Hugging Face's logo

bespokelabs
/
qwen3-1.7b-dabstep-reasoning-108-fixed-reasoning-sharegpt-sft

Text Generation
Transformers
Safetensors
qwen3
llama-factory
full
Generated from Trainer
conversational
text-generation-inference
Model card Files Community
CharlieJi commited on
Commit
07c2e59
·
verified ·
1 Parent(s): bde9343

Training in progress, step 42

Browse files
Files changed (3) hide show
  1. model.safetensors +1 -1
  2. trainer_log.jsonl +67 -36
  3. training_args.bin +1 -1
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9ed43599d048ffebfab7e72b7395c3bcd899989ae12a5a725ad394c22ca81c9e
3
  size 3441185608
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:497db5d132eaa270b0002bd6cf6b6ced840e8096df45acfa7bfcb3b1c7eb2069
3
  size 3441185608
trainer_log.jsonl CHANGED
@@ -1,36 +1,67 @@
1
- {"current_steps": 1, "total_steps": 35, "loss": 1.6904, "lr": 0.0, "epoch": 0.14285714285714285, "percentage": 2.86, "elapsed_time": "0:00:10", "remaining_time": "0:05:57"}
2
- {"current_steps": 2, "total_steps": 35, "loss": 1.6039, "lr": 1e-05, "epoch": 0.2857142857142857, "percentage": 5.71, "elapsed_time": "0:00:13", "remaining_time": "0:03:38"}
3
- {"current_steps": 3, "total_steps": 35, "loss": 1.5826, "lr": 9.978670881475173e-06, "epoch": 0.42857142857142855, "percentage": 8.57, "elapsed_time": "0:00:13", "remaining_time": "0:02:29"}
4
- {"current_steps": 4, "total_steps": 35, "loss": 1.4068, "lr": 9.91486549841951e-06, "epoch": 0.5714285714285714, "percentage": 11.43, "elapsed_time": "0:00:14", "remaining_time": "0:01:54"}
5
- {"current_steps": 5, "total_steps": 35, "loss": 1.4604, "lr": 9.809128215864096e-06, "epoch": 0.7142857142857143, "percentage": 14.29, "elapsed_time": "0:00:15", "remaining_time": "0:01:32"}
6
- {"current_steps": 6, "total_steps": 35, "loss": 1.3071, "lr": 9.66236114702178e-06, "epoch": 0.8571428571428571, "percentage": 17.14, "elapsed_time": "0:00:16", "remaining_time": "0:01:18"}
7
- {"current_steps": 7, "total_steps": 35, "loss": 1.2656, "lr": 9.475816456775313e-06, "epoch": 1.0, "percentage": 20.0, "elapsed_time": "0:00:17", "remaining_time": "0:01:09"}
8
- {"current_steps": 8, "total_steps": 35, "loss": 1.1997, "lr": 9.251085678648072e-06, "epoch": 1.1428571428571428, "percentage": 22.86, "elapsed_time": "0:00:25", "remaining_time": "0:01:25"}
9
- {"current_steps": 9, "total_steps": 35, "loss": 1.1369, "lr": 8.990086136401199e-06, "epoch": 1.2857142857142856, "percentage": 25.71, "elapsed_time": "0:00:25", "remaining_time": "0:01:15"}
10
- {"current_steps": 10, "total_steps": 35, "loss": 1.153, "lr": 8.695044586103297e-06, "epoch": 1.4285714285714286, "percentage": 28.57, "elapsed_time": "0:00:26", "remaining_time": "0:01:06"}
11
- {"current_steps": 11, "total_steps": 35, "loss": 1.0615, "lr": 8.368478218232787e-06, "epoch": 1.5714285714285714, "percentage": 31.43, "elapsed_time": "0:00:27", "remaining_time": "0:00:59"}
12
- {"current_steps": 12, "total_steps": 35, "loss": 1.015, "lr": 8.013173181896283e-06, "epoch": 1.7142857142857144, "percentage": 34.29, "elapsed_time": "0:00:28", "remaining_time": "0:00:54"}
