| W0227 04:02:23.285000 1606219 site-packages/torch/distributed/run.py:766] |
| W0227 04:02:23.285000 1606219 site-packages/torch/distributed/run.py:766] ***************************************** |
| W0227 04:02:23.285000 1606219 site-packages/torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. |
| W0227 04:02:23.285000 1606219 site-packages/torch/distributed/run.py:766] ***************************************** |
| The image processor of type `Qwen2VLImageProcessor` is now loaded as a fast processor by default, even if the model checkpoint was saved with a slow processor. This is a breaking change and may produce slightly different outputs. To continue using the slow processor, instantiate this class with `use_fast=False`. Note that this behavior will be extended to all models in a future release. |
| The image processor of type `Qwen2VLImageProcessor` is now loaded as a fast processor by default, even if the model checkpoint was saved with a slow processor. This is a breaking change and may produce slightly different outputs. To continue using the slow processor, instantiate this class with `use_fast=False`. Note that this behavior will be extended to all models in a future release. |
| The image processor of type `Qwen2VLImageProcessor` is now loaded as a fast processor by default, even if the model checkpoint was saved with a slow processor. This is a breaking change and may produce slightly different outputs. To continue using the slow processor, instantiate this class with `use_fast=False`. Note that this behavior will be extended to all models in a future release. |
| `torch_dtype` is deprecated! Use `dtype` instead! |
|
Loading checkpoint shards: 0%| | 0/5 [00:00<?, ?it/s]
Loading checkpoint shards: 100%|ββββββββββ| 5/5 [00:00<00:00, 50.32it/s] |
| `torch_dtype` is deprecated! Use `dtype` instead! |
|
Loading checkpoint shards: 0%| | 0/5 [00:00<?, ?it/s]`torch_dtype` is deprecated! Use `dtype` instead! |
|
Loading checkpoint shards: 0%| | 0/5 [00:00<?, ?it/s]
Loading checkpoint shards: 80%|ββββββββ | 4/5 [00:00<00:00, 38.99it/s]
Loading checkpoint shards: 100%|ββββββββββ| 5/5 [00:00<00:00, 44.74it/s] |
|
Loading checkpoint shards: 80%|ββββββββ | 4/5 [00:00<00:00, 37.11it/s]
Loading checkpoint shards: 100%|ββββββββββ| 5/5 [00:00<00:00, 43.32it/s] |
| The image processor of type `Qwen2VLImageProcessor` is now loaded as a fast processor by default, even if the model checkpoint was saved with a slow processor. This is a breaking change and may produce slightly different outputs. To continue using the slow processor, instantiate this class with `use_fast=False`. Note that this behavior will be extended to all models in a future release. |
|
Loading checkpoint shards: 0%| | 0/5 [00:00<?, ?it/s]
Loading checkpoint shards: 100%|ββββββββββ| 5/5 [00:00<00:00, 47.82it/s]
Loading checkpoint shards: 100%|ββββββββββ| 5/5 [00:00<00:00, 47.75it/s] |
|
Loading checkpoint shards: 0%| | 0/5 [00:00<?, ?it/s]
Loading checkpoint shards: 80%|ββββββββ | 4/5 [00:00<00:00, 36.50it/s]
Loading checkpoint shards: 100%|ββββββββββ| 5/5 [00:00<00:00, 42.27it/s] |
|
Loading checkpoint shards: 0%| | 0/5 [00:00<?, ?it/s]
Loading checkpoint shards: 100%|ββββββββββ| 5/5 [00:00<00:00, 51.89it/s] |
| `torch_dtype` is deprecated! Use `dtype` instead! |
|
Loading checkpoint shards: 0%| | 0/5 [00:00<?, ?it/s]
Loading checkpoint shards: 100%|ββββββββββ| 5/5 [00:00<00:00, 47.29it/s]
Loading checkpoint shards: 100%|ββββββββββ| 5/5 [00:00<00:00, 47.17it/s] |
|
Loading checkpoint shards: 0%| | 0/5 [00:00<?, ?it/s]
Loading checkpoint shards: 100%|ββββββββββ| 5/5 [00:00<00:00, 49.40it/s]
Loading checkpoint shards: 100%|ββββββββββ| 5/5 [00:00<00:00, 49.29it/s] |
|
SFT Epoch 1/2: 0%| | 0/205 [00:00<?, ?it/s]
SFT Epoch 1/2: 0%| | 0/205 [00:03<?, ?it/s, loss=0.346, sft=0.346, clean=0.187, trig=0.415, accum=4, step=1, kl=0.000]
SFT Epoch 1/2: 0%| | 1/205 [00:03<13:31, 3.98s/it, loss=0.346, sft=0.346, clean=0.187, trig=0.415, accum=4, step=1, kl=0.000]
SFT Epoch 1/2: 0%| | 1/205 [00:06<13:31, 3.98s/it, loss=0.464, sft=0.464, clean=0.249, trig=0.556, accum=4, step=2, kl=-0.000]
SFT Epoch 1/2: 1%| | 2/205 [00:06<10:22, 3.07s/it, loss=0.464, sft=0.464, clean=0.249, trig=0.556, accum=4, step=2, kl=-0.000]
SFT Epoch 1/2: 1%| | 2/205 [00:08<10:22, 3.07s/it, loss=0.463, sft=0.462, clean=0.287, trig=0.538, accum=4, step=3, kl=0.002]
SFT Epoch 1/2: 1%|β | 3/205 [00:08<09:12, 2.73s/it, loss=0.463, sft=0.462, clean=0.287, trig=0.538, accum=4, step=3, kl=0.002]
SFT Epoch 1/2: 1%|β | 3/205 [00:10<09:12, 2.73s/it, loss=0.504, sft=0.498, clean=0.254, trig=0.603, accum=4, step=4, kl=0.011]
SFT Epoch 1/2: 2%|β | 4/205 [00:10<08:22, 2.50s/it, loss=0.504, sft=0.498, clean=0.254, trig=0.603, accum=4, step=4, kl=0.011]
SFT Epoch 1/2: 2%|β | 4/205 [00:12<08:22, 2.50s/it, loss=0.475, sft=0.445, clean=0.260, trig=0.524, accum=4, step=5, kl=0.060]
SFT Epoch 1/2: 2%|β | 5/205 [00:12<07:49, 2.35s/it, loss=0.475, sft=0.445, clean=0.260, trig=0.524, accum=4, step=5, kl=0.060]
SFT Epoch 1/2: 2%|β | 5/205 [00:15<07:49, 2.35s/it, loss=0.388, sft=0.362, clean=0.255, trig=0.408, accum=4, step=6, kl=0.052]
SFT Epoch 1/2: 3%|β | 6/205 [00:15<07:40, 2.31s/it, loss=0.388, sft=0.362, clean=0.255, trig=0.408, accum=4, step=6, kl=0.052]
SFT Epoch 1/2: 3%|β | 6/205 [00:17<07:40, 2.31s/it, loss=0.373, sft=0.349, clean=0.251, trig=0.390, accum=4, step=7, kl=0.048]
SFT Epoch 1/2: 3%|β | 7/205 [00:17<07:32, 2.29s/it, loss=0.373, sft=0.349, clean=0.251, trig=0.390, accum=4, step=7, kl=0.048]
SFT Epoch 1/2: 3%|β | 7/205 [00:19<07:32, 2.29s/it, loss=0.228, sft=0.203, clean=0.174, trig=0.215, accum=4, step=8, kl=0.050]
SFT Epoch 1/2: 4%|β | 8/205 [00:19<07:15, 2.21s/it, loss=0.228, sft=0.203, clean=0.174, trig=0.215, accum=4, step=8, kl=0.050]
SFT Epoch 1/2: 4%|β | 8/205 [00:21<07:15, 2.21s/it, loss=0.270, sft=0.244, clean=0.247, trig=0.243, accum=4, step=9, kl=0.052]
SFT Epoch 1/2: 4%|β | 9/205 [00:21<07:16, 2.23s/it, loss=0.270, sft=0.244, clean=0.247, trig=0.243, accum=4, step=9, kl=0.052]
SFT Epoch 1/2: 4%|β | 9/205 [00:23<07:16, 2.23s/it, loss=0.293, sft=0.251, clean=0.264, trig=0.245, accum=4, step=10, kl=0.083]
SFT Epoch 1/2: 5%|β | 10/205 [00:23<07:06, 2.19s/it, loss=0.293, sft=0.251, clean=0.264, trig=0.245, accum=4, step=10, kl=0.083]
SFT Epoch 1/2: 5%|β | 10/205 [00:25<07:06, 2.19s/it, loss=0.191, sft=0.149, clean=0.165, trig=0.142, accum=4, step=11, kl=0.083]
SFT Epoch 1/2: 5%|β | 11/205 [00:25<06:59, 2.16s/it, loss=0.191, sft=0.149, clean=0.165, trig=0.142, accum=4, step=11, kl=0.083]
SFT Epoch 1/2: 5%|β | 11/205 [00:28<06:59, 2.16s/it, loss=0.220, sft=0.183, clean=0.191, trig=0.180, accum=4, step=12, kl=0.074]
SFT Epoch 1/2: 6%|β | 12/205 [00:28<07:08, 2.22s/it, loss=0.220, sft=0.183, clean=0.191, trig=0.180, accum=4, step=12, kl=0.074]
SFT Epoch 1/2: 6%|β | 12/205 [00:30<07:08, 2.22s/it, loss=0.237, sft=0.201, clean=0.210, trig=0.197, accum=4, step=13, kl=0.072]
SFT Epoch 1/2: 6%|β | 13/205 [00:30<06:56, 2.17s/it, loss=0.237, sft=0.201, clean=0.210, trig=0.197, accum=4, step=13, kl=0.072]
SFT Epoch 1/2: 6%|β | 13/205 [00:32<06:56, 2.17s/it, loss=0.234, sft=0.195, clean=0.194, trig=0.196, accum=4, step=14, kl=0.078]
SFT Epoch 1/2: 7%|β | 14/205 [00:32<06:41, 2.10s/it, loss=0.234, sft=0.195, clean=0.194, trig=0.196, accum=4, step=14, kl=0.078]
SFT Epoch 1/2: 7%|β | 14/205 [00:34<06:41, 2.10s/it, loss=0.286, sft=0.244, clean=0.243, trig=0.245, accum=4, step=15, kl=0.084]
SFT Epoch 1/2: 7%|β | 15/205 [00:34<06:41, 2.11s/it, loss=0.286, sft=0.244, clean=0.243, trig=0.245, accum=4, step=15, kl=0.084]
SFT Epoch 1/2: 7%|β | 15/205 [00:36<06:41, 2.11s/it, loss=0.287, sft=0.247, clean=0.240, trig=0.250, accum=4, step=16, kl=0.081]
SFT Epoch 1/2: 8%|β | 16/205 [00:36<06:48, 2.16s/it, loss=0.287, sft=0.247, clean=0.240, trig=0.250, accum=4, step=16, kl=0.081]
SFT Epoch 1/2: 8%|β | 16/205 [00:38<06:48, 2.16s/it, loss=0.241, sft=0.207, clean=0.205, trig=0.208, accum=4, step=17, kl=0.069]
SFT Epoch 1/2: 8%|β | 17/205 [00:38<06:53, 2.20s/it, loss=0.241, sft=0.207, clean=0.205, trig=0.208, accum=4, step=17, kl=0.069]
SFT Epoch 1/2: 8%|β | 17/205 [00:41<06:53, 2.20s/it, loss=0.239, sft=0.210, clean=0.212, trig=0.210, accum=4, step=18, kl=0.056]
SFT Epoch 1/2: 9%|β | 18/205 [00:41<06:47, 2.18s/it, loss=0.239, sft=0.210, clean=0.212, trig=0.210, accum=4, step=18, kl=0.056]
SFT Epoch 1/2: 9%|β | 18/205 [00:43<06:47, 2.18s/it, loss=0.234, sft=0.199, clean=0.203, trig=0.198, accum=4, step=19, kl=0.069]
SFT Epoch 1/2: 9%|β | 19/205 [00:43<06:51, 2.21s/it, loss=0.234, sft=0.199, clean=0.203, trig=0.198, accum=4, step=19, kl=0.069]
SFT Epoch 1/2: 9%|β | 19/205 [00:45<06:51, 2.21s/it, loss=0.236, sft=0.210, clean=0.212, trig=0.209, accum=4, step=20, kl=0.053]
SFT Epoch 1/2: 10%|β | 20/205 [00:45<06:55, 2.25s/it, loss=0.236, sft=0.210, clean=0.212, trig=0.209, accum=4, step=20, kl=0.053]
SFT Epoch 1/2: 10%|β | 20/205 [00:47<06:55, 2.25s/it, loss=0.209, sft=0.184, clean=0.190, trig=0.181, accum=4, step=21, kl=0.051]
SFT Epoch 1/2: 10%|β | 21/205 [00:47<06:43, 2.20s/it, loss=0.209, sft=0.184, clean=0.190, trig=0.181, accum=4, step=21, kl=0.051]
SFT Epoch 1/2: 10%|β | 21/205 [00:49<06:43, 2.20s/it, loss=0.260, sft=0.230, clean=0.231, trig=0.229, accum=4, step=22, kl=0.060]
SFT Epoch 1/2: 11%|β | 22/205 [00:49<06:28, 2.12s/it, loss=0.260, sft=0.230, clean=0.231, trig=0.229, accum=4, step=22, kl=0.060]
SFT Epoch 1/2: 11%|β | 22/205 [00:51<06:28, 2.12s/it, loss=0.139, sft=0.120, clean=0.118, trig=0.121, accum=4, step=23, kl=0.037]
SFT Epoch 1/2: 11%|β | 23/205 [00:51<06:25, 2.12s/it, loss=0.139, sft=0.120, clean=0.118, trig=0.121, accum=4, step=23, kl=0.037]
SFT Epoch 1/2: 11%|β | 23/205 [00:54<06:25, 2.12s/it, loss=0.289, sft=0.257, clean=0.260, trig=0.256, accum=4, step=24, kl=0.064]
