "4:08:30"}
101
+ {"current_steps": 1000, "total_steps": 2724, "loss": 0.4405, "learning_rate": 5e-06, "epoch": 1.100715465052284, "percentage": 36.71, "elapsed_time": "2:23:16", "remaining_time": "4:07:00"}
102
+ {"current_steps": 1010, "total_steps": 2724, "loss": 0.4371, "learning_rate": 5e-06, "epoch": 1.1117226197028067, "percentage": 37.08, "elapsed_time": "2:24:39", "remaining_time": "4:05:29"}
103
+ {"current_steps": 1020, "total_steps": 2724, "loss": 0.4345, "learning_rate": 5e-06, "epoch": 1.1227297743533298, "percentage": 37.44, "elapsed_time": "2:26:02", "remaining_time": "4:03:58"}
104
+ {"current_steps": 1030, "total_steps": 2724, "loss": 0.4405, "learning_rate": 5e-06, "epoch": 1.1337369290038526, "percentage": 37.81, "elapsed_time": "2:27:25", "remaining_time": "4:02:28"}
105
+ {"current_steps": 1040, "total_steps": 2724, "loss": 0.4381, "learning_rate": 5e-06, "epoch": 1.1447440836543754, "percentage": 38.18, "elapsed_time": "2:28:49", "remaining_time": "4:00:58"}
106
+ {"current_steps": 1050, "total_steps": 2724, "loss": 0.4387, "learning_rate": 5e-06, "epoch": 1.1557512383048982, "percentage": 38.55, "elapsed_time": "2:30:12", "remaining_time": "3:59:28"}
107
+ {"current_steps": 1060, "total_steps": 2724, "loss": 0.4422, "learning_rate": 5e-06, "epoch": 1.166758392955421, "percentage": 38.91, "elapsed_time": "2:31:35", "remaining_time": "3:57:58"}
108
+ {"current_steps": 1070, "total_steps": 2724, "loss": 0.4341, "learning_rate": 5e-06, "epoch": 1.1777655476059439, "percentage": 39.28, "elapsed_time": "2:32:59", "remaining_time": "3:56:29"}
109
+ {"current_steps": 1080, "total_steps": 2724, "loss": 0.4353, "learning_rate": 5e-06, "epoch": 1.1887727022564667, "percentage": 39.65, "elapsed_time": "2:34:22", "remaining_time": "3:55:00"}
110
+ {"current_steps": 1090, "total_steps": 2724, "loss": 0.4425, "learning_rate": 5e-06, "epoch": 1.1997798569069895, "percentage": 40.01, "elapsed_time": "2:35:46", "remaining_time": "3:53:31"}
111
+ {"current_steps": 1100, "total_steps": 2724, "loss": 0.4376, "learning_rate": 5e-06, "epoch": 1.2107870115575123, "percentage": 40.38, "elapsed_time": "2:37:09", "remaining_time": "3:52:01"}
112
+ {"current_steps": 1110, "total_steps": 2724, "loss": 0.4323, "learning_rate": 5e-06, "epoch": 1.2217941662080352, "percentage": 40.75, "elapsed_time": "2:38:33", "remaining_time": "3:50:33"}
113
+ {"current_steps": 1120, "total_steps": 2724, "loss": 0.4394, "learning_rate": 5e-06, "epoch": 1.232801320858558, "percentage": 41.12, "elapsed_time": "2:39:57", "remaining_time": "3:49:04"}
114
+ {"current_steps": 1130, "total_steps": 2724, "loss": 0.4395, "learning_rate": 5e-06, "epoch": 1.243808475509081, "percentage": 41.48, "elapsed_time": "2:41:20", "remaining_time": "3:47:36"}
115
+ {"current_steps": 1140, "total_steps": 2724, "loss": 0.4361, "learning_rate": 5e-06, "epoch": 1.2548156301596038, "percentage": 41.85, "elapsed_time": "2:42:44", "remaining_time": "3:46:07"}
116
+ {"current_steps": 1150, "total_steps": 2724, "loss": 0.4359, "learning_rate": 5e-06, "epoch": 1.2658227848101267, "percentage": 42.22, "elapsed_time": "2:44:07", "remaining_time": "3:44:38"}
117
+ {"current_steps": 1160, "total_steps": 2724, "loss": 0.4381, "learning_rate": 5e-06, "epoch": 1.2768299394606495, "percentage": 42.58, "elapsed_time": "2:45:31", "remaining_time": "3:43:09"}