13
- {"current_steps": 13, "total_steps": 35, "loss": 1.06, "lr": 7.63216081438678e-06, "epoch": 1.8571428571428572, "percentage": 37.14, "elapsed_time": "0:00:28", "remaining_time": "0:00:48"}
14
- {"current_steps": 14, "total_steps": 35, "loss": 0.9481, "lr": 7.2286917788826926e-06, "epoch": 2.0, "percentage": 40.0, "elapsed_time": "0:00:30", "remaining_time": "0:00:45"}
15
- {"current_steps": 15, "total_steps": 35, "loss": 0.9303, "lr": 6.806208330935766e-06, "epoch": 2.142857142857143, "percentage": 42.86, "elapsed_time": "0:00:38", "remaining_time": "0:00:50"}
16
- {"current_steps": 16, "total_steps": 35, "loss": 0.9709, "lr": 6.368314950360416e-06, "epoch": 2.2857142857142856, "percentage": 45.71, "elapsed_time": "0:00:38", "remaining_time": "0:00:46"}
17
- {"current_steps": 17, "total_steps": 35, "loss": 0.9213, "lr": 5.918747589082853e-06, "epoch": 2.4285714285714284, "percentage": 48.57, "elapsed_time": "0:00:39", "remaining_time": "0:00:41"}
18
- {"current_steps": 18, "total_steps": 35, "loss": 0.8741, "lr": 5.46134179731651e-06, "epoch": 2.571428571428571, "percentage": 51.43, "elapsed_time": "0:00:40", "remaining_time": "0:00:37"}
19
- {"current_steps": 19, "total_steps": 35, "loss": 0.8563, "lr": 5e-06, "epoch": 2.7142857142857144, "percentage": 54.29, "elapsed_time": "0:00:40", "remaining_time": "0:00:34"}
20
- {"current_steps": 20, "total_steps": 35, "loss": 0.8587, "lr": 4.53865820268349e-06, "epoch": 2.857142857142857, "percentage": 57.14, "elapsed_time": "0:00:41", "remaining_time": "0:00:31"}
21
- {"current_steps": 21, "total_steps": 35, "loss": 0.7558, "lr": 4.081252410917148e-06, "epoch": 3.0, "percentage": 60.0, "elapsed_time": "0:00:42", "remaining_time": "0:00:28"}
22
- {"current_steps": 22, "total_steps": 35, "loss": 0.7808, "lr": 3.6316850496395863e-06, "epoch": 3.142857142857143, "percentage": 62.86, "elapsed_time": "0:00:45", "remaining_time": "0:00:27"}
23
- {"current_steps": 23, "total_steps": 35, "loss": 0.8018, "lr": 3.1937916690642356e-06, "epoch": 3.2857142857142856, "percentage": 65.71, "elapsed_time": "0:00:46", "remaining_time": "0:00:24"}
24
- {"current_steps": 24, "total_steps": 35, "loss": 0.785, "lr": 2.771308221117309e-06, "epoch": 3.4285714285714284, "percentage": 68.57, "elapsed_time": "0:00:50", "remaining_time": "0:00:23"}
25
- {"current_steps": 25, "total_steps": 35, "loss": 0.7427, "lr": 2.3678391856132203e-06, "epoch": 3.571428571428571, "percentage": 71.43, "elapsed_time": "0:00:51", "remaining_time": "0:00:20"}
26
- {"current_steps": 26, "total_steps": 35, "loss": 0.7829, "lr": 1.9868268181037186e-06, "epoch": 3.7142857142857144, "percentage": 74.29, "elapsed_time": "0:00:52", "remaining_time": "0:00:18"}
27
- {"current_steps": 27, "total_steps": 35, "loss": 0.785, "lr": 1.6315217817672142e-06, "epoch": 3.857142857142857, "percentage": 77.14, "elapsed_time": "0:00:53", "remaining_time": "0:00:15"}
28
