train=3000 eval=30


processor_config.json:   0%|          | 0.00/1.02k [00:00<?, ?B/s]
processor_config.json: 100%|██████████| 1.02k/1.02k [00:00<00:00, 743kB/s]


chat_template.jinja:   0%|          | 0.00/720 [00:00<?, ?B/s]
chat_template.jinja: 100%|██████████| 720/720 [00:00<00:00, 4.24MB/s]


config.json:   0%|          | 0.00/2.36k [00:00<?, ?B/s]
config.json: 100%|██████████| 2.36k/2.36k [00:00<00:00, 15.0MB/s]


tokenizer_config.json:   0%|          | 0.00/5.56k [00:00<?, ?B/s]
tokenizer_config.json: 100%|██████████| 5.56k/5.56k [00:00<00:00, 31.5MB/s]


tokenizer.json: reconstructing file:   0%|          |  0.00B / 11.4MB            
tokenizer.json: downloading bytes:           |  0.00B            
tokenizer.json: downloading bytes:           | 33.6kB            

tokenizer.json: reconstructing file:   0%|          | 28.5kB / 11.4MB            
tokenizer.json: downloading bytes: ██████████| 3.40MB,  323kB/s  
tokenizer.json: downloading bytes: ██████████| 3.40MB,  323kB/s  

tokenizer.json: reconstructing file: 100%|██████████| 11.4MB / 11.4MB, 1.09MB/s  


added_tokens.json:   0%|          | 0.00/707 [00:00<?, ?B/s]
added_tokens.json: 100%|██████████| 707/707 [00:00<00:00, 5.24MB/s]


special_tokens_map.json:   0%|          | 0.00/613 [00:00<?, ?B/s]
special_tokens_map.json: 100%|██████████| 613/613 [00:00<00:00, 3.49MB/s]
[transformers] You are using a model of type `mistral3` to instantiate a model of type `lighton_ocr`. This may be expected if you are loading a checkpoint that shares a subset of the architecture (e.g., loading a `sam2_video` checkpoint into `Sam2Model`), but is otherwise not supported and can yield errors. Please verify that the checkpoint is compatible with the model you are instantiating.


model.safetensors: reconstructing file:   0%|          |  0.00B / 2.01GB            
model.safetensors: downloading bytes:           |  0.00B            

model.safetensors: reconstructing file:   0%|          | 39.4kB / 2.01GB            
model.safetensors: downloading bytes:           |  103kB            
model.safetensors: downloading bytes: ▏         | 34.7MB, 2.41MB/s  
model.safetensors: downloading bytes: ▋         |  147MB, 12.4MB/s  

model.safetensors: reconstructing file:   7%|▋         |  134MB / 2.01GB, 19.1kB/s  
model.safetensors: downloading bytes: █         |  209MB, 17.3MB/s  

model.safetensors: reconstructing file:  12%|█▏        |  243MB / 2.01GB, 19.4MB/s  
model.safetensors: downloading bytes: █▍        |  288MB, 19.9MB/s  

model.safetensors: reconstructing file:  18%|█▊        |  354MB / 2.01GB, 24.1MB/s  
model.safetensors: downloading bytes: ███▌      |  722MB, 51.3MB/s  

model.safetensors: reconstructing file:  48%|████▊     |  957MB / 2.01GB, 30.1MB/s  
model.safetensors: downloading bytes: █████▉    | 1.19GB, 84.6MB/s  
model.safetensors: downloading bytes: ████████▎ | 1.68GB,  120MB/s  

model.safetensors: reconstructing file:  68%|██████▊   | 1.36GB / 2.01GB, 78.8MB/s  
model.safetensors: downloading bytes: ██████████| 1.75GB,  122MB/s  
model.safetensors: downloading bytes: ██████████| 1.75GB,  122MB/s  

model.safetensors: reconstructing file: 100%|██████████| 2.01GB / 2.01GB,  150MB/s  


