Hugging Face's logo

mlfoundations-dev
/
no_pipeline_code_100k

Text Generation
Transformers
Safetensors
qwen2
llama-factory
full
Generated from Trainer
conversational
text-generation-inference
Model card Files
xet
Community
neginr commited on
Commit
ca9993d
·
verified ·
1 Parent(s): e059d46

Training in progress, epoch 2

Browse files
Files changed (6) hide show
  1. model-00001-of-00004.safetensors +1 -1
  2. model-00002-of-00004.safetensors +1 -1
  3. model-00003-of-00004.safetensors +1 -1
  4. model-00004-of-00004.safetensors +1 -1
  5. trainer_log.jsonl +81 -0
  6. training_args.bin +2 -2
model-00001-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:583c02d70cfe2a97958027c94138b5e3c277d5538b60b62ee0e665c34c58cfef
3
  size 4877660776
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c4fd2f8d9d156f2fe9f4ffa68e9185b1799c3ccf263172d07487a87697827b97
3
  size 4877660776
model-00002-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:aa59c8e28910353a9027b41657225abda9337062c47e0a8e95cf86e51ba1be76
3
  size 4932751008
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc2318b37bcef8e13e2e08212e8325a343a838a2e4b41f1cfdfef719329638c1
3
  size 4932751008
model-00003-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ee3ebf88c615843d76dae09c1bb0a83cccb91c1851d37c78156d2fb39d9b5e85
3
  size 4330865200
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:360723db3185f51c40f6e85f7271941aa4d69dc97f9e84b5879b39eb8e2f6cce
3
  size 4330865200
model-00004-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c56380b670c3e823fbf3d092f4dac3d6d4f81c04b86700811399e4be4c023223
3
  size 1089994880
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e78cdebb2a9c7ce2895ae9c756579e1e8471427a73e3fe9da6b1e63b31cf803
3
  size 1089994880
trainer_log.jsonl CHANGED
@@ -117,3 +117,84 @@
117
  {"current_steps": 117, "total_steps": 295, "loss": 0.5942, "lr": 6.054504078761882e-05, "epoch": 1.9788583509513742, "percentage": 39.66, "elapsed_time": "3:16:20", "remaining_time": "4:58:42"}
118
  {"current_steps": 118, "total_steps": 295, "loss": 0.5989, "lr": 6.013673400353261e-05, "epoch": 1.9957716701902748, "percentage": 40.0, "elapsed_time": "3:17:59", "remaining_time": "4:56:59"}
119
  {"current_steps": 119, "total_steps": 295, "loss": 0.5599, "lr": 5.972559718354869e-05, "epoch": 2.0126849894291756, "percentage": 40.34, "elapsed_time": "3:21:32", "remaining_time": "4:58:04"}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
117
  {"current_steps": 117, "total_steps": 295, "loss": 0.5942, "lr": 6.054504078761882e-05, "epoch": 1.9788583509513742, "percentage": 39.66, "elapsed_time": "3:16:20", "remaining_time": "4:58:42"}
118
  {"current_steps": 118, "total_steps": 295, "loss": 0.5989, "lr": 6.013673400353261e-05, "epoch": 1.9957716701902748, "percentage": 40.0, "elapsed_time": "3:17:59", "remaining_time": "4:56:59"}
119
  {"current_steps": 119, "total_steps": 295, "loss": 0.5599, "lr": 5.972559718354869e-05, "epoch": 2.0126849894291756, "percentage": 40.34, "elapsed_time": "3:21:32", "remaining_time": "4:58:04"}
120
+ {"current_steps": 120, "total_steps": 295, "loss": 0.5491, "lr": 5.93116881092298e-05, "epoch": 2.029598308668076, "percentage": 40.68, "elapsed_time": "3:23:13", "remaining_time": "4:56:22"}