- {"current_steps": 28, "total_steps": 35, "loss": 0.7273, "lr": 1.3049554138967052e-06, "epoch": 4.0, "percentage": 80.0, "elapsed_time": "0:00:54", "remaining_time": "0:00:13"}
29
- {"current_steps": 29, "total_steps": 35, "loss": 0.7737, "lr": 1.0099138635988026e-06, "epoch": 4.142857142857143, "percentage": 82.86, "elapsed_time": "0:01:03", "remaining_time": "0:00:13"}
30
- {"current_steps": 30, "total_steps": 35, "loss": 0.6784, "lr": 7.489143213519301e-07, "epoch": 4.285714285714286, "percentage": 85.71, "elapsed_time": "0:01:03", "remaining_time": "0:00:10"}
31
- {"current_steps": 31, "total_steps": 35, "loss": 0.7267, "lr": 5.241835432246888e-07, "epoch": 4.428571428571429, "percentage": 88.57, "elapsed_time": "0:01:04", "remaining_time": "0:00:08"}
32
- {"current_steps": 32, "total_steps": 35, "loss": 0.767, "lr": 3.3763885297822153e-07, "epoch": 4.571428571428571, "percentage": 91.43, "elapsed_time": "0:01:05", "remaining_time": "0:00:06"}
33
- {"current_steps": 33, "total_steps": 35, "loss": 0.6971, "lr": 1.908717841359048e-07, "epoch": 4.714285714285714, "percentage": 94.29, "elapsed_time": "0:01:06", "remaining_time": "0:00:04"}
34
- {"current_steps": 34, "total_steps": 35, "loss": 0.7647, "lr": 8.513450158049109e-08, "epoch": 4.857142857142857, "percentage": 97.14, "elapsed_time": "0:01:06", "remaining_time": "0:00:01"}
35
- {"current_steps": 35, "total_steps": 35, "loss": 0.7352, "lr": 2.1329118524827662e-08, "epoch": 5.0, "percentage": 100.0, "elapsed_time": "0:01:08", "remaining_time": "0:00:00"}
36
- {"current_steps": 35, "total_steps": 35, "epoch": 5.0, "percentage": 100.0, "elapsed_time": "0:02:09", "remaining_time": "0:00:00"}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"current_steps": 1, "total_steps": 255, "loss": 1.5712, "lr": 0.0, "epoch": 0.0196078431372549, "percentage": 0.39, "elapsed_time": "0:00:03", "remaining_time": "0:14:04"}
2
+ {"current_steps": 2, "total_steps": 255, "loss": 1.8824, "lr": 3.3333333333333333e-06, "epoch": 0.0392156862745098, "percentage": 0.78, "elapsed_time": "0:00:05", "remaining_time": "0:11:43"}
3
+ {"current_steps": 3, "total_steps": 255, "loss": 1.7331, "lr": 6.666666666666667e-06, "epoch": 0.058823529411764705, "percentage": 1.18, "elapsed_time": "0:00:06", "remaining_time": "0:08:50"}
4
+ {"current_steps": 4, "total_steps": 255, "loss": 1.4968, "lr": 1e-05, "epoch": 0.0784313725490196, "percentage": 1.57, "elapsed_time": "0:00:07", "remaining_time": "0:07:23"}
5
+ {"current_steps": 5, "total_steps": 255, "loss": 1.6074, "lr": 9.999611462404874e-06, "epoch": 0.09803921568627451, "percentage": 1.96, "elapsed_time": "0:00:07", "remaining_time": "0:06:31"}
6
+ {"current_steps": 6, "total_steps": 255, "loss": 1.3465, "lr": 9.998445910004082e-06, "epoch": 0.11764705882352941, "percentage": 2.35, "elapsed_time": "0:00:08", "remaining_time": "0:05:56"}
7
+ {"current_steps": 7, "total_steps": 255, "loss": 1.4326, "lr": 9.996503523941994e-06, "epoch": 0.13725490196078433, "percentage": 2.75, "elapsed_time": "0:00:09", "remaining_time": "0:05:30"}
8