Loading weights:   0%|          | 0/532 [00:00<?, ?it/s]
Loading weights: 100%|██████████| 532/532 [00:00<00:00, 15770.96it/s]


generation_config.json:   0%|          | 0.00/219 [00:00<?, ?B/s]
generation_config.json: 100%|██████████| 219/219 [00:00<00:00, 1.62MB/s]
[eval base] 1/30 cer=0.1851
[eval base] 2/30 cer=0.2617
[eval base] 3/30 cer=0.2774
[eval base] 4/30 cer=0.2012
[eval base] 5/30 cer=0.2754
[eval base] 6/30 cer=0.4906
[eval base] 7/30 cer=0.5578
[eval base] 8/30 cer=0.4953
[eval base] 9/30 cer=0.6164
[eval base] 10/30 cer=0.6604
[eval base] 11/30 cer=0.6920
[eval base] 12/30 cer=0.6875
[eval base] 13/30 cer=0.5397
[eval base] 14/30 cer=0.5463
[eval base] 15/30 cer=0.6926
[eval base] 16/30 cer=0.3421
[eval base] 17/30 cer=0.4329
[eval base] 18/30 cer=0.4298
[eval base] 19/30 cer=0.3960
[eval base] 20/30 cer=0.3732
[eval base] 21/30 cer=0.7938
[eval base] 22/30 cer=0.7542
[eval base] 23/30 cer=0.7598
[eval base] 24/30 cer=0.8000
[eval base] 25/30 cer=1.0146
[eval base] 26/30 cer=1.1248
[eval base] 27/30 cer=1.1429
[eval base] 28/30 cer=1.1569
[eval base] 29/30 cer=1.1456
[eval base] 30/30 cer=1.4049
BASE CER mean=0.6417 median=0.6164
epoch=0 step=5/375 loss=0.1145 lr=4.54e-06
epoch=0 step=10/375 loss=0.0793 lr=9.07e-06
epoch=0 step=15/375 loss=0.0548 lr=9.96e-06
epoch=0 step=20/375 loss=0.0353 lr=9.93e-06
epoch=0 step=25/375 loss=0.0220 lr=9.89e-06
epoch=0 step=30/375 loss=0.0150 lr=9.84e-06
epoch=0 step=35/375 loss=0.0105 lr=9.79e-06
epoch=0 step=40/375 loss=0.0083 lr=9.72e-06
epoch=0 step=45/375 loss=0.0066 lr=9.65e-06
epoch=0 step=50/375 loss=0.0049 lr=9.57e-06
epoch=0 step=55/375 loss=0.0038 lr=9.48e-06
epoch=0 step=60/375 loss=0.0038 lr=9.38e-06
epoch=0 step=65/375 loss=0.0040 lr=9.28e-06
epoch=0 step=70/375 loss=0.0035 lr=9.16e-06
epoch=0 step=75/375 loss=0.0035 lr=9.05e-06
epoch=0 step=80/375 loss=0.0025 lr=8.92e-06
epoch=0 step=85/375 loss=0.0028 lr=8.78e-06
epoch=0 step=90/375 loss=0.0030 lr=8.64e-06
epoch=0 step=95/375 loss=0.0025 lr=8.50e-06
epoch=0 step=100/375 loss=0.0024 lr=8.35e-06
epoch=0 step=105/375 loss=0.0022 lr=8.19e-06
epoch=0 step=110/375 loss=0.0025 lr=8.02e-06
epoch=0 step=115/375 loss=0.0021 lr=7.85e-06
epoch=0 step=120/375 loss=0.0018 lr=7.68e-06
epoch=0 step=125/375 loss=0.0021 lr=7.50e-06
epoch=0 step=130/375 loss=0.0019 lr=7.32e-06
epoch=0 step=135/375 loss=0.0018 lr=7.13e-06
epoch=0 step=140/375 loss=0.0017 lr=6.94e-06
epoch=0 step=145/375 loss=0.0019 lr=6.74e-06
epoch=0 step=150/375 loss=0.0016 lr=6.55e-06
epoch=0 step=155/375 loss=0.0017 lr=6.34e-06
epoch=0 step=160/375 loss=0.0014 lr=6.14e-06
epoch=0 step=165/375 loss=0.0018 lr=5.94e-06
epoch=0 step=170/375 loss=0.0017 lr=5.73e-06
epoch=0 step=175/375 loss=0.0016 lr=5.52e-06
epoch=0 step=180/375 loss=0.0014 lr=5.31e-06
epoch=0 step=185/375 loss=0.0014 lr=5.10e-06
epoch=1 step=190/375 loss=0.0014 lr=4.90e-06
epoch=1 step=195/375 loss=0.0014 lr=4.69e-06