SFT Epoch 1/2: 12%|ββ | 24/205 [00:54<06:25, 2.13s/it, loss=0.289, sft=0.257, clean=0.260, trig=0.256, accum=4, step=24, kl=0.064]
SFT Epoch 1/2: 12%|ββ | 24/205 [00:56<06:25, 2.13s/it, loss=0.234, sft=0.208, clean=0.211, trig=0.207, accum=4, step=25, kl=0.051]
SFT Epoch 1/2: 12%|ββ | 25/205 [00:56<06:23, 2.13s/it, loss=0.234, sft=0.208, clean=0.211, trig=0.207, accum=4, step=25, kl=0.051]
SFT Epoch 1/2: 12%|ββ | 25/205 [00:58<06:23, 2.13s/it, loss=0.315, sft=0.274, clean=0.278, trig=0.272, accum=4, step=26, kl=0.083]
SFT Epoch 1/2: 13%|ββ | 26/205 [00:58<06:39, 2.23s/it, loss=0.315, sft=0.274, clean=0.278, trig=0.272, accum=4, step=26, kl=0.083]
SFT Epoch 1/2: 13%|ββ | 26/205 [01:00<06:39, 2.23s/it, loss=0.233, sft=0.200, clean=0.203, trig=0.199, accum=4, step=27, kl=0.066]
SFT Epoch 1/2: 13%|ββ | 27/205 [01:00<06:42, 2.26s/it, loss=0.233, sft=0.200, clean=0.203, trig=0.199, accum=4, step=27, kl=0.066]
SFT Epoch 1/2: 13%|ββ | 27/205 [01:02<06:42, 2.26s/it, loss=0.284, sft=0.245, clean=0.242, trig=0.247, accum=4, step=28, kl=0.076]
SFT Epoch 1/2: 14%|ββ | 28/205 [01:02<06:27, 2.19s/it, loss=0.284, sft=0.245, clean=0.242, trig=0.247, accum=4, step=28, kl=0.076]
SFT Epoch 1/2: 14%|ββ | 28/205 [01:05<06:27, 2.19s/it, loss=0.172, sft=0.146, clean=0.143, trig=0.148, accum=4, step=29, kl=0.052]
SFT Epoch 1/2: 14%|ββ | 29/205 [01:05<06:21, 2.17s/it, loss=0.172, sft=0.146, clean=0.143, trig=0.148, accum=4, step=29, kl=0.052]
SFT Epoch 1/2: 14%|ββ | 29/205 [01:07<06:21, 2.17s/it, loss=0.208, sft=0.182, clean=0.182, trig=0.183, accum=4, step=30, kl=0.051]
SFT Epoch 1/2: 15%|ββ | 30/205 [01:07<06:14, 2.14s/it, loss=0.208, sft=0.182, clean=0.182, trig=0.183, accum=4, step=30, kl=0.051]
SFT Epoch 1/2: 15%|ββ | 30/205 [01:09<06:14, 2.14s/it, loss=0.258, sft=0.220, clean=0.224, trig=0.219, accum=4, step=31, kl=0.075]
SFT Epoch 1/2: 15%|ββ | 31/205 [01:09<06:15, 2.16s/it, loss=0.258, sft=0.220, clean=0.224, trig=0.219, accum=4, step=31, kl=0.075]
SFT Epoch 1/2: 15%|ββ | 31/205 [01:11<06:15, 2.16s/it, loss=0.151, sft=0.125, clean=0.129, trig=0.123, accum=4, step=32, kl=0.051]
SFT Epoch 1/2: 16%|ββ | 32/205 [01:11<06:08, 2.13s/it, loss=0.151, sft=0.125, clean=0.129, trig=0.123, accum=4, step=32, kl=0.051]
SFT Epoch 1/2: 16%|ββ | 32/205 [01:13<06:08, 2.13s/it, loss=0.265, sft=0.217, clean=0.223, trig=0.215, accum=4, step=33, kl=0.096]
SFT Epoch 1/2: 16%|ββ | 33/205 [01:13<06:17, 2.19s/it, loss=0.265, sft=0.217, clean=0.223, trig=0.215, accum=4, step=33, kl=0.096]
SFT Epoch 1/2: 16%|ββ | 33/205 [01:16<06:17, 2.19s/it, loss=0.228, sft=0.194, clean=0.198, trig=0.193, accum=4, step=34, kl=0.067]
SFT Epoch 1/2: 17%|ββ | 34/205 [01:16<06:23, 2.24s/it, loss=0.228, sft=0.194, clean=0.198, trig=0.193, accum=4, step=34, kl=0.067]
SFT Epoch 1/2: 17%|ββ | 34/205 [01:18<06:23, 2.24s/it, loss=0.194, sft=0.162, clean=0.166, trig=0.161, accum=4, step=35, kl=0.063]
SFT Epoch 1/2: 17%|ββ | 35/205 [01:18<06:15, 2.21s/it, loss=0.194, sft=0.162, clean=0.166, trig=0.161, accum=4, step=35, kl=0.063]
SFT Epoch 1/2: 17%|ββ | 35/205 [01:20<06:15, 2.21s/it, loss=0.210, sft=0.179, clean=0.183, trig=0.177, accum=4, step=36, kl=0.062]
SFT Epoch 1/2: 18%|ββ | 36/205 [01:20<06:12, 2.21s/it, loss=0.210, sft=0.179, clean=0.183, trig=0.177, accum=4, step=36, kl=0.062]
SFT Epoch 1/2: 18%|ββ | 36/205 [01:22<06:12, 2.21s/it, loss=0.269, sft=0.236, clean=0.246, trig=0.232, accum=4, step=37, kl=0.064]
SFT Epoch 1/2: 18%|ββ | 37/205 [01:22<06:11, 2.21s/it, loss=0.269, sft=0.236, clean=0.246, trig=0.232, accum=4, step=37, kl=0.064]
SFT Epoch 1/2: 18%|ββ | 37/205 [01:24<06:11, 2.21s/it, loss=0.212, sft=0.181, clean=0.178, trig=0.182, accum=4, step=38, kl=0.062]
SFT Epoch 1/2: 19%|ββ | 38/205 [01:24<06:03, 2.18s/it, loss=0.212, sft=0.181, clean=0.178, trig=0.182, accum=4, step=38, kl=0.062]
SFT Epoch 1/2: 19%|ββ | 38/205 [01:26<06:03, 2.18s/it, loss=0.166, sft=0.135, clean=0.141, trig=0.132, accum=4, step=39, kl=0.063]
SFT Epoch 1/2: 19%|ββ | 39/205 [01:26<05:54, 2.13s/it, loss=0.166, sft=0.135, clean=0.141, trig=0.132, accum=4, step=39, kl=0.063]
SFT Epoch 1/2: 19%|ββ | 39/205 [01:28<05:54, 2.13s/it, loss=0.265, sft=0.230, clean=0.235, trig=0.229, accum=4, step=40, kl=0.069]
SFT Epoch 1/2: 20%|ββ | 40/205 [01:28<05:44, 2.09s/it, loss=0.265, sft=0.230, clean=0.235, trig=0.229, accum=4, step=40, kl=0.069]
SFT Epoch 1/2: 20%|ββ | 40/205 [01:31<05:44, 2.09s/it, loss=0.256, sft=0.217, clean=0.222, trig=0.215, accum=4, step=41, kl=0.077]
SFT Epoch 1/2: 20%|ββ | 41/205 [01:31<05:53, 2.16s/it, loss=0.256, sft=0.217, clean=0.222, trig=0.215, accum=4, step=41, kl=0.077]
SFT Epoch 1/2: 20%|ββ | 41/205 [01:33<05:53, 2.16s/it, loss=0.251, sft=0.218, clean=0.218, trig=0.218, accum=4, step=42, kl=0.065]
SFT Epoch 1/2: 20%|ββ | 42/205 [01:33<05:47, 2.13s/it, loss=0.251, sft=0.218, clean=0.218, trig=0.218, accum=4, step=42, kl=0.065]
SFT Epoch 1/2: 20%|ββ | 42/205 [01:35<05:47, 2.13s/it, loss=0.265, sft=0.230, clean=0.234, trig=0.228, accum=4, step=43, kl=0.071]
SFT Epoch 1/2: 21%|ββ | 43/205 [01:35<05:49, 2.16s/it, loss=0.265, sft=0.230, clean=0.234, trig=0.228, accum=4, step=43, kl=0.071]
SFT Epoch 1/2: 21%|ββ | 43/205 [01:37<05:49, 2.16s/it, loss=0.276, sft=0.244, clean=0.249, trig=0.242, accum=4, step=44, kl=0.064]
SFT Epoch 1/2: 21%|βββ | 44/205 [01:37<05:42, 2.12s/it, loss=0.276, sft=0.244, clean=0.249, trig=0.242, accum=4, step=44, kl=0.064]
SFT Epoch 1/2: 21%|βββ | 44/205 [01:39<05:42, 2.12s/it, loss=0.161, sft=0.132, clean=0.135, trig=0.131, accum=4, step=45, kl=0.057]
SFT Epoch 1/2: 22%|βββ | 45/205 [01:39<05:52, 2.20s/it, loss=0.161, sft=0.132, clean=0.135, trig=0.131, accum=4, step=45, kl=0.057]
SFT Epoch 1/2: 22%|βββ | 45/205 [01:41<05:52, 2.20s/it, loss=0.213, sft=0.184, clean=0.190, trig=0.181, accum=4, step=46, kl=0.058]
SFT Epoch 1/2: 22%|βββ | 46/205 [01:41<05:46, 2.18s/it, loss=0.213, sft=0.184, clean=0.190, trig=0.181, accum=4, step=46, kl=0.058]
SFT Epoch 1/2: 22%|βββ | 46/205 [01:43<05:46, 2.18s/it, loss=0.275, sft=0.234, clean=0.236, trig=0.233, accum=4, step=47, kl=0.083]
SFT Epoch 1/2: 23%|βββ | 47/205 [01:43<05:31, 2.10s/it, loss=0.275, sft=0.234, clean=0.236, trig=0.233, accum=4, step=47, kl=0.083]
SFT Epoch 1/2: 23%|βββ | 47/205 [01:45<05:31, 2.10s/it, loss=0.227, sft=0.197, clean=0.201, trig=0.196, accum=4, step=48, kl=0.060]
SFT Epoch 1/2: 23%|βββ | 48/205 [01:45<05:28, 2.09s/it, loss=0.227, sft=0.197, clean=0.201, trig=0.196, accum=4, step=48, kl=0.060]
SFT Epoch 1/2: 23%|βββ | 48/205 [01:48<05:28, 2.09s/it, loss=0.187, sft=0.164, clean=0.167, trig=0.162, accum=4, step=49, kl=0.047]
SFT Epoch 1/2: 24%|βββ | 49/205 [01:48<05:41, 2.19s/it, loss=0.187, sft=0.164, clean=0.167, trig=0.162, accum=4, step=49, kl=0.047]
SFT Epoch 1/2: 24%|βββ | 49/205 [01:50<05:41, 2.19s/it, loss=0.282, sft=0.248, clean=0.245, trig=0.249, accum=4, step=50, kl=0.070]
SFT Epoch 1/2: 24%|βββ | 50/205 [01:50<05:46, 2.24s/it, loss=0.282, sft=0.248, clean=0.245, trig=0.249, accum=4, step=50, kl=0.070]
SFT Epoch 1/2: 24%|βββ | 50/205 [01:52<05:46, 2.24s/it, loss=0.265, sft=0.236, clean=0.241, trig=0.234, accum=4, step=51, kl=0.058]
SFT Epoch 1/2: 25%|βββ | 51/205 [01:52<05:31, 2.15s/it, loss=0.265, sft=0.236, clean=0.241, trig=0.234, accum=4, step=51, kl=0.058]
SFT Epoch 1/2: 25%|βββ | 51/205 [01:55<05:31, 2.15s/it, loss=0.224, sft=0.194, clean=0.196, trig=0.194, accum=4, step=52, kl=0.059]
SFT Epoch 1/2: 25%|βββ | 52/205 [01:55<05:41, 2.23s/it, loss=0.224, sft=0.194, clean=0.196, trig=0.194, accum=4, step=52, kl=0.059]
SFT Epoch 1/2: 25%|βββ | 52/205 [01:57<05:41, 2.23s/it, loss=0.225, sft=0.193, clean=0.196, trig=0.192, accum=4, step=53, kl=0.065]
SFT Epoch 1/2: 26%|βββ | 53/205 [01:57<05:36, 2.21s/it, loss=0.225, sft=0.193, clean=0.196, trig=0.192, accum=4, step=53, kl=0.065]
SFT Epoch 1/2: 26%|βββ | 53/205 [01:59<05:36, 2.21s/it, loss=0.307, sft=0.272, clean=0.273, trig=0.272, accum=4, step=54, kl=0.069]
SFT Epoch 1/2: 26%|βββ | 54/205 [01:59<05:34, 2.21s/it, loss=0.307, sft=0.272, clean=0.273, trig=0.272, accum=4, step=54, kl=0.069]
SFT Epoch 1/2: 26%|βββ | 54/205 [02:01<05:34, 2.21s/it, loss=0.252, sft=0.224, clean=0.217, trig=0.227, accum=4, step=55, kl=0.058]
SFT Epoch 1/2: 27%|βββ | 55/205 [02:01<05:27, 2.19s/it, loss=0.252, sft=0.224, clean=0.217, trig=0.227, accum=4, step=55, kl=0.058]
SFT Epoch 1/2: 27%|βββ | 55/205 [02:03<05:27, 2.19s/it, loss=0.251, sft=0.218, clean=0.213, trig=0.220, accum=4, step=56, kl=0.065]
SFT Epoch 1/2: 27%|βββ | 56/205 [02:03<05:30, 2.22s/it, loss=0.251, sft=0.218, clean=0.213, trig=0.220, accum=4, step=56, kl=0.065]
SFT Epoch 1/2: 27%|βββ | 56/205 [02:06<05:30, 2.22s/it, loss=0.208, sft=0.179, clean=0.181, trig=0.177, accum=4, step=57, kl=0.058]
SFT Epoch 1/2: 28%|βββ | 57/205 [02:06<05:26, 2.21s/it, loss=0.208, sft=0.179, clean=0.181, trig=0.177, accum=4, step=57, kl=0.058]
SFT Epoch 1/2: 28%|βββ | 57/205 [02:08<05:26, 2.21s/it, loss=0.257, sft=0.228, clean=0.233, trig=0.226, accum=4, step=58, kl=0.056]
SFT Epoch 1/2: 28%|βββ | 58/205 [02:08<05:22, 2.20s/it, loss=0.257, sft=0.228, clean=0.233, trig=0.226, accum=4, step=58, kl=0.056]
SFT Epoch 1/2: 28%|βββ | 58/205 [02:10<05:22, 2.20s/it, loss=0.238, sft=0.205, clean=0.209, trig=0.203, accum=4, step=59, kl=0.066]
SFT Epoch 1/2: 29%|βββ | 59/205 [02:10<05:12, 2.14s/it, loss=0.238, sft=0.205, clean=0.209, trig=0.203, accum=4, step=59, kl=0.066]
SFT Epoch 1/2: 29%|βββ | 59/205 [02:12<05:12, 2.14s/it, loss=0.222, sft=0.195, clean=0.192, trig=0.196, accum=4, step=60, kl=0.054]
SFT Epoch 1/2: 29%|βββ | 60/205 [02:12<05:04, 2.10s/it, loss=0.222, sft=0.195, clean=0.192, trig=0.196, accum=4, step=60, kl=0.054]
SFT Epoch 1/2: 29%|βββ | 60/205 [02:14<05:04, 2.10s/it, loss=0.251, sft=0.213, clean=0.213, trig=0.212, accum=4, step=61, kl=0.078]
SFT Epoch 1/2: 30%|βββ | 61/205 [02:14<05:10, 2.15s/it, loss=0.251, sft=0.213, clean=0.213, trig=0.212, accum=4, step=61, kl=0.078]