121
+ {"current_steps": 121, "total_steps": 295, "loss": 0.5543, "lr": 5.889506495175398e-05, "epoch": 2.046511627906977, "percentage": 41.02, "elapsed_time": "3:24:54", "remaining_time": "4:54:39"}
122
+ {"current_steps": 122, "total_steps": 295, "loss": 0.5504, "lr": 5.84757862637391e-05, "epoch": 2.063424947145877, "percentage": 41.36, "elapsed_time": "3:26:35", "remaining_time": "4:52:57"}
123
+ {"current_steps": 123, "total_steps": 295, "loss": 0.5463, "lr": 5.805391097101386e-05, "epoch": 2.080338266384778, "percentage": 41.69, "elapsed_time": "3:28:17", "remaining_time": "4:51:15"}
124
+ {"current_steps": 124, "total_steps": 295, "loss": 0.5542, "lr": 5.7629498364336256e-05, "epoch": 2.097251585623679, "percentage": 42.03, "elapsed_time": "3:29:57", "remaining_time": "4:49:32"}
125
+ {"current_steps": 125, "total_steps": 295, "loss": 0.5445, "lr": 5.720260809106083e-05, "epoch": 2.1141649048625792, "percentage": 42.37, "elapsed_time": "3:31:38", "remaining_time": "4:47:49"}
126
+ {"current_steps": 126, "total_steps": 295, "loss": 0.5462, "lr": 5.677330014675576e-05, "epoch": 2.13107822410148, "percentage": 42.71, "elapsed_time": "3:33:18", "remaining_time": "4:46:06"}
127
+ {"current_steps": 127, "total_steps": 295, "loss": 0.5464, "lr": 5.6341634866771e-05, "epoch": 2.1479915433403804, "percentage": 43.05, "elapsed_time": "3:34:59", "remaining_time": "4:44:23"}
128
+ {"current_steps": 128, "total_steps": 295, "loss": 0.5473, "lr": 5.590767291775879e-05, "epoch": 2.1649048625792813, "percentage": 43.39, "elapsed_time": "3:36:39", "remaining_time": "4:42:40"}
129
+ {"current_steps": 129, "total_steps": 295, "loss": 0.5485, "lr": 5.547147528914735e-05, "epoch": 2.1818181818181817, "percentage": 43.73, "elapsed_time": "3:38:19", "remaining_time": "4:40:57"}
130
+ {"current_steps": 130, "total_steps": 295, "loss": 0.5485, "lr": 5.503310328456953e-05, "epoch": 2.1987315010570825, "percentage": 44.07, "elapsed_time": "3:40:00", "remaining_time": "4:39:14"}
131
+ {"current_steps": 131, "total_steps": 295, "loss": 0.5433, "lr": 5.4592618513247016e-05, "epoch": 2.2156448202959833, "percentage": 44.41, "elapsed_time": "3:41:40", "remaining_time": "4:37:31"}
132
+ {"current_steps": 132, "total_steps": 295, "loss": 0.545, "lr": 5.415008288133177e-05, "epoch": 2.2325581395348837, "percentage": 44.75, "elapsed_time": "3:43:20", "remaining_time": "4:35:47"}
133
+ {"current_steps": 133, "total_steps": 295, "loss": 0.5398, "lr": 5.370555858320566e-05, "epoch": 2.2494714587737845, "percentage": 45.08, "elapsed_time": "3:45:01", "remaining_time": "4:34:05"}
134
+ {"current_steps": 134, "total_steps": 295, "loss": 0.548, "lr": 5.3259108092739626e-05, "epoch": 2.266384778012685, "percentage": 45.42, "elapsed_time": "3:46:41", "remaining_time": "4:32:22"}
135
+ {"current_steps": 135, "total_steps": 295, "loss": 0.5439, "lr": 5.2810794154513503e-05, "epoch": 2.2832980972515857, "percentage": 45.76, "elapsed_time": "3:48:21", "remaining_time": "4:30:39"}
136
+ {"current_steps": 136, "total_steps": 295, "loss": 0.5399, "lr": 5.23606797749979e-05, "epoch": 2.300211416490486, "percentage": 46.1, "elapsed_time": "3:50:02", "remaining_time": "4:28:56"}
137