118
+ {"current_steps": 1170, "total_steps": 2724, "loss": 0.4375, "learning_rate": 5e-06, "epoch": 1.2878370941111723, "percentage": 42.95, "elapsed_time": "2:46:54", "remaining_time": "3:41:41"}
119
+ {"current_steps": 1180, "total_steps": 2724, "loss": 0.44, "learning_rate": 5e-06, "epoch": 1.2988442487616951, "percentage": 43.32, "elapsed_time": "2:48:17", "remaining_time": "3:40:12"}
120
+ {"current_steps": 1190, "total_steps": 2724, "loss": 0.4321, "learning_rate": 5e-06, "epoch": 1.309851403412218, "percentage": 43.69, "elapsed_time": "2:49:41", "remaining_time": "3:38:44"}
121
+ {"current_steps": 1200, "total_steps": 2724, "loss": 0.4391, "learning_rate": 5e-06, "epoch": 1.3208585580627408, "percentage": 44.05, "elapsed_time": "2:51:04", "remaining_time": "3:37:16"}
122
+ {"current_steps": 1210, "total_steps": 2724, "loss": 0.4299, "learning_rate": 5e-06, "epoch": 1.3318657127132636, "percentage": 44.42, "elapsed_time": "2:52:28", "remaining_time": "3:35:47"}
123
+ {"current_steps": 1220, "total_steps": 2724, "loss": 0.4401, "learning_rate": 5e-06, "epoch": 1.3428728673637864, "percentage": 44.79, "elapsed_time": "2:53:51", "remaining_time": "3:34:19"}
124
+ {"current_steps": 1230, "total_steps": 2724, "loss": 0.4368, "learning_rate": 5e-06, "epoch": 1.3538800220143092, "percentage": 45.15, "elapsed_time": "2:55:14", "remaining_time": "3:32:51"}
125
+ {"current_steps": 1240, "total_steps": 2724, "loss": 0.4382, "learning_rate": 5e-06, "epoch": 1.364887176664832, "percentage": 45.52, "elapsed_time": "2:56:38", "remaining_time": "3:31:24"}
126
+ {"current_steps": 1250, "total_steps": 2724, "loss": 0.4418, "learning_rate": 5e-06, "epoch": 1.3758943313153549, "percentage": 45.89, "elapsed_time": "2:58:02", "remaining_time": "3:29:56"}
127
+ {"current_steps": 1260, "total_steps": 2724, "loss": 0.4402, "learning_rate": 5e-06, "epoch": 1.3869014859658777, "percentage": 46.26, "elapsed_time": "2:59:26", "remaining_time": "3:28:30"}
128
+ {"current_steps": 1270, "total_steps": 2724, "loss": 0.4384, "learning_rate": 5e-06, "epoch": 1.3979086406164005, "percentage": 46.62, "elapsed_time": "3:00:51", "remaining_time": "3:27:03"}
129
+ {"current_steps": 1280, "total_steps": 2724, "loss": 0.4358, "learning_rate": 5e-06, "epoch": 1.4089157952669236, "percentage": 46.99, "elapsed_time": "3:02:14", "remaining_time": "3:25:35"}
130
+ {"current_steps": 1290, "total_steps": 2724, "loss": 0.4366, "learning_rate": 5e-06, "epoch": 1.4199229499174464, "percentage": 47.36, "elapsed_time": "3:03:38", "remaining_time": "3:24:08"}
131
+ {"current_steps": 1300, "total_steps": 2724, "loss": 0.4333, "learning_rate": 5e-06, "epoch": 1.4309301045679692, "percentage": 47.72, "elapsed_time": "3:05:02", "remaining_time": "3:22:41"}
132
+ {"current_steps": 1310, "total_steps": 2724, "loss": 0.4307, "learning_rate": 5e-06, "epoch": 1.441937259218492, "percentage": 48.09, "elapsed_time": "3:06:26", "remaining_time": "3:21:14"}
133
+ {"current_steps": 1320, "total_steps": 2724, "loss": 0.4375, "learning_rate": 5e-06, "epoch": 1.4529444138690149, "percentage": 48.46, "elapsed_time": "3:07:49", "remaining_time": "3:19:46"}
134
+ {"current_steps": 1330, "total_steps": 2724, "loss": 0.4318, "learning_rate": 5e-06, "epoch": 1.4639515685195377, "percentage": 48.83, "elapsed_time": "3:09:13", "remaining_time": "3:18:19"}
135