SFT Epoch 1/2: 30%|βββ | 61/205 [02:16<05:10, 2.15s/it, loss=0.225, sft=0.190, clean=0.184, trig=0.193, accum=4, step=62, kl=0.069]
SFT Epoch 1/2: 30%|βββ | 62/205 [02:16<05:11, 2.18s/it, loss=0.225, sft=0.190, clean=0.184, trig=0.193, accum=4, step=62, kl=0.069]
SFT Epoch 1/2: 30%|βββ | 62/205 [02:18<05:11, 2.18s/it, loss=0.318, sft=0.277, clean=0.278, trig=0.277, accum=4, step=63, kl=0.082]
SFT Epoch 1/2: 31%|βββ | 63/205 [02:18<05:04, 2.15s/it, loss=0.318, sft=0.277, clean=0.278, trig=0.277, accum=4, step=63, kl=0.082]
SFT Epoch 1/2: 31%|βββ | 63/205 [02:20<05:04, 2.15s/it, loss=0.301, sft=0.258, clean=0.260, trig=0.257, accum=4, step=64, kl=0.087]
SFT Epoch 1/2: 31%|βββ | 64/205 [02:20<05:00, 2.13s/it, loss=0.301, sft=0.258, clean=0.260, trig=0.257, accum=4, step=64, kl=0.087]
SFT Epoch 1/2: 31%|βββ | 64/205 [02:22<05:00, 2.13s/it, loss=0.251, sft=0.224, clean=0.226, trig=0.223, accum=4, step=65, kl=0.055]
SFT Epoch 1/2: 32%|ββββ | 65/205 [02:22<04:54, 2.10s/it, loss=0.251, sft=0.224, clean=0.226, trig=0.223, accum=4, step=65, kl=0.055]
SFT Epoch 1/2: 32%|ββββ | 65/205 [02:25<04:54, 2.10s/it, loss=0.272, sft=0.238, clean=0.242, trig=0.236, accum=4, step=66, kl=0.069]
SFT Epoch 1/2: 32%|ββββ | 66/205 [02:25<04:55, 2.12s/it, loss=0.272, sft=0.238, clean=0.242, trig=0.236, accum=4, step=66, kl=0.069]
SFT Epoch 1/2: 32%|ββββ | 66/205 [02:27<04:55, 2.12s/it, loss=0.252, sft=0.217, clean=0.215, trig=0.218, accum=4, step=67, kl=0.069]
SFT Epoch 1/2: 33%|ββββ | 67/205 [02:27<05:01, 2.18s/it, loss=0.252, sft=0.217, clean=0.215, trig=0.218, accum=4, step=67, kl=0.069]
SFT Epoch 1/2: 33%|ββββ | 67/205 [02:29<05:01, 2.18s/it, loss=0.222, sft=0.192, clean=0.195, trig=0.191, accum=4, step=68, kl=0.060]
SFT Epoch 1/2: 33%|ββββ | 68/205 [02:29<04:47, 2.10s/it, loss=0.222, sft=0.192, clean=0.195, trig=0.191, accum=4, step=68, kl=0.060]
SFT Epoch 1/2: 33%|ββββ | 68/205 [02:31<04:47, 2.10s/it, loss=0.238, sft=0.203, clean=0.205, trig=0.202, accum=4, step=69, kl=0.071]
SFT Epoch 1/2: 34%|ββββ | 69/205 [02:31<04:47, 2.12s/it, loss=0.238, sft=0.203, clean=0.205, trig=0.202, accum=4, step=69, kl=0.071]
SFT Epoch 1/2: 34%|ββββ | 69/205 [02:33<04:47, 2.12s/it, loss=0.215, sft=0.185, clean=0.185, trig=0.185, accum=4, step=70, kl=0.060]
SFT Epoch 1/2: 34%|ββββ | 70/205 [02:33<04:39, 2.07s/it, loss=0.215, sft=0.185, clean=0.185, trig=0.185, accum=4, step=70, kl=0.060]
SFT Epoch 1/2: 34%|ββββ | 70/205 [02:35<04:39, 2.07s/it, loss=0.196, sft=0.164, clean=0.167, trig=0.162, accum=4, step=71, kl=0.065]
SFT Epoch 1/2: 35%|ββββ | 71/205 [02:35<04:52, 2.18s/it, loss=0.196, sft=0.164, clean=0.167, trig=0.162, accum=4, step=71, kl=0.065]
SFT Epoch 1/2: 35%|ββββ | 71/205 [02:38<04:52, 2.18s/it, loss=0.246, sft=0.221, clean=0.224, trig=0.220, accum=4, step=72, kl=0.049]
SFT Epoch 1/2: 35%|ββββ | 72/205 [02:38<04:57, 2.24s/it, loss=0.246, sft=0.221, clean=0.224, trig=0.220, accum=4, step=72, kl=0.049]
SFT Epoch 1/2: 35%|ββββ | 72/205 [02:40<04:57, 2.24s/it, loss=0.294, sft=0.260, clean=0.263, trig=0.258, accum=4, step=73, kl=0.068]
SFT Epoch 1/2: 36%|ββββ | 73/205 [02:40<04:39, 2.12s/it, loss=0.294, sft=0.260, clean=0.263, trig=0.258, accum=4, step=73, kl=0.068]
SFT Epoch 1/2: 36%|ββββ | 73/205 [02:42<04:39, 2.12s/it, loss=0.162, sft=0.133, clean=0.134, trig=0.132, accum=4, step=74, kl=0.059]
SFT Epoch 1/2: 36%|ββββ | 74/205 [02:42<04:32, 2.08s/it, loss=0.162, sft=0.133, clean=0.134, trig=0.132, accum=4, step=74, kl=0.059]
SFT Epoch 1/2: 36%|ββββ | 74/205 [02:44<04:32, 2.08s/it, loss=0.250, sft=0.216, clean=0.223, trig=0.212, accum=4, step=75, kl=0.070]
SFT Epoch 1/2: 37%|ββββ | 75/205 [02:44<04:35, 2.12s/it, loss=0.250, sft=0.216, clean=0.223, trig=0.212, accum=4, step=75, kl=0.070]
SFT Epoch 1/2: 37%|ββββ | 75/205 [02:46<04:35, 2.12s/it, loss=0.241, sft=0.207, clean=0.211, trig=0.205, accum=4, step=76, kl=0.068]
SFT Epoch 1/2: 37%|ββββ | 76/205 [02:46<04:37, 2.15s/it, loss=0.241, sft=0.207, clean=0.211, trig=0.205, accum=4, step=76, kl=0.068]
SFT Epoch 1/2: 37%|ββββ | 76/205 [02:48<04:37, 2.15s/it, loss=0.163, sft=0.137, clean=0.138, trig=0.136, accum=4, step=77, kl=0.052]
SFT Epoch 1/2: 38%|ββββ | 77/205 [02:48<04:38, 2.18s/it, loss=0.163, sft=0.137, clean=0.138, trig=0.136, accum=4, step=77, kl=0.052]
SFT Epoch 1/2: 38%|ββββ | 77/205 [02:50<04:38, 2.18s/it, loss=0.222, sft=0.191, clean=0.185, trig=0.193, accum=4, step=78, kl=0.064]
SFT Epoch 1/2: 38%|ββββ | 78/205 [02:50<04:29, 2.12s/it, loss=0.222, sft=0.191, clean=0.185, trig=0.193, accum=4, step=78, kl=0.064]
SFT Epoch 1/2: 38%|ββββ | 78/205 [02:53<04:29, 2.12s/it, loss=0.246, sft=0.209, clean=0.215, trig=0.206, accum=4, step=79, kl=0.074]
SFT Epoch 1/2: 39%|ββββ | 79/205 [02:53<04:29, 2.14s/it, loss=0.246, sft=0.209, clean=0.215, trig=0.206, accum=4, step=79, kl=0.074]
SFT Epoch 1/2: 39%|ββββ | 79/205 [02:55<04:29, 2.14s/it, loss=0.265, sft=0.228, clean=0.232, trig=0.226, accum=4, step=80, kl=0.074]
SFT Epoch 1/2: 39%|ββββ | 80/205 [02:55<04:24, 2.12s/it, loss=0.265, sft=0.228, clean=0.232, trig=0.226, accum=4, step=80, kl=0.074]
SFT Epoch 1/2: 39%|ββββ | 80/205 [02:57<04:24, 2.12s/it, loss=0.172, sft=0.143, clean=0.149, trig=0.141, accum=4, step=81, kl=0.057]
SFT Epoch 1/2: 40%|ββββ | 81/205 [02:57<04:24, 2.14s/it, loss=0.172, sft=0.143, clean=0.149, trig=0.141, accum=4, step=81, kl=0.057]
SFT Epoch 1/2: 40%|ββββ | 81/205 [02:59<04:24, 2.14s/it, loss=0.170, sft=0.143, clean=0.147, trig=0.141, accum=4, step=82, kl=0.054]
SFT Epoch 1/2: 40%|ββββ | 82/205 [02:59<04:17, 2.10s/it, loss=0.170, sft=0.143, clean=0.147, trig=0.141, accum=4, step=82, kl=0.054]
SFT Epoch 1/2: 40%|ββββ | 82/205 [03:01<04:17, 2.10s/it, loss=0.222, sft=0.194, clean=0.194, trig=0.194, accum=4, step=83, kl=0.056]
SFT Epoch 1/2: 40%|ββββ | 83/205 [03:01<04:09, 2.04s/it, loss=0.222, sft=0.194, clean=0.194, trig=0.194, accum=4, step=83, kl=0.056]
SFT Epoch 1/2: 40%|ββββ | 83/205 [03:03<04:09, 2.04s/it, loss=0.186, sft=0.161, clean=0.166, trig=0.159, accum=4, step=84, kl=0.049]
SFT Epoch 1/2: 41%|ββββ | 84/205 [03:03<04:09, 2.06s/it, loss=0.186, sft=0.161, clean=0.166, trig=0.159, accum=4, step=84, kl=0.049]
SFT Epoch 1/2: 41%|ββββ | 84/205 [03:05<04:09, 2.06s/it, loss=0.227, sft=0.196, clean=0.199, trig=0.195, accum=4, step=85, kl=0.061]
SFT Epoch 1/2: 41%|βββββ | 85/205 [03:05<04:16, 2.14s/it, loss=0.227, sft=0.196, clean=0.199, trig=0.195, accum=4, step=85, kl=0.061]
SFT Epoch 1/2: 41%|βββββ | 85/205 [03:07<04:16, 2.14s/it, loss=0.196, sft=0.167, clean=0.172, trig=0.165, accum=4, step=86, kl=0.058]
SFT Epoch 1/2: 42%|βββββ | 86/205 [03:07<04:09, 2.10s/it, loss=0.196, sft=0.167, clean=0.172, trig=0.165, accum=4, step=86, kl=0.058]
SFT Epoch 1/2: 42%|βββββ | 86/205 [03:09<04:09, 2.10s/it, loss=0.224, sft=0.198, clean=0.200, trig=0.196, accum=4, step=87, kl=0.053]
SFT Epoch 1/2: 42%|βββββ | 87/205 [03:09<04:07, 2.10s/it, loss=0.224, sft=0.198, clean=0.200, trig=0.196, accum=4, step=87, kl=0.053]
SFT Epoch 1/2: 42%|βββββ | 87/205 [03:11<04:07, 2.10s/it, loss=0.136, sft=0.119, clean=0.119, trig=0.119, accum=4, step=88, kl=0.033]
SFT Epoch 1/2: 43%|βββββ | 88/205 [03:11<04:09, 2.13s/it, loss=0.136, sft=0.119, clean=0.119, trig=0.119, accum=4, step=88, kl=0.033]
SFT Epoch 1/2: 43%|βββββ | 88/205 [03:13<04:09, 2.13s/it, loss=0.184, sft=0.149, clean=0.152, trig=0.148, accum=4, step=89, kl=0.069]
SFT Epoch 1/2: 43%|βββββ | 89/205 [03:13<03:56, 2.04s/it, loss=0.184, sft=0.149, clean=0.152, trig=0.148, accum=4, step=89, kl=0.069]
SFT Epoch 1/2: 43%|βββββ | 89/205 [03:15<03:56, 2.04s/it, loss=0.180, sft=0.156, clean=0.158, trig=0.156, accum=4, step=90, kl=0.048]
SFT Epoch 1/2: 44%|βββββ | 90/205 [03:15<03:54, 2.04s/it, loss=0.180, sft=0.156, clean=0.158, trig=0.156, accum=4, step=90, kl=0.048]
SFT Epoch 1/2: 44%|βββββ | 90/205 [03:18<03:54, 2.04s/it, loss=0.178, sft=0.155, clean=0.155, trig=0.155, accum=4, step=91, kl=0.045]
SFT Epoch 1/2: 44%|βββββ | 91/205 [03:18<04:01, 2.12s/it, loss=0.178, sft=0.155, clean=0.155, trig=0.155, accum=4, step=91, kl=0.045]
SFT Epoch 1/2: 44%|βββββ | 91/205 [03:19<04:01, 2.12s/it, loss=0.132, sft=0.113, clean=0.114, trig=0.112, accum=4, step=92, kl=0.038]
SFT Epoch 1/2: 45%|βββββ | 92/205 [03:19<03:50, 2.04s/it, loss=0.132, sft=0.113, clean=0.114, trig=0.112, accum=4, step=92, kl=0.038]
SFT Epoch 1/2: 45%|βββββ | 92/205 [03:22<03:50, 2.04s/it, loss=0.240, sft=0.209, clean=0.214, trig=0.207, accum=4, step=93, kl=0.063]
SFT Epoch 1/2: 45%|βββββ | 93/205 [03:22<03:52, 2.07s/it, loss=0.240, sft=0.209, clean=0.214, trig=0.207, accum=4, step=93, kl=0.063]
SFT Epoch 1/2: 45%|βββββ | 93/205 [03:24<03:52, 2.07s/it, loss=0.265, sft=0.234, clean=0.241, trig=0.231, accum=4, step=94, kl=0.062]
SFT Epoch 1/2: 46%|βββββ | 94/205 [03:24<03:48, 2.06s/it, loss=0.265, sft=0.234, clean=0.241, trig=0.231, accum=4, step=94, kl=0.062]
SFT Epoch 1/2: 46%|βββββ | 94/205 [03:25<03:48, 2.06s/it, loss=0.215, sft=0.189, clean=0.192, trig=0.188, accum=4, step=95, kl=0.052]
SFT Epoch 1/2: 46%|βββββ | 95/205 [03:25<03:36, 1.97s/it, loss=0.215, sft=0.189, clean=0.192, trig=0.188, accum=4, step=95, kl=0.052]
SFT Epoch 1/2: 46%|βββββ | 95/205 [03:27<03:36, 1.97s/it, loss=0.194, sft=0.165, clean=0.167, trig=0.164, accum=4, step=96, kl=0.058]
SFT Epoch 1/2: 47%|βββββ | 96/205 [03:27<03:34, 1.96s/it, loss=0.194, sft=0.165, clean=0.167, trig=0.164, accum=4, step=96, kl=0.058]
SFT Epoch 1/2: 47%|βββββ | 96/205 [03:30<03:34, 1.96s/it, loss=0.207, sft=0.178, clean=0.185, trig=0.175, accum=4, step=97, kl=0.057]
SFT Epoch 1/2: 47%|βββββ | 97/205 [03:30<03:44, 2.08s/it, loss=0.207, sft=0.178, clean=0.185, trig=0.175, accum=4, step=97, kl=0.057]
SFT Epoch 1/2: 47%|βββββ | 97/205 [03:32<03:44, 2.08s/it, loss=0.244, sft=0.218, clean=0.222, trig=0.216, accum=4, step=98, kl=0.051]