+ {"current_steps": 137, "total_steps": 295, "loss": 0.5413, "lr": 5.1908828213699125e-05, "epoch": 2.317124735729387, "percentage": 46.44, "elapsed_time": "3:51:42", "remaining_time": "4:27:13"}
138
+ {"current_steps": 138, "total_steps": 295, "loss": 0.5435, "lr": 5.1455302974268694e-05, "epoch": 2.3340380549682873, "percentage": 46.78, "elapsed_time": "3:53:22", "remaining_time": "4:25:30"}
139
+ {"current_steps": 139, "total_steps": 295, "loss": 0.5408, "lr": 5.1000167795578484e-05, "epoch": 2.350951374207188, "percentage": 47.12, "elapsed_time": "3:55:02", "remaining_time": "4:23:47"}
140
+ {"current_steps": 140, "total_steps": 295, "loss": 0.5449, "lr": 5.054348664276271e-05, "epoch": 2.367864693446089, "percentage": 47.46, "elapsed_time": "3:56:43", "remaining_time": "4:22:05"}
141
+ {"current_steps": 119, "total_steps": 295, "loss": 0.5492, "lr": 5.972559718354869e-05, "epoch": 2.016913319238901, "percentage": 40.34, "elapsed_time": "0:01:54", "remaining_time": "0:02:48"}
142
+ {"current_steps": 120, "total_steps": 295, "loss": 0.5505, "lr": 5.93116881092298e-05, "epoch": 2.033826638477801, "percentage": 40.68, "elapsed_time": "0:03:37", "remaining_time": "0:05:17"}
143
+ {"current_steps": 121, "total_steps": 295, "loss": 0.5567, "lr": 5.889506495175398e-05, "epoch": 2.050739957716702, "percentage": 41.02, "elapsed_time": "0:05:19", "remaining_time": "0:07:39"}
144
+ {"current_steps": 122, "total_steps": 295, "loss": 0.5475, "lr": 5.84757862637391e-05, "epoch": 2.0676532769556024, "percentage": 41.36, "elapsed_time": "0:07:02", "remaining_time": "0:09:58"}
145
+ {"current_steps": 123, "total_steps": 295, "loss": 0.5529, "lr": 5.805391097101386e-05, "epoch": 2.0845665961945032, "percentage": 41.69, "elapsed_time": "0:08:44", "remaining_time": "0:12:14"}
146
+ {"current_steps": 124, "total_steps": 295, "loss": 0.5516, "lr": 5.7629498364336256e-05, "epoch": 2.1014799154334036, "percentage": 42.03, "elapsed_time": "0:10:26", "remaining_time": "0:14:24"}
147
+ {"current_steps": 125, "total_steps": 295, "loss": 0.5445, "lr": 5.720260809106083e-05, "epoch": 2.1183932346723044, "percentage": 42.37, "elapsed_time": "0:12:10", "remaining_time": "0:16:33"}
148
+ {"current_steps": 126, "total_steps": 295, "loss": 0.5479, "lr": 5.677330014675576e-05, "epoch": 2.1353065539112053, "percentage": 42.71, "elapsed_time": "0:13:54", "remaining_time": "0:18:39"}
149
+ {"current_steps": 127, "total_steps": 295, "loss": 0.5441, "lr": 5.6341634866771e-05, "epoch": 2.1522198731501057, "percentage": 43.05, "elapsed_time": "0:15:37", "remaining_time": "0:20:40"}
150
+ {"current_steps": 128, "total_steps": 295, "loss": 0.5476, "lr": 5.590767291775879e-05, "epoch": 2.1691331923890065, "percentage": 43.39, "elapsed_time": "0:17:20", "remaining_time": "0:22:37"}
151
+ {"current_steps": 129, "total_steps": 295, "loss": 0.5495, "lr": 5.547147528914735e-05, "epoch": 2.186046511627907, "percentage": 43.73, "elapsed_time": "0:19:03", "remaining_time": "0:24:31"}
152
+ {"current_steps": 130, "total_steps": 295, "loss": 0.5479, "lr": 5.503310328456953e-05, "epoch": 2.2029598308668077, "percentage": 44.07, "elapsed_time": "0:20:46", "remaining_time": "0:26:22"}
153
+ {"current_steps": 131, "total_steps": 295, "loss": 0.5392, "lr": 5.4592618513247016e-05, "epoch": 2.219873150105708, "percentage": 44.41, "elapsed_time": "0:22:28", "remaining_time": "0:28:08"}