+ {"current_steps": 1340, "total_steps": 2724, "loss": 0.4395, "learning_rate": 5e-06, "epoch": 1.4749587231700605, "percentage": 49.19, "elapsed_time": "3:10:36", "remaining_time": "3:16:52"}
136
+ {"current_steps": 1350, "total_steps": 2724, "loss": 0.4343, "learning_rate": 5e-06, "epoch": 1.4859658778205833, "percentage": 49.56, "elapsed_time": "3:12:00", "remaining_time": "3:15:25"}
137
+ {"current_steps": 1360, "total_steps": 2724, "loss": 0.4358, "learning_rate": 5e-06, "epoch": 1.4969730324711064, "percentage": 49.93, "elapsed_time": "3:13:24", "remaining_time": "3:13:58"}
138
+ {"current_steps": 1370, "total_steps": 2724, "loss": 0.4365, "learning_rate": 5e-06, "epoch": 1.5079801871216292, "percentage": 50.29, "elapsed_time": "3:14:47", "remaining_time": "3:12:30"}
139
+ {"current_steps": 1380, "total_steps": 2724, "loss": 0.4387, "learning_rate": 5e-06, "epoch": 1.518987341772152, "percentage": 50.66, "elapsed_time": "3:16:10", "remaining_time": "3:11:03"}
140
+ {"current_steps": 1390, "total_steps": 2724, "loss": 0.4275, "learning_rate": 5e-06, "epoch": 1.5299944964226748, "percentage": 51.03, "elapsed_time": "3:17:34", "remaining_time": "3:09:37"}
141
+ {"current_steps": 1400, "total_steps": 2724, "loss": 0.4302, "learning_rate": 5e-06, "epoch": 1.5410016510731976, "percentage": 51.4, "elapsed_time": "3:18:59", "remaining_time": "3:08:10"}
142
+ {"current_steps": 1410, "total_steps": 2724, "loss": 0.4344, "learning_rate": 5e-06, "epoch": 1.5520088057237205, "percentage": 51.76, "elapsed_time": "3:20:23", "remaining_time": "3:06:44"}
143
+ {"current_steps": 1420, "total_steps": 2724, "loss": 0.4366, "learning_rate": 5e-06, "epoch": 1.5630159603742433, "percentage": 52.13, "elapsed_time": "3:21:47", "remaining_time": "3:05:18"}
144
+ {"current_steps": 1430, "total_steps": 2724, "loss": 0.4353, "learning_rate": 5e-06, "epoch": 1.574023115024766, "percentage": 52.5, "elapsed_time": "3:23:11", "remaining_time": "3:03:51"}
145
+ {"current_steps": 1440, "total_steps": 2724, "loss": 0.4332, "learning_rate": 5e-06, "epoch": 1.585030269675289, "percentage": 52.86, "elapsed_time": "3:24:35", "remaining_time": "3:02:25"}
146
+ {"current_steps": 1450, "total_steps": 2724, "loss": 0.4297, "learning_rate": 5e-06, "epoch": 1.5960374243258117, "percentage": 53.23, "elapsed_time": "3:25:59", "remaining_time": "3:00:59"}
147
+ {"current_steps": 1460, "total_steps": 2724, "loss": 0.4369, "learning_rate": 5e-06, "epoch": 1.6070445789763346, "percentage": 53.6, "elapsed_time": "3:27:23", "remaining_time": "2:59:33"}
148
+ {"current_steps": 1470, "total_steps": 2724, "loss": 0.4364, "learning_rate": 5e-06, "epoch": 1.6180517336268574, "percentage": 53.96, "elapsed_time": "3:28:47", "remaining_time": "2:58:06"}
149
+ {"current_steps": 1480, "total_steps": 2724, "loss": 0.4313, "learning_rate": 5e-06, "epoch": 1.6290588882773802, "percentage": 54.33, "elapsed_time": "3:30:11", "remaining_time": "2:56:40"}
150
+ {"current_steps": 1490, "total_steps": 2724, "loss": 0.4314, "learning_rate": 5e-06, "epoch": 1.640066042927903, "percentage": 54.7, "elapsed_time": "3:31:35", "remaining_time": "2:55:14"}
151
+ {"current_steps": 1500, "total_steps": 2724, "loss": 0.4334, "learning_rate": 5e-06, "epoch": 1.6510731975784259, "percentage": 55.07, "elapsed_time": "3:32:59", "remaining_time": "2:53:48"}
152