SFT Epoch 1/2: 48%|βββββ | 98/205 [03:32<03:37, 2.03s/it, loss=0.244, sft=0.218, clean=0.222, trig=0.216, accum=4, step=98, kl=0.051]
SFT Epoch 1/2: 48%|βββββ | 98/205 [03:34<03:37, 2.03s/it, loss=0.284, sft=0.253, clean=0.265, trig=0.248, accum=4, step=99, kl=0.062]
SFT Epoch 1/2: 48%|βββββ | 99/205 [03:34<03:40, 2.08s/it, loss=0.284, sft=0.253, clean=0.265, trig=0.248, accum=4, step=99, kl=0.062]
SFT Epoch 1/2: 48%|βββββ | 99/205 [03:36<03:40, 2.08s/it, loss=0.174, sft=0.145, clean=0.147, trig=0.144, accum=4, step=100, kl=0.057]
SFT Epoch 1/2: 49%|βββββ | 100/205 [03:36<03:35, 2.05s/it, loss=0.174, sft=0.145, clean=0.147, trig=0.144, accum=4, step=100, kl=0.057]
SFT Epoch 1/2: 49%|βββββ | 100/205 [03:38<03:35, 2.05s/it, loss=0.224, sft=0.193, clean=0.198, trig=0.191, accum=4, step=101, kl=0.062]
SFT Epoch 1/2: 49%|βββββ | 101/205 [03:38<03:36, 2.08s/it, loss=0.224, sft=0.193, clean=0.198, trig=0.191, accum=4, step=101, kl=0.062]
SFT Epoch 1/2: 49%|βββββ | 101/205 [03:40<03:36, 2.08s/it, loss=0.249, sft=0.217, clean=0.223, trig=0.215, accum=4, step=102, kl=0.064]
SFT Epoch 1/2: 50%|βββββ | 102/205 [03:40<03:27, 2.02s/it, loss=0.249, sft=0.217, clean=0.223, trig=0.215, accum=4, step=102, kl=0.064]
SFT Epoch 1/2: 50%|βββββ | 102/205 [03:42<03:27, 2.02s/it, loss=0.165, sft=0.140, clean=0.142, trig=0.140, accum=4, step=103, kl=0.049]
SFT Epoch 1/2: 50%|βββββ | 103/205 [03:42<03:21, 1.97s/it, loss=0.165, sft=0.140, clean=0.142, trig=0.140, accum=4, step=103, kl=0.049]
SFT Epoch 1/2: 50%|βββββ | 103/205 [03:44<03:21, 1.97s/it, loss=0.213, sft=0.183, clean=0.183, trig=0.183, accum=4, step=104, kl=0.060]
SFT Epoch 1/2: 51%|βββββ | 104/205 [03:44<03:26, 2.04s/it, loss=0.213, sft=0.183, clean=0.183, trig=0.183, accum=4, step=104, kl=0.060]
SFT Epoch 1/2: 51%|βββββ | 104/205 [03:46<03:26, 2.04s/it, loss=0.213, sft=0.188, clean=0.188, trig=0.188, accum=4, step=105, kl=0.051]
SFT Epoch 1/2: 51%|βββββ | 105/205 [03:46<03:31, 2.12s/it, loss=0.213, sft=0.188, clean=0.188, trig=0.188, accum=4, step=105, kl=0.051]
SFT Epoch 1/2: 51%|βββββ | 105/205 [03:49<03:31, 2.12s/it, loss=0.232, sft=0.206, clean=0.209, trig=0.204, accum=4, step=106, kl=0.053]
SFT Epoch 1/2: 52%|ββββββ | 106/205 [03:49<03:43, 2.26s/it, loss=0.232, sft=0.206, clean=0.209, trig=0.204, accum=4, step=106, kl=0.053]
SFT Epoch 1/2: 52%|ββββββ | 106/205 [03:51<03:43, 2.26s/it, loss=0.226, sft=0.191, clean=0.195, trig=0.190, accum=4, step=107, kl=0.070]
SFT Epoch 1/2: 52%|ββββββ | 107/205 [03:51<03:34, 2.18s/it, loss=0.226, sft=0.191, clean=0.195, trig=0.190, accum=4, step=107, kl=0.070]
SFT Epoch 1/2: 52%|ββββββ | 107/205 [03:53<03:34, 2.18s/it, loss=0.275, sft=0.244, clean=0.249, trig=0.242, accum=4, step=108, kl=0.062]
SFT Epoch 1/2: 53%|ββββββ | 108/205 [03:53<03:27, 2.14s/it, loss=0.275, sft=0.244, clean=0.249, trig=0.242, accum=4, step=108, kl=0.062]
SFT Epoch 1/2: 53%|ββββββ | 108/205 [03:55<03:27, 2.14s/it, loss=0.192, sft=0.163, clean=0.164, trig=0.162, accum=4, step=109, kl=0.058]
SFT Epoch 1/2: 53%|ββββββ | 109/205 [03:55<03:14, 2.02s/it, loss=0.192, sft=0.163, clean=0.164, trig=0.162, accum=4, step=109, kl=0.058]
SFT Epoch 1/2: 53%|ββββββ | 109/205 [03:57<03:14, 2.02s/it, loss=0.280, sft=0.245, clean=0.245, trig=0.244, accum=4, step=110, kl=0.071]
SFT Epoch 1/2: 54%|ββββββ | 110/205 [03:57<03:11, 2.01s/it, loss=0.280, sft=0.245, clean=0.245, trig=0.244, accum=4, step=110, kl=0.071]
SFT Epoch 1/2: 54%|ββββββ | 110/205 [03:58<03:11, 2.01s/it, loss=0.267, sft=0.231, clean=0.239, trig=0.228, accum=4, step=111, kl=0.072]
SFT Epoch 1/2: 54%|ββββββ | 111/205 [03:58<03:08, 2.00s/it, loss=0.267, sft=0.231, clean=0.239, trig=0.228, accum=4, step=111, kl=0.072]
SFT Epoch 1/2: 54%|ββββββ | 111/205 [04:01<03:08, 2.00s/it, loss=0.177, sft=0.150, clean=0.154, trig=0.149, accum=4, step=112, kl=0.053]
SFT Epoch 1/2: 55%|ββββββ | 112/205 [04:01<03:08, 2.03s/it, loss=0.177, sft=0.150, clean=0.154, trig=0.149, accum=4, step=112, kl=0.053]
SFT Epoch 1/2: 55%|ββββββ | 112/205 [04:03<03:08, 2.03s/it, loss=0.311, sft=0.268, clean=0.276, trig=0.265, accum=4, step=113, kl=0.085]
SFT Epoch 1/2: 55%|ββββββ | 113/205 [04:03<03:14, 2.11s/it, loss=0.311, sft=0.268, clean=0.276, trig=0.265, accum=4, step=113, kl=0.085]
SFT Epoch 1/2: 55%|ββββββ | 113/205 [04:05<03:14, 2.11s/it, loss=0.283, sft=0.250, clean=0.257, trig=0.246, accum=4, step=114, kl=0.066]
SFT Epoch 1/2: 56%|ββββββ | 114/205 [04:05<03:23, 2.23s/it, loss=0.283, sft=0.250, clean=0.257, trig=0.246, accum=4, step=114, kl=0.066]
SFT Epoch 1/2: 56%|ββββββ | 114/205 [04:07<03:23, 2.23s/it, loss=0.259, sft=0.224, clean=0.229, trig=0.222, accum=4, step=115, kl=0.069]
SFT Epoch 1/2: 56%|ββββββ | 115/205 [04:07<03:14, 2.16s/it, loss=0.259, sft=0.224, clean=0.229, trig=0.222, accum=4, step=115, kl=0.069]
SFT Epoch 1/2: 56%|ββββββ | 115/205 [04:09<03:14, 2.16s/it, loss=0.256, sft=0.226, clean=0.226, trig=0.226, accum=4, step=116, kl=0.061]
SFT Epoch 1/2: 57%|ββββββ | 116/205 [04:09<03:06, 2.10s/it, loss=0.256, sft=0.226, clean=0.226, trig=0.226, accum=4, step=116, kl=0.061]
SFT Epoch 1/2: 57%|ββββββ | 116/205 [04:11<03:06, 2.10s/it, loss=0.263, sft=0.233, clean=0.229, trig=0.234, accum=4, step=117, kl=0.061]
SFT Epoch 1/2: 57%|ββββββ | 117/205 [04:11<02:58, 2.02s/it, loss=0.263, sft=0.233, clean=0.229, trig=0.234, accum=4, step=117, kl=0.061]
SFT Epoch 1/2: 57%|ββββββ | 117/205 [04:13<02:58, 2.02s/it, loss=0.214, sft=0.182, clean=0.186, trig=0.181, accum=4, step=118, kl=0.064]
SFT Epoch 1/2: 58%|ββββββ | 118/205 [04:13<02:54, 2.01s/it, loss=0.214, sft=0.182, clean=0.186, trig=0.181, accum=4, step=118, kl=0.064]
SFT Epoch 1/2: 58%|ββββββ | 118/205 [04:15<02:54, 2.01s/it, loss=0.203, sft=0.171, clean=0.177, trig=0.169, accum=4, step=119, kl=0.063]
SFT Epoch 1/2: 58%|ββββββ | 119/205 [04:15<03:00, 2.10s/it, loss=0.203, sft=0.171, clean=0.177, trig=0.169, accum=4, step=119, kl=0.063]
SFT Epoch 1/2: 58%|ββββββ | 119/205 [04:17<03:00, 2.10s/it, loss=0.297, sft=0.258, clean=0.264, trig=0.255, accum=4, step=120, kl=0.078]
SFT Epoch 1/2: 59%|ββββββ | 120/205 [04:17<02:50, 2.00s/it, loss=0.297, sft=0.258, clean=0.264, trig=0.255, accum=4, step=120, kl=0.078]
SFT Epoch 1/2: 59%|ββββββ | 120/205 [04:19<02:50, 2.00s/it, loss=0.210, sft=0.183, clean=0.182, trig=0.183, accum=4, step=121, kl=0.055]
SFT Epoch 1/2: 59%|ββββββ | 121/205 [04:19<02:49, 2.02s/it, loss=0.210, sft=0.183, clean=0.182, trig=0.183, accum=4, step=121, kl=0.055]
SFT Epoch 1/2: 59%|ββββββ | 121/205 [04:21<02:49, 2.02s/it, loss=0.183, sft=0.154, clean=0.160, trig=0.151, accum=4, step=122, kl=0.058]
SFT Epoch 1/2: 60%|ββββββ | 122/205 [04:21<02:42, 1.95s/it, loss=0.183, sft=0.154, clean=0.160, trig=0.151, accum=4, step=122, kl=0.058]
SFT Epoch 1/2: 60%|ββββββ | 122/205 [04:23<02:42, 1.95s/it, loss=0.156, sft=0.133, clean=0.136, trig=0.132, accum=4, step=123, kl=0.045]
SFT Epoch 1/2: 60%|ββββββ | 123/205 [04:23<02:40, 1.96s/it, loss=0.156, sft=0.133, clean=0.136, trig=0.132, accum=4, step=123, kl=0.045]
SFT Epoch 1/2: 60%|ββββββ | 123/205 [04:25<02:40, 1.96s/it, loss=0.206, sft=0.184, clean=0.172, trig=0.189, accum=4, step=124, kl=0.044]
SFT Epoch 1/2: 60%|ββββββ | 124/205 [04:25<02:47, 2.06s/it, loss=0.206, sft=0.184, clean=0.172, trig=0.189, accum=4, step=124, kl=0.044]
SFT Epoch 1/2: 60%|ββββββ | 124/205 [04:27<02:47, 2.06s/it, loss=0.114, sft=0.095, clean=0.095, trig=0.095, accum=4, step=125, kl=0.038]
SFT Epoch 1/2: 61%|ββββββ | 125/205 [04:27<02:41, 2.02s/it, loss=0.114, sft=0.095, clean=0.095, trig=0.095, accum=4, step=125, kl=0.038]
SFT Epoch 1/2: 61%|ββββββ | 125/205 [04:29<02:41, 2.02s/it, loss=0.236, sft=0.204, clean=0.203, trig=0.204, accum=4, step=126, kl=0.065]
SFT Epoch 1/2: 61%|βββββββ | 126/205 [04:29<02:39, 2.01s/it, loss=0.236, sft=0.204, clean=0.203, trig=0.204, accum=4, step=126, kl=0.065]
SFT Epoch 1/2: 61%|βββββββ | 126/205 [04:32<02:39, 2.01s/it, loss=0.240, sft=0.210, clean=0.214, trig=0.208, accum=4, step=127, kl=0.061]
SFT Epoch 1/2: 62%|βββββββ | 127/205 [04:32<02:42, 2.09s/it, loss=0.240, sft=0.210, clean=0.214, trig=0.208, accum=4, step=127, kl=0.061]
SFT Epoch 1/2: 62%|βββββββ | 127/205 [04:34<02:42, 2.09s/it, loss=0.252, sft=0.217, clean=0.222, trig=0.216, accum=4, step=128, kl=0.069]
SFT Epoch 1/2: 62%|βββββββ | 128/205 [04:34<02:37, 2.05s/it, loss=0.252, sft=0.217, clean=0.222, trig=0.216, accum=4, step=128, kl=0.069]
SFT Epoch 1/2: 62%|βββββββ | 128/205 [04:36<02:37, 2.05s/it, loss=0.112, sft=0.092, clean=0.094, trig=0.091, accum=4, step=129, kl=0.040]
SFT Epoch 1/2: 63%|βββββββ | 129/205 [04:36<02:43, 2.15s/it, loss=0.112, sft=0.092, clean=0.094, trig=0.091, accum=4, step=129, kl=0.040]
SFT Epoch 1/2: 63%|βββββββ | 129/205 [04:38<02:43, 2.15s/it, loss=0.229, sft=0.200, clean=0.200, trig=0.200, accum=4, step=130, kl=0.058]
SFT Epoch 1/2: 63%|βββββββ | 130/205 [04:38<02:44, 2.20s/it, loss=0.229, sft=0.200, clean=0.200, trig=0.200, accum=4, step=130, kl=0.058]
SFT Epoch 1/2: 63%|βββββββ | 130/205 [04:40<02:44, 2.20s/it, loss=0.196, sft=0.169, clean=0.171, trig=0.168, accum=4, step=131, kl=0.054]
SFT Epoch 1/2: 64%|βββββββ | 131/205 [04:40<02:38, 2.14s/it, loss=0.196, sft=0.169, clean=0.171, trig=0.168, accum=4, step=131, kl=0.054]
SFT Epoch 1/2: 64%|βββββββ | 131/205 [04:43<02:38, 2.14s/it, loss=0.267, sft=0.237, clean=0.236, trig=0.238, accum=4, step=132, kl=0.060]
SFT Epoch 1/2: 64%|βββββββ | 132/205 [04:43<02:41, 2.21s/it, loss=0.267, sft=0.237, clean=0.236, trig=0.238, accum=4, step=132, kl=0.060]
SFT Epoch 1/2: 64%|βββββββ | 132/205 [04:45<02:41, 2.21s/it, loss=0.236, sft=0.206, clean=0.207, trig=0.206, accum=4, step=133, kl=0.059]
SFT Epoch 1/2: 65%|βββββββ | 133/205 [04:45<02:34, 2.14s/it, loss=0.236, sft=0.206, clean=0.207, trig=0.206, accum=4, step=133, kl=0.059]
SFT Epoch 1/2: 65%|βββββββ | 133/205 [04:47<02:34, 2.14s/it, loss=0.217, sft=0.192, clean=0.192, trig=0.192, accum=4, step=134, kl=0.049]