+ {"current_steps": 8, "total_steps": 255, "loss": 1.3606, "lr": 9.993784606094612e-06, "epoch": 0.1568627450980392, "percentage": 3.14, "elapsed_time": "0:00:10", "remaining_time": "0:05:11"}
9
+ {"current_steps": 9, "total_steps": 255, "loss": 1.3686, "lr": 9.990289579022661e-06, "epoch": 0.17647058823529413, "percentage": 3.53, "elapsed_time": "0:00:10", "remaining_time": "0:04:56"}
10
+ {"current_steps": 10, "total_steps": 255, "loss": 1.2102, "lr": 9.986018985905901e-06, "epoch": 0.19607843137254902, "percentage": 3.92, "elapsed_time": "0:00:11", "remaining_time": "0:04:44"}
11
+ {"current_steps": 11, "total_steps": 255, "loss": 1.2232, "lr": 9.980973490458728e-06, "epoch": 0.21568627450980393, "percentage": 4.31, "elapsed_time": "0:00:12", "remaining_time": "0:04:34"}
12
+ {"current_steps": 12, "total_steps": 255, "loss": 1.1679, "lr": 9.975153876827008e-06, "epoch": 0.23529411764705882, "percentage": 4.71, "elapsed_time": "0:00:13", "remaining_time": "0:04:25"}
13
+ {"current_steps": 13, "total_steps": 255, "loss": 0.9655, "lr": 9.968561049466214e-06, "epoch": 0.2549019607843137, "percentage": 5.1, "elapsed_time": "0:00:13", "remaining_time": "0:04:18"}
14
+ {"current_steps": 14, "total_steps": 255, "loss": 1.0298, "lr": 9.961196033000862e-06, "epoch": 0.27450980392156865, "percentage": 5.49, "elapsed_time": "0:00:14", "remaining_time": "0:04:11"}
15
+ {"current_steps": 15, "total_steps": 255, "loss": 1.3297, "lr": 9.953059972065264e-06, "epoch": 0.29411764705882354, "percentage": 5.88, "elapsed_time": "0:00:15", "remaining_time": "0:04:06"}
16
+ {"current_steps": 16, "total_steps": 255, "loss": 1.1663, "lr": 9.944154131125643e-06, "epoch": 0.3137254901960784, "percentage": 6.27, "elapsed_time": "0:00:16", "remaining_time": "0:04:01"}
17
+ {"current_steps": 17, "total_steps": 255, "loss": 1.0548, "lr": 9.934479894283607e-06, "epoch": 0.3333333333333333, "percentage": 6.67, "elapsed_time": "0:00:16", "remaining_time": "0:03:56"}
18
+ {"current_steps": 18, "total_steps": 255, "loss": 1.2631, "lr": 9.924038765061042e-06, "epoch": 0.35294117647058826, "percentage": 7.06, "elapsed_time": "0:00:17", "remaining_time": "0:03:52"}
19
+ {"current_steps": 19, "total_steps": 255, "loss": 1.1965, "lr": 9.912832366166443e-06, "epoch": 0.37254901960784315, "percentage": 7.45, "elapsed_time": "0:00:18", "remaining_time": "0:03:48"}
20
+ {"current_steps": 20, "total_steps": 255, "loss": 1.0466, "lr": 9.900862439242719e-06, "epoch": 0.39215686274509803, "percentage": 7.84, "elapsed_time": "0:00:19", "remaining_time": "0:03:45"}
21
+ {"current_steps": 21, "total_steps": 255, "loss": 0.9809, "lr": 9.888130844596525e-06, "epoch": 0.4117647058823529, "percentage": 8.24, "elapsed_time": "0:00:19", "remaining_time": "0:03:41"}
22
+ {"current_steps": 22, "total_steps": 255, "loss": 1.1545, "lr": 9.874639560909118e-06, "epoch": 0.43137254901960786, "percentage": 8.63, "elapsed_time": "0:00:20", "remaining_time": "0:03:38"}
23