154
+ {"current_steps": 132, "total_steps": 295, "loss": 0.5468, "lr": 5.415008288133177e-05, "epoch": 2.236786469344609, "percentage": 44.75, "elapsed_time": "0:24:10", "remaining_time": "0:29:51"}
155
+ {"current_steps": 133, "total_steps": 295, "loss": 0.5409, "lr": 5.370555858320566e-05, "epoch": 2.2536997885835097, "percentage": 45.08, "elapsed_time": "0:25:52", "remaining_time": "0:31:31"}
156
+ {"current_steps": 134, "total_steps": 295, "loss": 0.5514, "lr": 5.3259108092739626e-05, "epoch": 2.27061310782241, "percentage": 45.42, "elapsed_time": "0:27:35", "remaining_time": "0:33:09"}
157
+ {"current_steps": 135, "total_steps": 295, "loss": 0.5424, "lr": 5.2810794154513503e-05, "epoch": 2.287526427061311, "percentage": 45.76, "elapsed_time": "0:29:19", "remaining_time": "0:34:45"}
158
+ {"current_steps": 136, "total_steps": 295, "loss": 0.5415, "lr": 5.23606797749979e-05, "epoch": 2.3044397463002113, "percentage": 46.1, "elapsed_time": "0:31:01", "remaining_time": "0:36:16"}
159
+ {"current_steps": 137, "total_steps": 295, "loss": 0.541, "lr": 5.1908828213699125e-05, "epoch": 2.321353065539112, "percentage": 46.44, "elapsed_time": "0:32:44", "remaining_time": "0:37:45"}
160
+ {"current_steps": 138, "total_steps": 295, "loss": 0.5427, "lr": 5.1455302974268694e-05, "epoch": 2.3382663847780125, "percentage": 46.78, "elapsed_time": "0:34:25", "remaining_time": "0:39:10"}
161
+ {"current_steps": 139, "total_steps": 295, "loss": 0.542, "lr": 5.1000167795578484e-05, "epoch": 2.3551797040169133, "percentage": 47.12, "elapsed_time": "0:36:08", "remaining_time": "0:40:33"}
162
+ {"current_steps": 140, "total_steps": 295, "loss": 0.5448, "lr": 5.054348664276271e-05, "epoch": 2.3720930232558137, "percentage": 47.46, "elapsed_time": "0:37:50", "remaining_time": "0:41:53"}
163
+ {"current_steps": 141, "total_steps": 295, "loss": 0.5441, "lr": 5.008532369822835e-05, "epoch": 2.3890063424947146, "percentage": 47.8, "elapsed_time": "0:39:33", "remaining_time": "0:43:12"}
164
+ {"current_steps": 142, "total_steps": 295, "loss": 0.5445, "lr": 4.962574335263474e-05, "epoch": 2.4059196617336154, "percentage": 48.14, "elapsed_time": "0:41:15", "remaining_time": "0:44:27"}
165
+ {"current_steps": 143, "total_steps": 295, "loss": 0.5428, "lr": 4.916481019584417e-05, "epoch": 2.4228329809725158, "percentage": 48.47, "elapsed_time": "0:42:56", "remaining_time": "0:45:38"}
166
+ {"current_steps": 144, "total_steps": 295, "loss": 0.5427, "lr": 4.870258900784432e-05, "epoch": 2.4397463002114166, "percentage": 48.81, "elapsed_time": "0:44:38", "remaining_time": "0:46:48"}
167
+ {"current_steps": 145, "total_steps": 295, "loss": 0.5461, "lr": 4.8239144749643936e-05, "epoch": 2.456659619450317, "percentage": 49.15, "elapsed_time": "0:46:20", "remaining_time": "0:47:55"}
168
+ {"current_steps": 146, "total_steps": 295, "loss": 0.5431, "lr": 4.777454255414331e-05, "epoch": 2.473572938689218, "percentage": 49.49, "elapsed_time": "0:48:01", "remaining_time": "0:49:01"}
169
+ {"current_steps": 147, "total_steps": 295, "loss": 0.5396, "lr": 4.73088477169803e-05, "epoch": 2.4904862579281186, "percentage": 49.83, "elapsed_time": "0:49:47", "remaining_time": "0:50:08"}