SFT Epoch 1/2: 65%|βββββββ | 134/205 [04:47<02:29, 2.10s/it, loss=0.217, sft=0.192, clean=0.192, trig=0.192, accum=4, step=134, kl=0.049]
SFT Epoch 1/2: 65%|βββββββ | 134/205 [04:49<02:29, 2.10s/it, loss=0.223, sft=0.190, clean=0.192, trig=0.189, accum=4, step=135, kl=0.066]
SFT Epoch 1/2: 66%|βββββββ | 135/205 [04:49<02:25, 2.09s/it, loss=0.223, sft=0.190, clean=0.192, trig=0.189, accum=4, step=135, kl=0.066]
SFT Epoch 1/2: 66%|βββββββ | 135/205 [04:51<02:25, 2.09s/it, loss=0.216, sft=0.184, clean=0.186, trig=0.183, accum=4, step=136, kl=0.064]
SFT Epoch 1/2: 66%|βββββββ | 136/205 [04:51<02:31, 2.20s/it, loss=0.216, sft=0.184, clean=0.186, trig=0.183, accum=4, step=136, kl=0.064]
SFT Epoch 1/2: 66%|βββββββ | 136/205 [04:53<02:31, 2.20s/it, loss=0.250, sft=0.215, clean=0.218, trig=0.213, accum=4, step=137, kl=0.070]
SFT Epoch 1/2: 67%|βββββββ | 137/205 [04:53<02:32, 2.24s/it, loss=0.250, sft=0.215, clean=0.218, trig=0.213, accum=4, step=137, kl=0.070]
SFT Epoch 1/2: 67%|βββββββ | 137/205 [04:56<02:32, 2.24s/it, loss=0.257, sft=0.224, clean=0.223, trig=0.225, accum=4, step=138, kl=0.065]
SFT Epoch 1/2: 67%|βββββββ | 138/205 [04:56<02:31, 2.26s/it, loss=0.257, sft=0.224, clean=0.223, trig=0.225, accum=4, step=138, kl=0.065]
SFT Epoch 1/2: 67%|βββββββ | 138/205 [04:58<02:31, 2.26s/it, loss=0.272, sft=0.236, clean=0.240, trig=0.234, accum=4, step=139, kl=0.072]
SFT Epoch 1/2: 68%|βββββββ | 139/205 [04:58<02:25, 2.20s/it, loss=0.272, sft=0.236, clean=0.240, trig=0.234, accum=4, step=139, kl=0.072]
SFT Epoch 1/2: 68%|βββββββ | 139/205 [05:00<02:25, 2.20s/it, loss=0.287, sft=0.252, clean=0.257, trig=0.250, accum=4, step=140, kl=0.070]
SFT Epoch 1/2: 68%|βββββββ | 140/205 [05:00<02:19, 2.14s/it, loss=0.287, sft=0.252, clean=0.257, trig=0.250, accum=4, step=140, kl=0.070]
SFT Epoch 1/2: 68%|βββββββ | 140/205 [05:02<02:19, 2.14s/it, loss=0.236, sft=0.205, clean=0.206, trig=0.205, accum=4, step=141, kl=0.061]
SFT Epoch 1/2: 69%|βββββββ | 141/205 [05:02<02:13, 2.09s/it, loss=0.236, sft=0.205, clean=0.206, trig=0.205, accum=4, step=141, kl=0.061]
SFT Epoch 1/2: 69%|βββββββ | 141/205 [05:04<02:13, 2.09s/it, loss=0.172, sft=0.149, clean=0.154, trig=0.147, accum=4, step=142, kl=0.047]
SFT Epoch 1/2: 69%|βββββββ | 142/205 [05:04<02:08, 2.04s/it, loss=0.172, sft=0.149, clean=0.154, trig=0.147, accum=4, step=142, kl=0.047]
SFT Epoch 1/2: 69%|βββββββ | 142/205 [05:06<02:08, 2.04s/it, loss=0.250, sft=0.220, clean=0.227, trig=0.217, accum=4, step=143, kl=0.061]
SFT Epoch 1/2: 70%|βββββββ | 143/205 [05:06<02:02, 1.97s/it, loss=0.250, sft=0.220, clean=0.227, trig=0.217, accum=4, step=143, kl=0.061]
SFT Epoch 1/2: 70%|βββββββ | 143/205 [05:08<02:02, 1.97s/it, loss=0.187, sft=0.157, clean=0.159, trig=0.156, accum=4, step=144, kl=0.059]
SFT Epoch 1/2: 70%|βββββββ | 144/205 [05:08<02:02, 2.01s/it, loss=0.187, sft=0.157, clean=0.159, trig=0.156, accum=4, step=144, kl=0.059]
SFT Epoch 1/2: 70%|βββββββ | 144/205 [05:10<02:02, 2.01s/it, loss=0.215, sft=0.186, clean=0.187, trig=0.185, accum=4, step=145, kl=0.059]
SFT Epoch 1/2: 71%|βββββββ | 145/205 [05:10<02:02, 2.05s/it, loss=0.215, sft=0.186, clean=0.187, trig=0.185, accum=4, step=145, kl=0.059]
SFT Epoch 1/2: 71%|βββββββ | 145/205 [05:12<02:02, 2.05s/it, loss=0.267, sft=0.230, clean=0.234, trig=0.229, accum=4, step=146, kl=0.074]
SFT Epoch 1/2: 71%|βββββββ | 146/205 [05:12<02:04, 2.10s/it, loss=0.267, sft=0.230, clean=0.234, trig=0.229, accum=4, step=146, kl=0.074]
SFT Epoch 1/2: 71%|βββββββ | 146/205 [05:14<02:04, 2.10s/it, loss=0.122, sft=0.097, clean=0.098, trig=0.097, accum=4, step=147, kl=0.050]
SFT Epoch 1/2: 72%|ββββββββ | 147/205 [05:14<01:58, 2.04s/it, loss=0.122, sft=0.097, clean=0.098, trig=0.097, accum=4, step=147, kl=0.050]
SFT Epoch 1/2: 72%|ββββββββ | 147/205 [05:16<01:58, 2.04s/it, loss=0.191, sft=0.158, clean=0.158, trig=0.158, accum=4, step=148, kl=0.066]
SFT Epoch 1/2: 72%|ββββββββ | 148/205 [05:16<02:01, 2.14s/it, loss=0.191, sft=0.158, clean=0.158, trig=0.158, accum=4, step=148, kl=0.066]
SFT Epoch 1/2: 72%|ββββββββ | 148/205 [05:19<02:01, 2.14s/it, loss=0.260, sft=0.226, clean=0.229, trig=0.225, accum=4, step=149, kl=0.068]
SFT Epoch 1/2: 73%|ββββββββ | 149/205 [05:19<02:02, 2.19s/it, loss=0.260, sft=0.226, clean=0.229, trig=0.225, accum=4, step=149, kl=0.068]
SFT Epoch 1/2: 73%|ββββββββ | 149/205 [05:21<02:02, 2.19s/it, loss=0.107, sft=0.089, clean=0.090, trig=0.088, accum=4, step=150, kl=0.037]
SFT Epoch 1/2: 73%|ββββββββ | 150/205 [05:21<02:00, 2.19s/it, loss=0.107, sft=0.089, clean=0.090, trig=0.088, accum=4, step=150, kl=0.037]
SFT Epoch 1/2: 73%|ββββββββ | 150/205 [05:22<02:00, 2.19s/it, loss=0.268, sft=0.230, clean=0.234, trig=0.229, accum=4, step=151, kl=0.075]
SFT Epoch 1/2: 74%|ββββββββ | 151/205 [05:22<01:49, 2.03s/it, loss=0.268, sft=0.230, clean=0.234, trig=0.229, accum=4, step=151, kl=0.075]
SFT Epoch 1/2: 74%|ββββββββ | 151/205 [05:24<01:49, 2.03s/it, loss=0.273, sft=0.239, clean=0.237, trig=0.239, accum=4, step=152, kl=0.069]
SFT Epoch 1/2: 74%|ββββββββ | 152/205 [05:24<01:47, 2.02s/it, loss=0.273, sft=0.239, clean=0.237, trig=0.239, accum=4, step=152, kl=0.069]
SFT Epoch 1/2: 74%|ββββββββ | 152/205 [05:27<01:47, 2.02s/it, loss=0.184, sft=0.155, clean=0.159, trig=0.153, accum=4, step=153, kl=0.059]
SFT Epoch 1/2: 75%|ββββββββ | 153/205 [05:27<01:49, 2.10s/it, loss=0.184, sft=0.155, clean=0.159, trig=0.153, accum=4, step=153, kl=0.059]
SFT Epoch 1/2: 75%|ββββββββ | 153/205 [05:29<01:49, 2.10s/it, loss=0.266, sft=0.233, clean=0.240, trig=0.230, accum=4, step=154, kl=0.067]
SFT Epoch 1/2: 75%|ββββββββ | 154/205 [05:29<01:43, 2.02s/it, loss=0.266, sft=0.233, clean=0.240, trig=0.230, accum=4, step=154, kl=0.067]
SFT Epoch 1/2: 75%|ββββββββ | 154/205 [05:31<01:43, 2.02s/it, loss=0.221, sft=0.193, clean=0.199, trig=0.191, accum=4, step=155, kl=0.054]
SFT Epoch 1/2: 76%|ββββββββ | 155/205 [05:31<01:41, 2.02s/it, loss=0.221, sft=0.193, clean=0.199, trig=0.191, accum=4, step=155, kl=0.054]
SFT Epoch 1/2: 76%|ββββββββ | 155/205 [05:33<01:41, 2.02s/it, loss=0.213, sft=0.181, clean=0.182, trig=0.180, accum=4, step=156, kl=0.064]
SFT Epoch 1/2: 76%|ββββββββ | 156/205 [05:33<01:41, 2.07s/it, loss=0.213, sft=0.181, clean=0.182, trig=0.180, accum=4, step=156, kl=0.064]
SFT Epoch 1/2: 76%|ββββββββ | 156/205 [05:35<01:41, 2.07s/it, loss=0.231, sft=0.205, clean=0.208, trig=0.203, accum=4, step=157, kl=0.052]
SFT Epoch 1/2: 77%|ββββββββ | 157/205 [05:35<01:36, 2.01s/it, loss=0.231, sft=0.205, clean=0.208, trig=0.203, accum=4, step=157, kl=0.052]
SFT Epoch 1/2: 77%|ββββββββ | 157/205 [05:37<01:36, 2.01s/it, loss=0.303, sft=0.264, clean=0.266, trig=0.263, accum=4, step=158, kl=0.078]
SFT Epoch 1/2: 77%|ββββββββ | 158/205 [05:37<01:35, 2.02s/it, loss=0.303, sft=0.264, clean=0.266, trig=0.263, accum=4, step=158, kl=0.078]
SFT Epoch 1/2: 77%|ββββββββ | 158/205 [05:39<01:35, 2.02s/it, loss=0.247, sft=0.214, clean=0.220, trig=0.212, accum=4, step=159, kl=0.065]
SFT Epoch 1/2: 78%|ββββββββ | 159/205 [05:39<01:31, 1.98s/it, loss=0.247, sft=0.214, clean=0.220, trig=0.212, accum=4, step=159, kl=0.065]
SFT Epoch 1/2: 78%|ββββββββ | 159/205 [05:41<01:31, 1.98s/it, loss=0.245, sft=0.212, clean=0.210, trig=0.212, accum=4, step=160, kl=0.066]
SFT Epoch 1/2: 78%|ββββββββ | 160/205 [05:41<01:29, 2.00s/it, loss=0.245, sft=0.212, clean=0.210, trig=0.212, accum=4, step=160, kl=0.066]
SFT Epoch 1/2: 78%|ββββββββ | 160/205 [05:43<01:29, 2.00s/it, loss=0.244, sft=0.211, clean=0.218, trig=0.208, accum=4, step=161, kl=0.065]
SFT Epoch 1/2: 79%|ββββββββ | 161/205 [05:43<01:26, 1.97s/it, loss=0.244, sft=0.211, clean=0.218, trig=0.208, accum=4, step=161, kl=0.065]
SFT Epoch 1/2: 79%|ββββββββ | 161/205 [05:45<01:26, 1.97s/it, loss=0.258, sft=0.224, clean=0.229, trig=0.223, accum=4, step=162, kl=0.067]
SFT Epoch 1/2: 79%|ββββββββ | 162/205 [05:45<01:28, 2.05s/it, loss=0.258, sft=0.224, clean=0.229, trig=0.223, accum=4, step=162, kl=0.067]
SFT Epoch 1/2: 79%|ββββββββ | 162/205 [05:47<01:28, 2.05s/it, loss=0.220, sft=0.189, clean=0.188, trig=0.189, accum=4, step=163, kl=0.062]
SFT Epoch 1/2: 80%|ββββββββ | 163/205 [05:47<01:27, 2.08s/it, loss=0.220, sft=0.189, clean=0.188, trig=0.189, accum=4, step=163, kl=0.062]
SFT Epoch 1/2: 80%|ββββββββ | 163/205 [05:49<01:27, 2.08s/it, loss=0.235, sft=0.199, clean=0.205, trig=0.196, accum=4, step=164, kl=0.073]
SFT Epoch 1/2: 80%|ββββββββ | 164/205 [05:49<01:26, 2.10s/it, loss=0.235, sft=0.199, clean=0.205, trig=0.196, accum=4, step=164, kl=0.073]
SFT Epoch 1/2: 80%|ββββββββ | 164/205 [05:51<01:26, 2.10s/it, loss=0.326, sft=0.290, clean=0.298, trig=0.287, accum=4, step=165, kl=0.073]
SFT Epoch 1/2: 80%|ββββββββ | 165/205 [05:51<01:21, 2.05s/it, loss=0.326, sft=0.290, clean=0.298, trig=0.287, accum=4, step=165, kl=0.073]
SFT Epoch 1/2: 80%|ββββββββ | 165/205 [05:53<01:21, 2.05s/it, loss=0.229, sft=0.196, clean=0.202, trig=0.194, accum=4, step=166, kl=0.065]
SFT Epoch 1/2: 81%|ββββββββ | 166/205 [05:53<01:19, 2.05s/it, loss=0.229, sft=0.196, clean=0.202, trig=0.194, accum=4, step=166, kl=0.065]
SFT Epoch 1/2: 81%|ββββββββ | 166/205 [05:55<01:19, 2.05s/it, loss=0.249, sft=0.215, clean=0.218, trig=0.214, accum=4, step=167, kl=0.068]
SFT Epoch 1/2: 81%|βββββββββ | 167/205 [05:55<01:19, 2.09s/it, loss=0.249, sft=0.215, clean=0.218, trig=0.214, accum=4, step=167, kl=0.068]
SFT Epoch 1/2: 81%|βββββββββ | 167/205 [05:57<01:19, 2.09s/it, loss=0.133, sft=0.108, clean=0.109, trig=0.107, accum=4, step=168, kl=0.050]
SFT Epoch 1/2: 82%|βββββββββ | 168/205 [05:57<01:18, 2.12s/it, loss=0.133, sft=0.108, clean=0.109, trig=0.107, accum=4, step=168, kl=0.050]
SFT Epoch 1/2: 82%|βββββββββ | 168/205 [05:59<01:18, 2.12s/it, loss=0.242, sft=0.212, clean=0.218, trig=0.209, accum=4, step=169, kl=0.061]