epoch=1 step=200/375 loss=0.0015 lr=4.48e-06
epoch=1 step=205/375 loss=0.0016 lr=4.27e-06
epoch=1 step=210/375 loss=0.0015 lr=4.06e-06
epoch=1 step=215/375 loss=0.0011 lr=3.86e-06
epoch=1 step=220/375 loss=0.0013 lr=3.66e-06
epoch=1 step=225/375 loss=0.0013 lr=3.45e-06
epoch=1 step=230/375 loss=0.0015 lr=3.26e-06
epoch=1 step=235/375 loss=0.0013 lr=3.06e-06
epoch=1 step=240/375 loss=0.0016 lr=2.87e-06
epoch=1 step=245/375 loss=0.0013 lr=2.68e-06
epoch=1 step=250/375 loss=0.0017 lr=2.50e-06
epoch=1 step=255/375 loss=0.0012 lr=2.32e-06
epoch=1 step=260/375 loss=0.0012 lr=2.15e-06
epoch=1 step=265/375 loss=0.0013 lr=1.98e-06
epoch=1 step=270/375 loss=0.0014 lr=1.81e-06
epoch=1 step=275/375 loss=0.0016 lr=1.65e-06
epoch=1 step=280/375 loss=0.0013 lr=1.50e-06
epoch=1 step=285/375 loss=0.0014 lr=1.36e-06
epoch=1 step=290/375 loss=0.0012 lr=1.22e-06
epoch=1 step=295/375 loss=0.0012 lr=1.08e-06
epoch=1 step=300/375 loss=0.0015 lr=1.00e-06
epoch=1 step=305/375 loss=0.0012 lr=1.00e-06
epoch=1 step=310/375 loss=0.0015 lr=1.00e-06
epoch=1 step=315/375 loss=0.0013 lr=1.00e-06
epoch=1 step=320/375 loss=0.0014 lr=1.00e-06
epoch=1 step=325/375 loss=0.0013 lr=1.00e-06
epoch=1 step=330/375 loss=0.0014 lr=1.00e-06
epoch=1 step=335/375 loss=0.0015 lr=1.00e-06
epoch=1 step=340/375 loss=0.0012 lr=1.00e-06
epoch=1 step=345/375 loss=0.0014 lr=1.00e-06
epoch=1 step=350/375 loss=0.0012 lr=1.00e-06
epoch=1 step=355/375 loss=0.0012 lr=1.00e-06
epoch=1 step=360/375 loss=0.0011 lr=1.00e-06
epoch=1 step=365/375 loss=0.0011 lr=1.00e-06
epoch=1 step=370/375 loss=0.0012 lr=1.00e-06
epoch=1 step=375/375 loss=0.0014 lr=1.00e-06
[eval finetuned] 1/30 cer=0.0007
[eval finetuned] 2/30 cer=0.0007
[eval finetuned] 3/30 cer=0.0007
[eval finetuned] 4/30 cer=0.0008
[eval finetuned] 5/30 cer=0.0007
[eval finetuned] 6/30 cer=0.0010
[eval finetuned] 7/30 cer=0.0010
[eval finetuned] 8/30 cer=0.0010
[eval finetuned] 9/30 cer=0.0011
[eval finetuned] 10/30 cer=0.0010
[eval finetuned] 11/30 cer=0.0008
[eval finetuned] 12/30 cer=0.0008
[eval finetuned] 13/30 cer=0.0009
[eval finetuned] 14/30 cer=0.0009
[eval finetuned] 15/30 cer=0.0008
[eval finetuned] 16/30 cer=0.0009
[eval finetuned] 17/30 cer=0.0008
[eval finetuned] 18/30 cer=0.0008
[eval finetuned] 19/30 cer=0.0009
[eval finetuned] 20/30 cer=0.0010
[eval finetuned] 21/30 cer=0.0019
[eval finetuned] 22/30 cer=0.0018
[eval finetuned] 23/30 cer=0.0019
[eval finetuned] 24/30 cer=0.0020
[eval finetuned] 25/30 cer=0.0018
[eval finetuned] 26/30 cer=0.0015
[eval finetuned] 27/30 cer=0.0016
[eval finetuned] 28/30 cer=0.0016
[eval finetuned] 29/30 cer=0.0016
[eval finetuned] 30/30 cer=0.0013
FINETUNED CER mean=0.0012 median=0.0010


Writing model shards:   0%|          | 0/1 [00:00<?, ?it/s]

Writing model shards: 100%|██████████| 1/1 [00:02<00:00,  2.02s/it]
Writing model shards: 100%|██████████| 1/1 [00:02<00:00,  2.02s/it]
SAVED /w/model