170
+ {"current_steps": 148, "total_steps": 295, "loss": 0.532, "lr": 4.684212568735369e-05, "epoch": 2.507399577167019, "percentage": 50.17, "elapsed_time": "0:51:30", "remaining_time": "0:51:09"}
171
+ {"current_steps": 149, "total_steps": 295, "loss": 0.543, "lr": 4.637444205882487e-05, "epoch": 2.52431289640592, "percentage": 50.51, "elapsed_time": "0:53:11", "remaining_time": "0:52:07"}
172
+ {"current_steps": 150, "total_steps": 295, "loss": 0.5422, "lr": 4.5905862560099255e-05, "epoch": 2.54122621564482, "percentage": 50.85, "elapsed_time": "0:54:54", "remaining_time": "0:53:04"}
173
+ {"current_steps": 151, "total_steps": 295, "loss": 0.5431, "lr": 4.5436453045788733e-05, "epoch": 2.558139534883721, "percentage": 51.19, "elapsed_time": "0:56:36", "remaining_time": "0:53:59"}
174
+ {"current_steps": 152, "total_steps": 295, "loss": 0.54, "lr": 4.496627948715638e-05, "epoch": 2.5750528541226214, "percentage": 51.53, "elapsed_time": "0:58:19", "remaining_time": "0:54:52"}
175
+ {"current_steps": 153, "total_steps": 295, "loss": 0.5452, "lr": 4.449540796284476e-05, "epoch": 2.5919661733615222, "percentage": 51.86, "elapsed_time": "1:00:01", "remaining_time": "0:55:42"}
176
+ {"current_steps": 154, "total_steps": 295, "loss": 0.5398, "lr": 4.4023904649589237e-05, "epoch": 2.6088794926004226, "percentage": 52.2, "elapsed_time": "1:01:42", "remaining_time": "0:56:30"}
177
+ {"current_steps": 155, "total_steps": 295, "loss": 0.5388, "lr": 4.3551835812917395e-05, "epoch": 2.6257928118393234, "percentage": 52.54, "elapsed_time": "1:03:24", "remaining_time": "0:57:16"}
178
+ {"current_steps": 156, "total_steps": 295, "loss": 0.5444, "lr": 4.307926779783601e-05, "epoch": 2.6427061310782243, "percentage": 52.88, "elapsed_time": "1:05:05", "remaining_time": "0:57:59"}
179
+ {"current_steps": 157, "total_steps": 295, "loss": 0.5406, "lr": 4.260626701950686e-05, "epoch": 2.6596194503171247, "percentage": 53.22, "elapsed_time": "1:06:46", "remaining_time": "0:58:41"}
180
+ {"current_steps": 158, "total_steps": 295, "loss": 0.5426, "lr": 4.2132899953912705e-05, "epoch": 2.6765327695560255, "percentage": 53.56, "elapsed_time": "1:08:29", "remaining_time": "0:59:23"}
181
+ {"current_steps": 159, "total_steps": 295, "loss": 0.544, "lr": 4.165923312851467e-05, "epoch": 2.693446088794926, "percentage": 53.9, "elapsed_time": "1:10:12", "remaining_time": "1:00:02"}
182
+ {"current_steps": 160, "total_steps": 295, "loss": 0.5376, "lr": 4.1185333112902394e-05, "epoch": 2.7103594080338267, "percentage": 54.24, "elapsed_time": "1:11:57", "remaining_time": "1:00:42"}
183
+ {"current_steps": 161, "total_steps": 295, "loss": 0.5411, "lr": 4.07112665094383e-05, "epoch": 2.7272727272727275, "percentage": 54.58, "elapsed_time": "1:13:39", "remaining_time": "1:01:18"}
184
+ {"current_steps": 162, "total_steps": 295, "loss": 0.5379, "lr": 4.023709994389722e-05, "epoch": 2.744186046511628, "percentage": 54.92, "elapsed_time": "1:15:22", "remaining_time": "1:01:52"}
185
+ {"current_steps": 163, "total_steps": 295, "loss": 0.539, "lr": 3.9762900056102785e-05, "epoch": 2.7610993657505283, "percentage": 55.25, "elapsed_time": "1:17:03", "remaining_time": "1:02:24"}
186