SFT Epoch 1/2: 82%|βββββββββ | 169/205 [05:59<01:14, 2.07s/it, loss=0.242, sft=0.212, clean=0.218, trig=0.209, accum=4, step=169, kl=0.061]
SFT Epoch 1/2: 82%|βββββββββ | 169/205 [06:01<01:14, 2.07s/it, loss=0.130, sft=0.113, clean=0.114, trig=0.112, accum=4, step=170, kl=0.035]
SFT Epoch 1/2: 83%|βββββββββ | 170/205 [06:01<01:13, 2.10s/it, loss=0.130, sft=0.113, clean=0.114, trig=0.112, accum=4, step=170, kl=0.035]
SFT Epoch 1/2: 83%|βββββββββ | 170/205 [06:04<01:13, 2.10s/it, loss=0.209, sft=0.174, clean=0.178, trig=0.172, accum=4, step=171, kl=0.070]
SFT Epoch 1/2: 83%|βββββββββ | 171/205 [06:04<01:12, 2.14s/it, loss=0.209, sft=0.174, clean=0.178, trig=0.172, accum=4, step=171, kl=0.070]
SFT Epoch 1/2: 83%|βββββββββ | 171/205 [06:06<01:12, 2.14s/it, loss=0.306, sft=0.267, clean=0.274, trig=0.265, accum=4, step=172, kl=0.078]
SFT Epoch 1/2: 84%|βββββββββ | 172/205 [06:06<01:07, 2.05s/it, loss=0.306, sft=0.267, clean=0.274, trig=0.265, accum=4, step=172, kl=0.078]
SFT Epoch 1/2: 84%|βββββββββ | 172/205 [06:08<01:07, 2.05s/it, loss=0.192, sft=0.160, clean=0.164, trig=0.159, accum=4, step=173, kl=0.064]
SFT Epoch 1/2: 84%|βββββββββ | 173/205 [06:08<01:06, 2.08s/it, loss=0.192, sft=0.160, clean=0.164, trig=0.159, accum=4, step=173, kl=0.064]
SFT Epoch 1/2: 84%|βββββββββ | 173/205 [06:10<01:06, 2.08s/it, loss=0.246, sft=0.208, clean=0.216, trig=0.205, accum=4, step=174, kl=0.077]
SFT Epoch 1/2: 85%|βββββββββ | 174/205 [06:10<01:02, 2.01s/it, loss=0.246, sft=0.208, clean=0.216, trig=0.205, accum=4, step=174, kl=0.077]
SFT Epoch 1/2: 85%|βββββββββ | 174/205 [06:12<01:02, 2.01s/it, loss=0.170, sft=0.139, clean=0.145, trig=0.137, accum=4, step=175, kl=0.061]
SFT Epoch 1/2: 85%|βββββββββ | 175/205 [06:12<01:01, 2.04s/it, loss=0.170, sft=0.139, clean=0.145, trig=0.137, accum=4, step=175, kl=0.061]
SFT Epoch 1/2: 85%|βββββββββ | 175/205 [06:14<01:01, 2.04s/it, loss=0.206, sft=0.173, clean=0.174, trig=0.172, accum=4, step=176, kl=0.066]
SFT Epoch 1/2: 86%|βββββββββ | 176/205 [06:14<00:57, 2.00s/it, loss=0.206, sft=0.173, clean=0.174, trig=0.172, accum=4, step=176, kl=0.066]
SFT Epoch 1/2: 86%|βββββββββ | 176/205 [06:16<00:57, 2.00s/it, loss=0.251, sft=0.213, clean=0.218, trig=0.210, accum=4, step=177, kl=0.077]
SFT Epoch 1/2: 86%|βββββββββ | 177/205 [06:16<00:55, 1.99s/it, loss=0.251, sft=0.213, clean=0.218, trig=0.210, accum=4, step=177, kl=0.077]
SFT Epoch 1/2: 86%|βββββββββ | 177/205 [06:18<00:55, 1.99s/it, loss=0.234, sft=0.203, clean=0.210, trig=0.200, accum=4, step=178, kl=0.062]
SFT Epoch 1/2: 87%|βββββββββ | 178/205 [06:18<00:54, 2.01s/it, loss=0.234, sft=0.203, clean=0.210, trig=0.200, accum=4, step=178, kl=0.062]
SFT Epoch 1/2: 87%|βββββββββ | 178/205 [06:20<00:54, 2.01s/it, loss=0.130, sft=0.107, clean=0.105, trig=0.108, accum=4, step=179, kl=0.046]
SFT Epoch 1/2: 87%|βββββββββ | 179/205 [06:20<00:55, 2.12s/it, loss=0.130, sft=0.107, clean=0.105, trig=0.108, accum=4, step=179, kl=0.046]
SFT Epoch 1/2: 87%|βββββββββ | 179/205 [06:22<00:55, 2.12s/it, loss=0.230, sft=0.202, clean=0.206, trig=0.200, accum=4, step=180, kl=0.055]
SFT Epoch 1/2: 88%|βββββββββ | 180/205 [06:22<00:52, 2.12s/it, loss=0.230, sft=0.202, clean=0.206, trig=0.200, accum=4, step=180, kl=0.055]
SFT Epoch 1/2: 88%|βββββββββ | 180/205 [06:24<00:52, 2.12s/it, loss=0.161, sft=0.132, clean=0.135, trig=0.131, accum=4, step=181, kl=0.058]
SFT Epoch 1/2: 88%|βββββββββ | 181/205 [06:24<00:51, 2.16s/it, loss=0.161, sft=0.132, clean=0.135, trig=0.131, accum=4, step=181, kl=0.058]
SFT Epoch 1/2: 88%|βββββββββ | 181/205 [06:26<00:51, 2.16s/it, loss=0.177, sft=0.144, clean=0.145, trig=0.144, accum=4, step=182, kl=0.066]
SFT Epoch 1/2: 89%|βββββββββ | 182/205 [06:26<00:48, 2.10s/it, loss=0.177, sft=0.144, clean=0.145, trig=0.144, accum=4, step=182, kl=0.066]
SFT Epoch 1/2: 89%|βββββββββ | 182/205 [06:28<00:48, 2.10s/it, loss=0.238, sft=0.202, clean=0.204, trig=0.201, accum=4, step=183, kl=0.073]
SFT Epoch 1/2: 89%|βββββββββ | 183/205 [06:28<00:44, 2.04s/it, loss=0.238, sft=0.202, clean=0.204, trig=0.201, accum=4, step=183, kl=0.073]
SFT Epoch 1/2: 89%|βββββββββ | 183/205 [06:30<00:44, 2.04s/it, loss=0.290, sft=0.254, clean=0.262, trig=0.251, accum=4, step=184, kl=0.071]
SFT Epoch 1/2: 90%|βββββββββ | 184/205 [06:30<00:43, 2.05s/it, loss=0.290, sft=0.254, clean=0.262, trig=0.251, accum=4, step=184, kl=0.071]
SFT Epoch 1/2: 90%|βββββββββ | 184/205 [06:32<00:43, 2.05s/it, loss=0.258, sft=0.219, clean=0.225, trig=0.216, accum=4, step=185, kl=0.078]
SFT Epoch 1/2: 90%|βββββββββ | 185/205 [06:32<00:40, 2.02s/it, loss=0.258, sft=0.219, clean=0.225, trig=0.216, accum=4, step=185, kl=0.078]
SFT Epoch 1/2: 90%|βββββββββ | 185/205 [06:34<00:40, 2.02s/it, loss=0.177, sft=0.154, clean=0.156, trig=0.153, accum=4, step=186, kl=0.047]
SFT Epoch 1/2: 91%|βββββββββ | 186/205 [06:34<00:38, 2.01s/it, loss=0.177, sft=0.154, clean=0.156, trig=0.153, accum=4, step=186, kl=0.047]
SFT Epoch 1/2: 91%|βββββββββ | 186/205 [06:36<00:38, 2.01s/it, loss=0.250, sft=0.220, clean=0.225, trig=0.217, accum=4, step=187, kl=0.060]
SFT Epoch 1/2: 91%|βββββββββ | 187/205 [06:36<00:36, 2.05s/it, loss=0.250, sft=0.220, clean=0.225, trig=0.217, accum=4, step=187, kl=0.060]
SFT Epoch 1/2: 91%|βββββββββ | 187/205 [06:39<00:36, 2.05s/it, loss=0.207, sft=0.179, clean=0.186, trig=0.176, accum=4, step=188, kl=0.056]
SFT Epoch 1/2: 92%|ββββββββββ| 188/205 [06:39<00:35, 2.11s/it, loss=0.207, sft=0.179, clean=0.186, trig=0.176, accum=4, step=188, kl=0.056]
SFT Epoch 1/2: 92%|ββββββββββ| 188/205 [06:41<00:35, 2.11s/it, loss=0.236, sft=0.204, clean=0.207, trig=0.202, accum=4, step=189, kl=0.065]
SFT Epoch 1/2: 92%|ββββββββββ| 189/205 [06:41<00:34, 2.15s/it, loss=0.236, sft=0.204, clean=0.207, trig=0.202, accum=4, step=189, kl=0.065]
SFT Epoch 1/2: 92%|ββββββββββ| 189/205 [06:43<00:34, 2.15s/it, loss=0.275, sft=0.231, clean=0.233, trig=0.231, accum=4, step=190, kl=0.088]
SFT Epoch 1/2: 93%|ββββββββββ| 190/205 [06:43<00:31, 2.09s/it, loss=0.275, sft=0.231, clean=0.233, trig=0.231, accum=4, step=190, kl=0.088]
SFT Epoch 1/2: 93%|ββββββββββ| 190/205 [06:45<00:31, 2.09s/it, loss=0.145, sft=0.114, clean=0.118, trig=0.113, accum=4, step=191, kl=0.061]
SFT Epoch 1/2: 93%|ββββββββββ| 191/205 [06:45<00:29, 2.09s/it, loss=0.145, sft=0.114, clean=0.118, trig=0.113, accum=4, step=191, kl=0.061]
SFT Epoch 1/2: 93%|ββββββββββ| 191/205 [06:47<00:29, 2.09s/it, loss=0.184, sft=0.156, clean=0.161, trig=0.154, accum=4, step=192, kl=0.056]
SFT Epoch 1/2: 94%|ββββββββββ| 192/205 [06:47<00:26, 2.04s/it, loss=0.184, sft=0.156, clean=0.161, trig=0.154, accum=4, step=192, kl=0.056]
SFT Epoch 1/2: 94%|ββββββββββ| 192/205 [06:49<00:26, 2.04s/it, loss=0.190, sft=0.161, clean=0.161, trig=0.161, accum=4, step=193, kl=0.057]
SFT Epoch 1/2: 94%|ββββββββββ| 193/205 [06:49<00:23, 1.99s/it, loss=0.190, sft=0.161, clean=0.161, trig=0.161, accum=4, step=193, kl=0.057]
SFT Epoch 1/2: 94%|ββββββββββ| 193/205 [06:51<00:23, 1.99s/it, loss=0.279, sft=0.234, clean=0.242, trig=0.230, accum=4, step=194, kl=0.090]
SFT Epoch 1/2: 95%|ββββββββββ| 194/205 [06:51<00:21, 1.94s/it, loss=0.279, sft=0.234, clean=0.242, trig=0.230, accum=4, step=194, kl=0.090]
SFT Epoch 1/2: 95%|ββββββββββ| 194/205 [06:53<00:21, 1.94s/it, loss=0.268, sft=0.233, clean=0.239, trig=0.230, accum=4, step=195, kl=0.071]
SFT Epoch 1/2: 95%|ββββββββββ| 195/205 [06:53<00:20, 2.03s/it, loss=0.268, sft=0.233, clean=0.239, trig=0.230, accum=4, step=195, kl=0.071]
SFT Epoch 1/2: 95%|ββββββββββ| 195/205 [06:55<00:20, 2.03s/it, loss=0.201, sft=0.169, clean=0.168, trig=0.169, accum=4, step=196, kl=0.065]
SFT Epoch 1/2: 96%|ββββββββββ| 196/205 [06:55<00:18, 2.10s/it, loss=0.201, sft=0.169, clean=0.168, trig=0.169, accum=4, step=196, kl=0.065]
SFT Epoch 1/2: 96%|ββββββββββ| 196/205 [06:58<00:18, 2.10s/it, loss=0.242, sft=0.203, clean=0.206, trig=0.202, accum=4, step=197, kl=0.078]
SFT Epoch 1/2: 96%|ββββββββββ| 197/205 [06:58<00:17, 2.25s/it, loss=0.242, sft=0.203, clean=0.206, trig=0.202, accum=4, step=197, kl=0.078]
SFT Epoch 1/2: 96%|ββββββββββ| 197/205 [06:59<00:17, 2.25s/it, loss=0.231, sft=0.190, clean=0.189, trig=0.191, accum=4, step=198, kl=0.082]
SFT Epoch 1/2: 97%|ββββββββββ| 198/205 [06:59<00:14, 2.11s/it, loss=0.231, sft=0.190, clean=0.189, trig=0.191, accum=4, step=198, kl=0.082]
SFT Epoch 1/2: 97%|ββββββββββ| 198/205 [07:01<00:14, 2.11s/it, loss=0.310, sft=0.265, clean=0.268, trig=0.263, accum=4, step=199, kl=0.092]
SFT Epoch 1/2: 97%|ββββββββββ| 199/205 [07:01<00:11, 1.99s/it, loss=0.310, sft=0.265, clean=0.268, trig=0.263, accum=4, step=199, kl=0.092]
SFT Epoch 1/2: 97%|ββββββββββ| 199/205 [07:03<00:11, 1.99s/it, loss=0.202, sft=0.171, clean=0.172, trig=0.171, accum=4, step=200, kl=0.062]
SFT Epoch 1/2: 98%|ββββββββββ| 200/205 [07:03<00:10, 2.02s/it, loss=0.202, sft=0.171, clean=0.172, trig=0.171, accum=4, step=200, kl=0.062]
SFT Epoch 1/2: 98%|ββββββββββ| 200/205 [07:05<00:10, 2.02s/it, loss=0.223, sft=0.190, clean=0.185, trig=0.192, accum=4, step=201, kl=0.066]
SFT Epoch 1/2: 98%|ββββββββββ| 201/205 [07:05<00:08, 2.03s/it, loss=0.223, sft=0.190, clean=0.185, trig=0.192, accum=4, step=201, kl=0.066]
SFT Epoch 1/2: 98%|ββββββββββ| 201/205 [07:07<00:08, 2.03s/it, loss=0.279, sft=0.242, clean=0.249, trig=0.239, accum=4, step=202, kl=0.075]
SFT Epoch 1/2: 99%|ββββββββββ| 202/205 [07:07<00:06, 2.01s/it, loss=0.279, sft=0.242, clean=0.249, trig=0.239, accum=4, step=202, kl=0.075]
SFT Epoch 1/2: 99%|ββββββββββ| 202/205 [07:09<00:06, 2.01s/it, loss=0.325, sft=0.284, clean=0.290, trig=0.282, accum=4, step=203, kl=0.081]
SFT Epoch 1/2: 99%|ββββββββββ| 203/205 [07:09<00:04, 2.02s/it, loss=0.325, sft=0.284, clean=0.290, trig=0.282, accum=4, step=203, kl=0.081]
SFT Epoch 1/2: 99%|ββββββββββ| 203/205 [07:11<00:04, 2.02s/it, loss=0.306, sft=0.261, clean=0.268, trig=0.258, accum=4, step=204, kl=0.090]