+ {"current_steps": 23, "total_steps": 255, "loss": 0.9701, "lr": 9.860390684928873e-06, "epoch": 0.45098039215686275, "percentage": 9.02, "elapsed_time": "0:00:21", "remaining_time": "0:03:36"}
24
+ {"current_steps": 24, "total_steps": 255, "loss": 0.7863, "lr": 9.84538643114539e-06, "epoch": 0.47058823529411764, "percentage": 9.41, "elapsed_time": "0:00:22", "remaining_time": "0:03:33"}
25
+ {"current_steps": 25, "total_steps": 255, "loss": 1.0002, "lr": 9.829629131445342e-06, "epoch": 0.49019607843137253, "percentage": 9.8, "elapsed_time": "0:00:22", "remaining_time": "0:03:30"}
26
+ {"current_steps": 26, "total_steps": 255, "loss": 0.8831, "lr": 9.81312123475006e-06, "epoch": 0.5098039215686274, "percentage": 10.2, "elapsed_time": "0:00:23", "remaining_time": "0:03:28"}
27
+ {"current_steps": 27, "total_steps": 255, "loss": 0.9045, "lr": 9.795865306634939e-06, "epoch": 0.5294117647058824, "percentage": 10.59, "elapsed_time": "0:00:24", "remaining_time": "0:03:26"}
28
+ {"current_steps": 28, "total_steps": 255, "loss": 0.8205, "lr": 9.777864028930705e-06, "epoch": 0.5490196078431373, "percentage": 10.98, "elapsed_time": "0:00:25", "remaining_time": "0:03:24"}
29
+ {"current_steps": 29, "total_steps": 255, "loss": 0.9167, "lr": 9.759120199306613e-06, "epoch": 0.5686274509803921, "percentage": 11.37, "elapsed_time": "0:00:25", "remaining_time": "0:03:22"}
30
+ {"current_steps": 30, "total_steps": 255, "loss": 0.9964, "lr": 9.73963673083566e-06, "epoch": 0.5882352941176471, "percentage": 11.76, "elapsed_time": "0:00:26", "remaining_time": "0:03:20"}
31
+ {"current_steps": 31, "total_steps": 255, "loss": 0.9844, "lr": 9.719416651541839e-06, "epoch": 0.6078431372549019, "percentage": 12.16, "elapsed_time": "0:00:27", "remaining_time": "0:03:18"}
32
+ {"current_steps": 32, "total_steps": 255, "loss": 0.971, "lr": 9.698463103929542e-06, "epoch": 0.6274509803921569, "percentage": 12.55, "elapsed_time": "0:00:28", "remaining_time": "0:03:16"}
33
+ {"current_steps": 33, "total_steps": 255, "loss": 1.0724, "lr": 9.67677934449517e-06, "epoch": 0.6470588235294118, "percentage": 12.94, "elapsed_time": "0:00:28", "remaining_time": "0:03:14"}
34
+ {"current_steps": 34, "total_steps": 255, "loss": 1.0107, "lr": 9.654368743221022e-06, "epoch": 0.6666666666666666, "percentage": 13.33, "elapsed_time": "0:00:29", "remaining_time": "0:03:13"}
35
+ {"current_steps": 35, "total_steps": 255, "loss": 0.9819, "lr": 9.631234783051544e-06, "epoch": 0.6862745098039216, "percentage": 13.73, "elapsed_time": "0:00:30", "remaining_time": "0:03:11"}
36
+ {"current_steps": 36, "total_steps": 255, "loss": 0.863, "lr": 9.60738105935204e-06, "epoch": 0.7058823529411765, "percentage": 14.12, "elapsed_time": "0:00:31", "remaining_time": "0:03:10"}
37
+ {"current_steps": 37, "total_steps": 255, "loss": 0.9197, "lr": 9.582811279349881e-06, "epoch": 0.7254901960784313, "percentage": 14.51, "elapsed_time": "0:00:31", "remaining_time": "0:03:08"}
38
+ {"current_steps": 38, "total_steps": 255, "loss": 1.0188, "lr": 9.557529261558367e-06, "epoch": 0.7450980392156863, "percentage": 14.9, "elapsed_time": "0:00:32", "remaining_time": "0:03:07"}