+ {"current_steps": 1510, "total_steps": 2724, "loss": 0.4332, "learning_rate": 5e-06, "epoch": 1.6620803522289487, "percentage": 55.43, "elapsed_time": "3:34:23", "remaining_time": "2:52:21"}
153
+ {"current_steps": 1520, "total_steps": 2724, "loss": 0.4326, "learning_rate": 5e-06, "epoch": 1.6730875068794715, "percentage": 55.8, "elapsed_time": "3:35:46", "remaining_time": "2:50:55"}
154
+ {"current_steps": 1530, "total_steps": 2724, "loss": 0.4366, "learning_rate": 5e-06, "epoch": 1.6840946615299945, "percentage": 56.17, "elapsed_time": "3:37:10", "remaining_time": "2:49:28"}
155
+ {"current_steps": 1540, "total_steps": 2724, "loss": 0.4316, "learning_rate": 5e-06, "epoch": 1.6951018161805174, "percentage": 56.53, "elapsed_time": "3:38:36", "remaining_time": "2:48:04"}
156
+ {"current_steps": 1550, "total_steps": 2724, "loss": 0.431, "learning_rate": 5e-06, "epoch": 1.7061089708310402, "percentage": 56.9, "elapsed_time": "3:39:59", "remaining_time": "2:46:37"}
157
+ {"current_steps": 1560, "total_steps": 2724, "loss": 0.4359, "learning_rate": 5e-06, "epoch": 1.717116125481563, "percentage": 57.27, "elapsed_time": "3:41:22", "remaining_time": "2:45:10"}
158
+ {"current_steps": 1570, "total_steps": 2724, "loss": 0.4332, "learning_rate": 5e-06, "epoch": 1.7281232801320858, "percentage": 57.64, "elapsed_time": "3:42:44", "remaining_time": "2:43:43"}
159
+ {"current_steps": 1580, "total_steps": 2724, "loss": 0.4295, "learning_rate": 5e-06, "epoch": 1.7391304347826086, "percentage": 58.0, "elapsed_time": "3:44:07", "remaining_time": "2:42:16"}
160
+ {"current_steps": 1590, "total_steps": 2724, "loss": 0.4366, "learning_rate": 5e-06, "epoch": 1.7501375894331317, "percentage": 58.37, "elapsed_time": "3:45:31", "remaining_time": "2:40:50"}
161
+ {"current_steps": 1600, "total_steps": 2724, "loss": 0.4298, "learning_rate": 5e-06, "epoch": 1.7611447440836545, "percentage": 58.74, "elapsed_time": "3:46:54", "remaining_time": "2:39:24"}
162
+ {"current_steps": 1610, "total_steps": 2724, "loss": 0.43, "learning_rate": 5e-06, "epoch": 1.7721518987341773, "percentage": 59.1, "elapsed_time": "3:48:18", "remaining_time": "2:37:58"}
163
+ {"current_steps": 1620, "total_steps": 2724, "loss": 0.43, "learning_rate": 5e-06, "epoch": 1.7831590533847002, "percentage": 59.47, "elapsed_time": "3:49:41", "remaining_time": "2:36:32"}
164
+ {"current_steps": 1630, "total_steps": 2724, "loss": 0.432, "learning_rate": 5e-06, "epoch": 1.794166208035223, "percentage": 59.84, "elapsed_time": "3:51:05", "remaining_time": "2:35:06"}
165
+ {"current_steps": 1640, "total_steps": 2724, "loss": 0.4339, "learning_rate": 5e-06, "epoch": 1.8051733626857458, "percentage": 60.21, "elapsed_time": "3:52:28", "remaining_time": "2:33:39"}
166
+ {"current_steps": 1650, "total_steps": 2724, "loss": 0.4372, "learning_rate": 5e-06, "epoch": 1.8161805173362686, "percentage": 60.57, "elapsed_time": "3:53:51", "remaining_time": "2:32:13"}
167
+ {"current_steps": 1660, "total_steps": 2724, "loss": 0.4374, "learning_rate": 5e-06, "epoch": 1.8271876719867914, "percentage": 60.94, "elapsed_time": "3:55:15", "remaining_time": "2:30:47"}
168
+ {"current_steps": 1670, "total_steps": 2724, "loss": 0.4271, "learning_rate": 5e-06, "epoch": 1.8381948266373143, "percentage": 61.31, "elapsed_time": "3:56:38", "remaining_time": "2:29:21"}
169