SFT Epoch 1/2: 100%|ββββββββββ| 204/205 [07:11<00:02, 2.01s/it, loss=0.306, sft=0.261, clean=0.268, trig=0.258, accum=4, step=204, kl=0.090]
SFT Epoch 1/2: 100%|ββββββββββ| 204/205 [07:13<00:02, 2.01s/it, loss=0.221, sft=0.190, clean=0.190, trig=0.190, accum=4, step=205, kl=0.062]
SFT Epoch 1/2: 100%|ββββββββββ| 205/205 [07:13<00:00, 2.07s/it, loss=0.221, sft=0.190, clean=0.190, trig=0.190, accum=4, step=205, kl=0.062]
SFT Epoch 1/2: 100%|ββββββββββ| 205/205 [07:13<00:00, 2.12s/it, loss=0.221, sft=0.190, clean=0.190, trig=0.190, accum=4, step=205, kl=0.062] |
|
SFT Epoch 2/2: 0%| | 0/205 [00:00<?, ?it/s]
SFT Epoch 2/2: 0%| | 0/205 [00:02<?, ?it/s, loss=0.262, sft=0.191, clean=0.196, trig=0.189, accum=4, step=206, kl=0.141]
SFT Epoch 2/2: 0%| | 1/205 [00:02<07:37, 2.24s/it, loss=0.262, sft=0.191, clean=0.196, trig=0.189, accum=4, step=206, kl=0.141]
SFT Epoch 2/2: 0%| | 1/205 [00:04<07:37, 2.24s/it, loss=0.217, sft=0.167, clean=0.166, trig=0.168, accum=4, step=207, kl=0.100]
SFT Epoch 2/2: 1%| | 2/205 [00:04<07:35, 2.24s/it, loss=0.217, sft=0.167, clean=0.166, trig=0.168, accum=4, step=207, kl=0.100]
SFT Epoch 2/2: 1%| | 2/205 [00:06<07:35, 2.24s/it, loss=0.205, sft=0.161, clean=0.164, trig=0.160, accum=4, step=208, kl=0.088]
SFT Epoch 2/2: 1%|β | 3/205 [00:06<07:12, 2.14s/it, loss=0.205, sft=0.161, clean=0.164, trig=0.160, accum=4, step=208, kl=0.088]
SFT Epoch 2/2: 1%|β | 3/205 [00:08<07:12, 2.14s/it, loss=0.187, sft=0.142, clean=0.142, trig=0.142, accum=4, step=209, kl=0.090]
SFT Epoch 2/2: 2%|β | 4/205 [00:08<06:44, 2.01s/it, loss=0.187, sft=0.142, clean=0.142, trig=0.142, accum=4, step=209, kl=0.090]
SFT Epoch 2/2: 2%|β | 4/205 [00:10<06:44, 2.01s/it, loss=0.172, sft=0.126, clean=0.126, trig=0.126, accum=4, step=210, kl=0.091]
SFT Epoch 2/2: 2%|β | 5/205 [00:10<06:47, 2.04s/it, loss=0.172, sft=0.126, clean=0.126, trig=0.126, accum=4, step=210, kl=0.091]
SFT Epoch 2/2: 2%|β | 5/205 [00:12<06:47, 2.04s/it, loss=0.179, sft=0.136, clean=0.139, trig=0.135, accum=4, step=211, kl=0.084]
SFT Epoch 2/2: 3%|β | 6/205 [00:12<06:48, 2.05s/it, loss=0.179, sft=0.136, clean=0.139, trig=0.135, accum=4, step=211, kl=0.084]
SFT Epoch 2/2: 3%|β | 6/205 [00:14<06:48, 2.05s/it, loss=0.189, sft=0.151, clean=0.155, trig=0.149, accum=4, step=212, kl=0.075]
SFT Epoch 2/2: 3%|β | 7/205 [00:14<07:06, 2.15s/it, loss=0.189, sft=0.151, clean=0.155, trig=0.149, accum=4, step=212, kl=0.075]
SFT Epoch 2/2: 3%|β | 7/205 [00:17<07:06, 2.15s/it, loss=0.206, sft=0.145, clean=0.143, trig=0.145, accum=4, step=213, kl=0.122]
SFT Epoch 2/2: 4%|β | 8/205 [00:17<07:06, 2.16s/it, loss=0.206, sft=0.145, clean=0.143, trig=0.145, accum=4, step=213, kl=0.122]
SFT Epoch 2/2: 4%|β | 8/205 [00:19<07:06, 2.16s/it, loss=0.239, sft=0.189, clean=0.194, trig=0.187, accum=4, step=214, kl=0.101]
SFT Epoch 2/2: 4%|β | 9/205 [00:19<07:20, 2.25s/it, loss=0.239, sft=0.189, clean=0.194, trig=0.187, accum=4, step=214, kl=0.101]
SFT Epoch 2/2: 4%|β | 9/205 [00:21<07:20, 2.25s/it, loss=0.239, sft=0.195, clean=0.199, trig=0.193, accum=4, step=215, kl=0.089]
SFT Epoch 2/2: 5%|β | 10/205 [00:21<07:20, 2.26s/it, loss=0.239, sft=0.195, clean=0.199, trig=0.193, accum=4, step=215, kl=0.089]
SFT Epoch 2/2: 5%|β | 10/205 [00:23<07:20, 2.26s/it, loss=0.212, sft=0.142, clean=0.146, trig=0.141, accum=4, step=216, kl=0.140]
SFT Epoch 2/2: 5%|β | 11/205 [00:23<06:49, 2.11s/it, loss=0.212, sft=0.142, clean=0.146, trig=0.141, accum=4, step=216, kl=0.140]
SFT Epoch 2/2: 5%|β | 11/205 [00:25<06:49, 2.11s/it, loss=0.220, sft=0.181, clean=0.186, trig=0.179, accum=4, step=217, kl=0.079]
SFT Epoch 2/2: 6%|β | 12/205 [00:25<06:55, 2.15s/it, loss=0.220, sft=0.181, clean=0.186, trig=0.179, accum=4, step=217, kl=0.079]
SFT Epoch 2/2: 6%|β | 12/205 [00:28<06:55, 2.15s/it, loss=0.259, sft=0.212, clean=0.216, trig=0.210, accum=4, step=218, kl=0.095]
SFT Epoch 2/2: 6%|β | 13/205 [00:28<06:58, 2.18s/it, loss=0.259, sft=0.212, clean=0.216, trig=0.210, accum=4, step=218, kl=0.095]
SFT Epoch 2/2: 6%|β | 13/205 [00:30<06:58, 2.18s/it, loss=0.284, sft=0.236, clean=0.241, trig=0.233, accum=4, step=219, kl=0.096]
SFT Epoch 2/2: 7%|β | 14/205 [00:30<06:56, 2.18s/it, loss=0.284, sft=0.236, clean=0.241, trig=0.233, accum=4, step=219, kl=0.096]
SFT Epoch 2/2: 7%|β | 14/205 [00:31<06:56, 2.18s/it, loss=0.192, sft=0.109, clean=0.113, trig=0.108, accum=4, step=220, kl=0.166]
SFT Epoch 2/2: 7%|β | 15/205 [00:31<06:32, 2.07s/it, loss=0.192, sft=0.109, clean=0.113, trig=0.108, accum=4, step=220, kl=0.166]
SFT Epoch 2/2: 7%|β | 15/205 [00:33<06:32, 2.07s/it, loss=0.165, sft=0.119, clean=0.125, trig=0.117, accum=4, step=221, kl=0.091]
SFT Epoch 2/2: 8%|β | 16/205 [00:33<06:26, 2.04s/it, loss=0.165, sft=0.119, clean=0.125, trig=0.117, accum=4, step=221, kl=0.091]
SFT Epoch 2/2: 8%|β | 16/205 [00:35<06:26, 2.04s/it, loss=0.210, sft=0.165, clean=0.169, trig=0.164, accum=4, step=222, kl=0.089]
SFT Epoch 2/2: 8%|β | 17/205 [00:35<06:15, 2.00s/it, loss=0.210, sft=0.165, clean=0.169, trig=0.164, accum=4, step=222, kl=0.089]
SFT Epoch 2/2: 8%|β | 17/205 [00:38<06:15, 2.00s/it, loss=0.201, sft=0.157, clean=0.160, trig=0.156, accum=4, step=223, kl=0.088]
SFT Epoch 2/2: 9%|β | 18/205 [00:38<06:28, 2.08s/it, loss=0.201, sft=0.157, clean=0.160, trig=0.156, accum=4, step=223, kl=0.088]
SFT Epoch 2/2: 9%|β | 18/205 [00:39<06:28, 2.08s/it, loss=0.143, sft=0.108, clean=0.110, trig=0.107, accum=4, step=224, kl=0.070]
SFT Epoch 2/2: 9%|β | 19/205 [00:39<06:12, 2.00s/it, loss=0.143, sft=0.108, clean=0.110, trig=0.107, accum=4, step=224, kl=0.070]
SFT Epoch 2/2: 9%|β | 19/205 [00:41<06:12, 2.00s/it, loss=0.137, sft=0.112, clean=0.100, trig=0.117, accum=4, step=225, kl=0.051]
SFT Epoch 2/2: 10%|β | 20/205 [00:41<06:08, 1.99s/it, loss=0.137, sft=0.112, clean=0.100, trig=0.117, accum=4, step=225, kl=0.051]
SFT Epoch 2/2: 10%|β | 20/205 [00:43<06:08, 1.99s/it, loss=0.235, sft=0.170, clean=0.174, trig=0.169, accum=4, step=226, kl=0.129]
SFT Epoch 2/2: 10%|β | 21/205 [00:43<06:01, 1.97s/it, loss=0.235, sft=0.170, clean=0.174, trig=0.169, accum=4, step=226, kl=0.129]
SFT Epoch 2/2: 10%|β | 21/205 [00:45<06:01, 1.97s/it, loss=0.255, sft=0.216, clean=0.217, trig=0.216, accum=4, step=227, kl=0.077]
SFT Epoch 2/2: 11%|β | 22/205 [00:45<06:02, 1.98s/it, loss=0.255, sft=0.216, clean=0.217, trig=0.216, accum=4, step=227, kl=0.077]
SFT Epoch 2/2: 11%|β | 22/205 [00:47<06:02, 1.98s/it, loss=0.164, sft=0.141, clean=0.140, trig=0.141, accum=4, step=228, kl=0.047]
SFT Epoch 2/2: 11%|β | 23/205 [00:47<06:03, 1.99s/it, loss=0.164, sft=0.141, clean=0.140, trig=0.141, accum=4, step=228, kl=0.047]
SFT Epoch 2/2: 11%|β | 23/205 [00:50<06:03, 1.99s/it, loss=0.272, sft=0.222, clean=0.226, trig=0.221, accum=4, step=229, kl=0.099]
SFT Epoch 2/2: 12%|ββ | 24/205 [00:50<06:11, 2.05s/it, loss=0.272, sft=0.222, clean=0.226, trig=0.221, accum=4, step=229, kl=0.099]
SFT Epoch 2/2: 12%|ββ | 24/205 [00:51<06:11, 2.05s/it, loss=0.200, sft=0.171, clean=0.171, trig=0.171, accum=4, step=230, kl=0.058]
SFT Epoch 2/2: 12%|ββ | 25/205 [00:51<05:58, 1.99s/it, loss=0.200, sft=0.171, clean=0.171, trig=0.171, accum=4, step=230, kl=0.058]
SFT Epoch 2/2: 12%|ββ | 25/205 [00:54<05:58, 1.99s/it, loss=0.178, sft=0.136, clean=0.139, trig=0.135, accum=4, step=231, kl=0.084]
SFT Epoch 2/2: 13%|ββ | 26/205 [00:54<06:12, 2.08s/it, loss=0.178, sft=0.136, clean=0.139, trig=0.135, accum=4, step=231, kl=0.084]
SFT Epoch 2/2: 13%|ββ | 26/205 [00:56<06:12, 2.08s/it, loss=0.252, sft=0.214, clean=0.221, trig=0.211, accum=4, step=232, kl=0.076]
SFT Epoch 2/2: 13%|ββ | 27/205 [00:56<06:18, 2.13s/it, loss=0.252, sft=0.214, clean=0.221, trig=0.211, accum=4, step=232, kl=0.076]
SFT Epoch 2/2: 13%|ββ | 27/205 [00:58<06:18, 2.13s/it, loss=0.226, sft=0.167, clean=0.170, trig=0.166, accum=4, step=233, kl=0.117]
SFT Epoch 2/2: 14%|ββ | 28/205 [00:58<06:07, 2.07s/it, loss=0.226, sft=0.167, clean=0.170, trig=0.166, accum=4, step=233, kl=0.117]
SFT Epoch 2/2: 14%|ββ | 28/205 [01:00<06:07, 2.07s/it, loss=0.208, sft=0.159, clean=0.164, trig=0.157, accum=4, step=234, kl=0.097]
SFT Epoch 2/2: 14%|ββ | 29/205 [01:00<05:58, 2.04s/it, loss=0.208, sft=0.159, clean=0.164, trig=0.157, accum=4, step=234, kl=0.097]
SFT Epoch 2/2: 14%|ββ | 29/205 [01:02<05:58, 2.04s/it, loss=0.200, sft=0.131, clean=0.131, trig=0.130, accum=4, step=235, kl=0.138]
SFT Epoch 2/2: 15%|ββ | 30/205 [01:02<05:43, 1.96s/it, loss=0.200, sft=0.131, clean=0.131, trig=0.130, accum=4, step=235, kl=0.138]
SFT Epoch 2/2: 15%|ββ | 30/205 [01:04<05:43, 1.96s/it, loss=0.210, sft=0.172, clean=0.179, trig=0.169, accum=4, step=236, kl=0.075]
SFT Epoch 2/2: 15%|ββ | 31/205 [01:04<05:46, 1.99s/it, loss=0.210, sft=0.172, clean=0.179, trig=0.169, accum=4, step=236, kl=0.075]
SFT Epoch 2/2: 15%|ββ | 31/205 [01:06<05:46, 1.99s/it, loss=0.231, sft=0.191, clean=0.195, trig=0.189, accum=4, step=237, kl=0.081]
SFT Epoch 2/2: 16%|ββ | 32/205 [01:06<05:46, 2.00s/it, loss=0.231, sft=0.191, clean=0.195, trig=0.189, accum=4, step=237, kl=0.081]
SFT Epoch 2/2: 16%|ββ | 32/205 [01:08<05:46, 2.00s/it, loss=0.172, sft=0.137, clean=0.141, trig=0.135, accum=4, step=238, kl=0.070]
SFT Epoch 2/2: 16%|ββ | 33/205 [01:08<05:42, 1.99s/it, loss=0.172, sft=0.137, clean=0.141, trig=0.135, accum=4, step=238, kl=0.070]
SFT Epoch 2/2: 16%|ββ | 33/205 [01:10<05:42, 1.99s/it, loss=0.181, sft=0.105, clean=0.108, trig=0.104, accum=4, step=239, kl=0.153]
SFT Epoch 2/2: 17%|ββ | 34/205 [01:10<05:49, 2.04s/it, loss=0.181, sft=0.105, clean=0.108, trig=0.104, accum=4, step=239, kl=0.153]
SFT Epoch 2/2: 17%|ββ | 34/205 [01:12<05:49, 2.04s/it, loss=0.193, sft=0.142, clean=0.142, trig=0.142, accum=4, step=240, kl=0.101]
SFT Epoch 2/2: 17%|ββ | 35/205 [01:12<05:52, 2.07s/it, loss=0.193, sft=0.142, clean=0.142, trig=0.142, accum=4, step=240, kl=0.101]