39
+ {"current_steps": 39, "total_steps": 255, "loss": 0.8321, "lr": 9.531538935183252e-06, "epoch": 0.7647058823529411, "percentage": 15.29, "elapsed_time": "0:00:33", "remaining_time": "0:03:05"}
40
+ {"current_steps": 40, "total_steps": 255, "loss": 0.9346, "lr": 9.504844339512096e-06, "epoch": 0.7843137254901961, "percentage": 15.69, "elapsed_time": "0:00:34", "remaining_time": "0:03:04"}
41
+ {"current_steps": 41, "total_steps": 255, "loss": 0.884, "lr": 9.477449623286505e-06, "epoch": 0.803921568627451, "percentage": 16.08, "elapsed_time": "0:00:35", "remaining_time": "0:03:02"}
42
+ {"current_steps": 42, "total_steps": 255, "loss": 0.8966, "lr": 9.449359044057344e-06, "epoch": 0.8235294117647058, "percentage": 16.47, "elapsed_time": "0:00:35", "remaining_time": "0:03:01"}
43
+ {"current_steps": 43, "total_steps": 255, "loss": 0.9415, "lr": 9.420576967523049e-06, "epoch": 0.8431372549019608, "percentage": 16.86, "elapsed_time": "0:01:55", "remaining_time": "0:09:27"}
44
+ {"current_steps": 44, "total_steps": 255, "loss": 0.9063, "lr": 9.391107866851143e-06, "epoch": 0.8627450980392157, "percentage": 17.25, "elapsed_time": "0:01:55", "remaining_time": "0:09:15"}
45
+ {"current_steps": 45, "total_steps": 255, "loss": 0.7435, "lr": 9.360956321983028e-06, "epoch": 0.8823529411764706, "percentage": 17.65, "elapsed_time": "0:01:56", "remaining_time": "0:09:04"}
46
+ {"current_steps": 46, "total_steps": 255, "loss": 0.9956, "lr": 9.330127018922195e-06, "epoch": 0.9019607843137255, "percentage": 18.04, "elapsed_time": "0:01:57", "remaining_time": "0:08:53"}
47
+ {"current_steps": 47, "total_steps": 255, "loss": 0.9303, "lr": 9.298624749005953e-06, "epoch": 0.9215686274509803, "percentage": 18.43, "elapsed_time": "0:01:58", "remaining_time": "0:08:43"}
48
+ {"current_steps": 48, "total_steps": 255, "loss": 0.7471, "lr": 9.266454408160779e-06, "epoch": 0.9411764705882353, "percentage": 18.82, "elapsed_time": "0:01:59", "remaining_time": "0:08:33"}
49
+ {"current_steps": 49, "total_steps": 255, "loss": 0.7865, "lr": 9.233620996141421e-06, "epoch": 0.9607843137254902, "percentage": 19.22, "elapsed_time": "0:01:59", "remaining_time": "0:08:24"}
50
+ {"current_steps": 50, "total_steps": 255, "loss": 0.7501, "lr": 9.200129615753858e-06, "epoch": 0.9803921568627451, "percentage": 19.61, "elapsed_time": "0:02:00", "remaining_time": "0:08:14"}
51
+ {"current_steps": 51, "total_steps": 255, "loss": 0.7297, "lr": 9.165985472062245e-06, "epoch": 1.0, "percentage": 20.0, "elapsed_time": "0:02:01", "remaining_time": "0:08:06"}
52
+ {"current_steps": 52, "total_steps": 255, "loss": 0.5673, "lr": 9.131193871579975e-06, "epoch": 1.0196078431372548, "percentage": 20.39, "elapsed_time": "0:02:04", "remaining_time": "0:08:06"}
53
+ {"current_steps": 53, "total_steps": 255, "loss": 0.7061, "lr": 9.09576022144496e-06, "epoch": 1.0392156862745099, "percentage": 20.78, "elapsed_time": "0:02:05", "remaining_time": "0:07:57"}
54