+ {"current_steps": 1680, "total_steps": 2724, "loss": 0.4297, "learning_rate": 5e-06, "epoch": 1.849201981287837, "percentage": 61.67, "elapsed_time": "3:58:01", "remaining_time": "2:27:55"}
170
+ {"current_steps": 1690, "total_steps": 2724, "loss": 0.436, "learning_rate": 5e-06, "epoch": 1.86020913593836, "percentage": 62.04, "elapsed_time": "3:59:25", "remaining_time": "2:26:29"}
171
+ {"current_steps": 1700, "total_steps": 2724, "loss": 0.4381, "learning_rate": 5e-06, "epoch": 1.8712162905888827, "percentage": 62.41, "elapsed_time": "4:00:48", "remaining_time": "2:25:03"}
172
+ {"current_steps": 1710, "total_steps": 2724, "loss": 0.4315, "learning_rate": 5e-06, "epoch": 1.8822234452394055, "percentage": 62.78, "elapsed_time": "4:02:12", "remaining_time": "2:23:37"}
173
+ {"current_steps": 1720, "total_steps": 2724, "loss": 0.4337, "learning_rate": 5e-06, "epoch": 1.8932305998899284, "percentage": 63.14, "elapsed_time": "4:03:35", "remaining_time": "2:22:11"}
174
+ {"current_steps": 1730, "total_steps": 2724, "loss": 0.4297, "learning_rate": 5e-06, "epoch": 1.9042377545404512, "percentage": 63.51, "elapsed_time": "4:04:59", "remaining_time": "2:20:45"}
175
+ {"current_steps": 1740, "total_steps": 2724, "loss": 0.4307, "learning_rate": 5e-06, "epoch": 1.915244909190974, "percentage": 63.88, "elapsed_time": "4:06:22", "remaining_time": "2:19:20"}
176
+ {"current_steps": 1750, "total_steps": 2724, "loss": 0.4295, "learning_rate": 5e-06, "epoch": 1.9262520638414968, "percentage": 64.24, "elapsed_time": "4:07:46", "remaining_time": "2:17:54"}
177
+ {"current_steps": 1760, "total_steps": 2724, "loss": 0.4276, "learning_rate": 5e-06, "epoch": 1.9372592184920197, "percentage": 64.61, "elapsed_time": "4:09:09", "remaining_time": "2:16:28"}
178
+ {"current_steps": 1770, "total_steps": 2724, "loss": 0.4306, "learning_rate": 5e-06, "epoch": 1.9482663731425427, "percentage": 64.98, "elapsed_time": "4:10:33", "remaining_time": "2:15:02"}
179
+ {"current_steps": 1780, "total_steps": 2724, "loss": 0.428, "learning_rate": 5e-06, "epoch": 1.9592735277930655, "percentage": 65.35, "elapsed_time": "4:11:56", "remaining_time": "2:13:37"}
180
+ {"current_steps": 1790, "total_steps": 2724, "loss": 0.4303, "learning_rate": 5e-06, "epoch": 1.9702806824435883, "percentage": 65.71, "elapsed_time": "4:13:20", "remaining_time": "2:12:11"}
181
+ {"current_steps": 1800, "total_steps": 2724, "loss": 0.4295, "learning_rate": 5e-06, "epoch": 1.9812878370941112, "percentage": 66.08, "elapsed_time": "4:14:44", "remaining_time": "2:10:46"}
182
+ {"current_steps": 1810, "total_steps": 2724, "loss": 0.4298, "learning_rate": 5e-06, "epoch": 1.992294991744634, "percentage": 66.45, "elapsed_time": "4:16:07", "remaining_time": "2:09:20"}
183
+ {"current_steps": 1817, "total_steps": 2724, "eval_loss": 0.4725576341152191, "epoch": 2.0, "percentage": 66.7, "elapsed_time": "4:19:46", "remaining_time": "2:09:40"}
184
+ {"current_steps": 1820, "total_steps": 2724, "loss": 0.4167, "learning_rate": 5e-06, "epoch": 2.003302146395157, "percentage": 66.81, "elapsed_time": "4:21:11", "remaining_time": "2:09:43"}
185
+ {"current_steps": 1830, "total_steps": 2724, "loss": 0.3791, "learning_rate": 5e-06, "epoch": 2.01430930104568, "percentage": 67.18, "elapsed_time": "4:22:33", "remaining_time": "2:08:16"}
training_args.bin.sagemaker-uploaded ADDED
File without changes