+ {"current_steps": 164, "total_steps": 295, "loss": 0.5422, "lr": 3.9288733490561705e-05, "epoch": 2.778012684989429, "percentage": 55.59, "elapsed_time": "1:18:47", "remaining_time": "1:02:55"}
187
+ {"current_steps": 165, "total_steps": 295, "loss": 0.547, "lr": 3.881466688709761e-05, "epoch": 2.79492600422833, "percentage": 55.93, "elapsed_time": "1:20:31", "remaining_time": "1:03:26"}
188
+ {"current_steps": 166, "total_steps": 295, "loss": 0.5402, "lr": 3.834076687148534e-05, "epoch": 2.8118393234672303, "percentage": 56.27, "elapsed_time": "1:22:13", "remaining_time": "1:03:53"}
189
+ {"current_steps": 167, "total_steps": 295, "loss": 0.5403, "lr": 3.78671000460873e-05, "epoch": 2.828752642706131, "percentage": 56.61, "elapsed_time": "1:23:54", "remaining_time": "1:04:18"}
190
+ {"current_steps": 168, "total_steps": 295, "loss": 0.5443, "lr": 3.739373298049316e-05, "epoch": 2.8456659619450315, "percentage": 56.95, "elapsed_time": "1:25:35", "remaining_time": "1:04:42"}
191
+ {"current_steps": 169, "total_steps": 295, "loss": 0.5369, "lr": 3.6920732202164004e-05, "epoch": 2.8625792811839323, "percentage": 57.29, "elapsed_time": "1:27:17", "remaining_time": "1:05:04"}
192
+ {"current_steps": 170, "total_steps": 295, "loss": 0.5429, "lr": 3.644816418708261e-05, "epoch": 2.879492600422833, "percentage": 57.63, "elapsed_time": "1:28:59", "remaining_time": "1:05:26"}
193
+ {"current_steps": 171, "total_steps": 295, "loss": 0.543, "lr": 3.597609535041078e-05, "epoch": 2.8964059196617336, "percentage": 57.97, "elapsed_time": "1:30:42", "remaining_time": "1:05:46"}
194
+ {"current_steps": 172, "total_steps": 295, "loss": 0.5382, "lr": 3.550459203715525e-05, "epoch": 2.9133192389006344, "percentage": 58.31, "elapsed_time": "1:32:24", "remaining_time": "1:06:04"}
195
+ {"current_steps": 173, "total_steps": 295, "loss": 0.5376, "lr": 3.503372051284363e-05, "epoch": 2.9302325581395348, "percentage": 58.64, "elapsed_time": "1:34:05", "remaining_time": "1:06:21"}
196
+ {"current_steps": 174, "total_steps": 295, "loss": 0.5421, "lr": 3.456354695421128e-05, "epoch": 2.9471458773784356, "percentage": 58.98, "elapsed_time": "1:35:46", "remaining_time": "1:06:36"}
197
+ {"current_steps": 175, "total_steps": 295, "loss": 0.5408, "lr": 3.409413743990076e-05, "epoch": 2.9640591966173364, "percentage": 59.32, "elapsed_time": "1:37:27", "remaining_time": "1:06:49"}
198
+ {"current_steps": 176, "total_steps": 295, "loss": 0.5425, "lr": 3.3625557941175145e-05, "epoch": 2.980972515856237, "percentage": 59.66, "elapsed_time": "1:39:09", "remaining_time": "1:07:02"}
199
+ {"current_steps": 177, "total_steps": 295, "loss": 0.5353, "lr": 3.3157874312646324e-05, "epoch": 2.997885835095137, "percentage": 60.0, "elapsed_time": "1:40:51", "remaining_time": "1:07:14"}
200
+ {"current_steps": 178, "total_steps": 295, "loss": 0.5039, "lr": 3.2691152283019707e-05, "epoch": 3.014799154334038, "percentage": 60.34, "elapsed_time": "1:44:24", "remaining_time": "1:08:37"}
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:131f7c530cb922d6c2e4bdb0a6b4a5ba8366b12c9708eddf234c0de686e628c7
3
- size 7352
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2ebd0d94ce7e8c6bbde1a9aa6f10421480bd47b60a8d0b2dff112c992073ea7e
3
+ size 7416