+ {"current_steps": 54, "total_steps": 255, "loss": 0.5602, "lr": 9.059690028579285e-06, "epoch": 1.0588235294117647, "percentage": 21.18, "elapsed_time": "0:02:06", "remaining_time": "0:07:49"}
55
+ {"current_steps": 55, "total_steps": 255, "loss": 0.5984, "lr": 9.022988898833342e-06, "epoch": 1.0784313725490196, "percentage": 21.57, "elapsed_time": "0:02:06", "remaining_time": "0:07:41"}
56
+ {"current_steps": 56, "total_steps": 255, "loss": 0.6954, "lr": 8.985662536114614e-06, "epoch": 1.0980392156862746, "percentage": 21.96, "elapsed_time": "0:02:07", "remaining_time": "0:07:33"}
57
+ {"current_steps": 57, "total_steps": 255, "loss": 0.6879, "lr": 8.947716741501178e-06, "epoch": 1.1176470588235294, "percentage": 22.35, "elapsed_time": "0:02:08", "remaining_time": "0:07:25"}
58
+ {"current_steps": 58, "total_steps": 255, "loss": 0.5973, "lr": 8.90915741234015e-06, "epoch": 1.1372549019607843, "percentage": 22.75, "elapsed_time": "0:02:09", "remaining_time": "0:07:18"}
59
+ {"current_steps": 59, "total_steps": 255, "loss": 0.6397, "lr": 8.869990541331137e-06, "epoch": 1.156862745098039, "percentage": 23.14, "elapsed_time": "0:02:09", "remaining_time": "0:07:11"}
60
+ {"current_steps": 60, "total_steps": 255, "loss": 0.676, "lr": 8.83022221559489e-06, "epoch": 1.1764705882352942, "percentage": 23.53, "elapsed_time": "0:02:10", "remaining_time": "0:07:04"}
61
+ {"current_steps": 61, "total_steps": 255, "loss": 0.6512, "lr": 8.789858615727266e-06, "epoch": 1.196078431372549, "percentage": 23.92, "elapsed_time": "0:02:11", "remaining_time": "0:06:58"}
62
+ {"current_steps": 62, "total_steps": 255, "loss": 0.6142, "lr": 8.748906014838672e-06, "epoch": 1.215686274509804, "percentage": 24.31, "elapsed_time": "0:02:12", "remaining_time": "0:06:51"}
63
+ {"current_steps": 63, "total_steps": 255, "loss": 0.6279, "lr": 8.707370777579134e-06, "epoch": 1.2352941176470589, "percentage": 24.71, "elapsed_time": "0:02:12", "remaining_time": "0:06:45"}
64
+ {"current_steps": 64, "total_steps": 255, "loss": 0.6442, "lr": 8.665259359149132e-06, "epoch": 1.2549019607843137, "percentage": 25.1, "elapsed_time": "0:02:13", "remaining_time": "0:06:39"}
65
+ {"current_steps": 65, "total_steps": 255, "loss": 0.6373, "lr": 8.622578304296364e-06, "epoch": 1.2745098039215685, "percentage": 25.49, "elapsed_time": "0:02:14", "remaining_time": "0:06:33"}
66
+ {"current_steps": 66, "total_steps": 255, "loss": 0.4815, "lr": 8.579334246298593e-06, "epoch": 1.2941176470588236, "percentage": 25.88, "elapsed_time": "0:02:15", "remaining_time": "0:06:27"}
67
+ {"current_steps": 67, "total_steps": 255, "loss": 0.5527, "lr": 8.535533905932739e-06, "epoch": 1.3137254901960784, "percentage": 26.27, "elapsed_time": "0:02:16", "remaining_time": "0:06:21"}
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5117a02f517b1668278867cbfcebcc30cadfe2696c021dfa485adf509c0df5b5
3
  size 8273
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d20515e18cbcd276b6ec9653705809e64b4d4559ce13c446d6735b639325180
3
  size 8273