SFT Epoch 2/2: 17%|ββ | 35/205 [01:14<05:52, 2.07s/it, loss=0.255, sft=0.169, clean=0.174, trig=0.167, accum=4, step=241, kl=0.171]
SFT Epoch 2/2: 18%|ββ | 36/205 [01:14<05:51, 2.08s/it, loss=0.255, sft=0.169, clean=0.174, trig=0.167, accum=4, step=241, kl=0.171]
SFT Epoch 2/2: 18%|ββ | 36/205 [01:16<05:51, 2.08s/it, loss=0.196, sft=0.153, clean=0.155, trig=0.152, accum=4, step=242, kl=0.085]
SFT Epoch 2/2: 18%|ββ | 37/205 [01:16<05:35, 2.00s/it, loss=0.196, sft=0.153, clean=0.155, trig=0.152, accum=4, step=242, kl=0.085]
SFT Epoch 2/2: 18%|ββ | 37/205 [01:18<05:35, 2.00s/it, loss=0.168, sft=0.136, clean=0.145, trig=0.133, accum=4, step=243, kl=0.064]
SFT Epoch 2/2: 19%|ββ | 38/205 [01:18<05:33, 1.99s/it, loss=0.168, sft=0.136, clean=0.145, trig=0.133, accum=4, step=243, kl=0.064]
SFT Epoch 2/2: 19%|ββ | 38/205 [01:20<05:33, 1.99s/it, loss=0.193, sft=0.137, clean=0.141, trig=0.135, accum=4, step=244, kl=0.112]
SFT Epoch 2/2: 19%|ββ | 39/205 [01:20<05:32, 2.00s/it, loss=0.193, sft=0.137, clean=0.141, trig=0.135, accum=4, step=244, kl=0.112]
SFT Epoch 2/2: 19%|ββ | 39/205 [01:22<05:32, 2.00s/it, loss=0.193, sft=0.158, clean=0.163, trig=0.156, accum=4, step=245, kl=0.070]
SFT Epoch 2/2: 20%|ββ | 40/205 [01:22<05:36, 2.04s/it, loss=0.193, sft=0.158, clean=0.163, trig=0.156, accum=4, step=245, kl=0.070]
SFT Epoch 2/2: 20%|ββ | 40/205 [01:24<05:36, 2.04s/it, loss=0.303, sft=0.248, clean=0.223, trig=0.259, accum=4, step=246, kl=0.109]
SFT Epoch 2/2: 20%|ββ | 41/205 [01:24<05:43, 2.10s/it, loss=0.303, sft=0.248, clean=0.223, trig=0.259, accum=4, step=246, kl=0.109]
SFT Epoch 2/2: 20%|ββ | 41/205 [01:26<05:43, 2.10s/it, loss=0.236, sft=0.202, clean=0.204, trig=0.200, accum=4, step=247, kl=0.069]
SFT Epoch 2/2: 20%|ββ | 42/205 [01:26<05:42, 2.10s/it, loss=0.236, sft=0.202, clean=0.204, trig=0.200, accum=4, step=247, kl=0.069]
SFT Epoch 2/2: 20%|ββ | 42/205 [01:28<05:42, 2.10s/it, loss=0.168, sft=0.108, clean=0.111, trig=0.107, accum=4, step=248, kl=0.120]
SFT Epoch 2/2: 21%|ββ | 43/205 [01:28<05:36, 2.08s/it, loss=0.168, sft=0.108, clean=0.111, trig=0.107, accum=4, step=248, kl=0.120]
SFT Epoch 2/2: 21%|ββ | 43/205 [01:31<05:36, 2.08s/it, loss=0.226, sft=0.181, clean=0.186, trig=0.179, accum=4, step=249, kl=0.089]
SFT Epoch 2/2: 21%|βββ | 44/205 [01:31<05:46, 2.15s/it, loss=0.226, sft=0.181, clean=0.186, trig=0.179, accum=4, step=249, kl=0.089]
SFT Epoch 2/2: 21%|βββ | 44/205 [01:33<05:46, 2.15s/it, loss=0.182, sft=0.128, clean=0.129, trig=0.128, accum=4, step=250, kl=0.108]
SFT Epoch 2/2: 22%|βββ | 45/205 [01:33<05:44, 2.15s/it, loss=0.182, sft=0.128, clean=0.129, trig=0.128, accum=4, step=250, kl=0.108]
SFT Epoch 2/2: 22%|βββ | 45/205 [01:35<05:44, 2.15s/it, loss=0.168, sft=0.114, clean=0.116, trig=0.113, accum=4, step=251, kl=0.108]
SFT Epoch 2/2: 22%|βββ | 46/205 [01:35<05:51, 2.21s/it, loss=0.168, sft=0.114, clean=0.116, trig=0.113, accum=4, step=251, kl=0.108]
SFT Epoch 2/2: 22%|βββ | 46/205 [01:37<05:51, 2.21s/it, loss=0.258, sft=0.199, clean=0.204, trig=0.197, accum=4, step=252, kl=0.118]
SFT Epoch 2/2: 23%|βββ | 47/205 [01:37<05:43, 2.17s/it, loss=0.258, sft=0.199, clean=0.204, trig=0.197, accum=4, step=252, kl=0.118]
SFT Epoch 2/2: 23%|βββ | 47/205 [01:39<05:43, 2.17s/it, loss=0.201, sft=0.160, clean=0.161, trig=0.159, accum=4, step=253, kl=0.082]
SFT Epoch 2/2: 23%|βββ | 48/205 [01:39<05:40, 2.17s/it, loss=0.201, sft=0.160, clean=0.161, trig=0.159, accum=4, step=253, kl=0.082]
SFT Epoch 2/2: 23%|βββ | 48/205 [01:41<05:40, 2.17s/it, loss=0.255, sft=0.187, clean=0.195, trig=0.184, accum=4, step=254, kl=0.135]
SFT Epoch 2/2: 24%|βββ | 49/205 [01:41<05:14, 2.02s/it, loss=0.255, sft=0.187, clean=0.195, trig=0.184, accum=4, step=254, kl=0.135]
SFT Epoch 2/2: 24%|βββ | 49/205 [01:43<05:14, 2.02s/it, loss=0.139, sft=0.096, clean=0.098, trig=0.095, accum=4, step=255, kl=0.086]
SFT Epoch 2/2: 24%|βββ | 50/205 [01:43<05:23, 2.08s/it, loss=0.139, sft=0.096, clean=0.098, trig=0.095, accum=4, step=255, kl=0.086]
SFT Epoch 2/2: 24%|βββ | 50/205 [01:45<05:23, 2.08s/it, loss=0.133, sft=0.095, clean=0.097, trig=0.094, accum=4, step=256, kl=0.076]
SFT Epoch 2/2: 25%|βββ | 51/205 [01:45<05:08, 2.00s/it, loss=0.133, sft=0.095, clean=0.097, trig=0.094, accum=4, step=256, kl=0.076]
SFT Epoch 2/2: 25%|βββ | 51/205 [01:47<05:08, 2.00s/it, loss=0.185, sft=0.132, clean=0.133, trig=0.132, accum=4, step=257, kl=0.107]
SFT Epoch 2/2: 25%|βββ | 52/205 [01:47<05:17, 2.07s/it, loss=0.185, sft=0.132, clean=0.133, trig=0.132, accum=4, step=257, kl=0.107]
SFT Epoch 2/2: 25%|βββ | 52/205 [01:50<05:17, 2.07s/it, loss=0.201, sft=0.133, clean=0.135, trig=0.132, accum=4, step=258, kl=0.135]
SFT Epoch 2/2: 26%|βββ | 53/205 [01:50<05:27, 2.15s/it, loss=0.201, sft=0.133, clean=0.135, trig=0.132, accum=4, step=258, kl=0.135]
SFT Epoch 2/2: 26%|βββ | 53/205 [01:52<05:27, 2.15s/it, loss=0.210, sft=0.177, clean=0.181, trig=0.175, accum=4, step=259, kl=0.066]
SFT Epoch 2/2: 26%|βββ | 54/205 [01:52<05:31, 2.20s/it, loss=0.210, sft=0.177, clean=0.181, trig=0.175, accum=4, step=259, kl=0.066]
SFT Epoch 2/2: 26%|βββ | 54/205 [01:54<05:31, 2.20s/it, loss=0.216, sft=0.182, clean=0.184, trig=0.182, accum=4, step=260, kl=0.066]
SFT Epoch 2/2: 27%|βββ | 55/205 [01:54<05:30, 2.20s/it, loss=0.216, sft=0.182, clean=0.184, trig=0.182, accum=4, step=260, kl=0.066]
SFT Epoch 2/2: 27%|βββ | 55/205 [01:56<05:30, 2.20s/it, loss=0.224, sft=0.192, clean=0.196, trig=0.190, accum=4, step=261, kl=0.064]
SFT Epoch 2/2: 27%|βββ | 56/205 [01:56<05:25, 2.18s/it, loss=0.224, sft=0.192, clean=0.196, trig=0.190, accum=4, step=261, kl=0.064]
SFT Epoch 2/2: 27%|βββ | 56/205 [01:59<05:25, 2.18s/it, loss=0.213, sft=0.157, clean=0.161, trig=0.155, accum=4, step=262, kl=0.111]
SFT Epoch 2/2: 28%|βββ | 57/205 [01:59<05:24, 2.19s/it, loss=0.213, sft=0.157, clean=0.161, trig=0.155, accum=4, step=262, kl=0.111]
SFT Epoch 2/2: 28%|βββ | 57/205 [02:01<05:24, 2.19s/it, loss=0.161, sft=0.139, clean=0.138, trig=0.139, accum=4, step=263, kl=0.045]
SFT Epoch 2/2: 28%|βββ | 58/205 [02:01<05:12, 2.12s/it, loss=0.161, sft=0.139, clean=0.138, trig=0.139, accum=4, step=263, kl=0.045]
SFT Epoch 2/2: 28%|βββ | 58/205 [02:03<05:12, 2.12s/it, loss=0.203, sft=0.167, clean=0.169, trig=0.166, accum=4, step=264, kl=0.073]
SFT Epoch 2/2: 29%|βββ | 59/205 [02:03<05:06, 2.10s/it, loss=0.203, sft=0.167, clean=0.169, trig=0.166, accum=4, step=264, kl=0.073]
SFT Epoch 2/2: 29%|βββ | 59/205 [02:04<05:06, 2.10s/it, loss=0.193, sft=0.122, clean=0.126, trig=0.120, accum=4, step=265, kl=0.142]
SFT Epoch 2/2: 29%|βββ | 60/205 [02:04<04:49, 2.00s/it, loss=0.193, sft=0.122, clean=0.126, trig=0.120, accum=4, step=265, kl=0.142]
SFT Epoch 2/2: 29%|βββ | 60/205 [02:06<04:49, 2.00s/it, loss=0.204, sft=0.146, clean=0.149, trig=0.144, accum=4, step=266, kl=0.116]
SFT Epoch 2/2: 30%|βββ | 61/205 [02:06<04:47, 2.00s/it, loss=0.204, sft=0.146, clean=0.149, trig=0.144, accum=4, step=266, kl=0.116]
SFT Epoch 2/2: 30%|βββ | 61/205 [02:08<04:47, 2.00s/it, loss=0.229, sft=0.186, clean=0.192, trig=0.184, accum=4, step=267, kl=0.085]
SFT Epoch 2/2: 30%|βββ | 62/205 [02:08<04:49, 2.02s/it, loss=0.229, sft=0.186, clean=0.192, trig=0.184, accum=4, step=267, kl=0.085]
SFT Epoch 2/2: 30%|βββ | 62/205 [02:10<04:49, 2.02s/it, loss=0.175, sft=0.145, clean=0.146, trig=0.145, accum=4, step=268, kl=0.060]
SFT Epoch 2/2: 31%|βββ | 63/205 [02:10<04:41, 1.98s/it, loss=0.175, sft=0.145, clean=0.146, trig=0.145, accum=4, step=268, kl=0.060]
SFT Epoch 2/2: 31%|βββ | 63/205 [02:13<04:41, 1.98s/it, loss=0.189, sft=0.128, clean=0.134, trig=0.126, accum=4, step=269, kl=0.121]
SFT Epoch 2/2: 31%|βββ | 64/205 [02:13<04:50, 2.06s/it, loss=0.189, sft=0.128, clean=0.134, trig=0.126, accum=4, step=269, kl=0.121]
SFT Epoch 2/2: 31%|βββ | 64/205 [02:15<04:50, 2.06s/it, loss=0.223, sft=0.150, clean=0.151, trig=0.150, accum=4, step=270, kl=0.146]
SFT Epoch 2/2: 32%|ββββ | 65/205 [02:15<04:45, 2.04s/it, loss=0.223, sft=0.150, clean=0.151, trig=0.150, accum=4, step=270, kl=0.146]
SFT Epoch 2/2: 32%|ββββ | 65/205 [02:16<04:45, 2.04s/it, loss=0.162, sft=0.122, clean=0.123, trig=0.122, accum=4, step=271, kl=0.079]
SFT Epoch 2/2: 32%|ββββ | 66/205 [02:16<04:37, 2.00s/it, loss=0.162, sft=0.122, clean=0.123, trig=0.122, accum=4, step=271, kl=0.079]
SFT Epoch 2/2: 32%|ββββ | 66/205 [02:19<04:37, 2.00s/it, loss=0.212, sft=0.159, clean=0.162, trig=0.159, accum=4, step=272, kl=0.104]
SFT Epoch 2/2: 33%|ββββ | 67/205 [02:19<04:41, 2.04s/it, loss=0.212, sft=0.159, clean=0.162, trig=0.159, accum=4, step=272, kl=0.104]
SFT Epoch 2/2: 33%|ββββ | 67/205 [02:21<04:41, 2.04s/it, loss=0.197, sft=0.169, clean=0.170, trig=0.168, accum=4, step=273, kl=0.056]
SFT Epoch 2/2: 33%|ββββ | 68/205 [02:21<04:37, 2.03s/it, loss=0.197, sft=0.169, clean=0.170, trig=0.168, accum=4, step=273, kl=0.056]
SFT Epoch 2/2: 33%|ββββ | 68/205 [02:23<04:37, 2.03s/it, loss=0.166, sft=0.134, clean=0.137, trig=0.132, accum=4, step=274, kl=0.065]
SFT Epoch 2/2: 34%|ββββ | 69/205 [02:23<04:32, 2.01s/it, loss=0.166, sft=0.134, clean=0.137, trig=0.132, accum=4, step=274, kl=0.065]
SFT Epoch 2/2: 34%|ββββ | 69/205 [02:25<04:32, 2.01s/it, loss=0.228, sft=0.175, clean=0.182, trig=0.171, accum=4, step=275, kl=0.106]
SFT Epoch 2/2: 34%|ββββ | 70/205 [02:25<04:40, 2.08s/it, loss=0.228, sft=0.175, clean=0.182, trig=0.171, accum=4, step=275, kl=0.106]
SFT Epoch 2/2: 34%|ββββ | 70/205 [02:27<04:40, 2.08s/it, loss=0.190, sft=0.146, clean=0.148, trig=0.145, accum=4, step=276, kl=0.088]
SFT Epoch 2/2: 35%|ββββ | 71/205 [02:27<04:39, 2.09s/it, loss=0.190, sft=0.146, clean=0.148, trig=0.145, accum=4, step=276, kl=0.088]
SFT Epoch 2/2: 35%|ββββ | 71/205 [02:29<04:39, 2.09s/it, loss=0.172, sft=0.138, clean=0.138, trig=0.138, accum=4, step=277, kl=0.068]
SFT Epoch 2/2: 35%|ββββ | 72/205 [02:29<04:27, 2.01s/it, loss=0.172, sft=0.138, clean=0.138, trig=0.138, accum=4, step=277, kl=0.068]
SFT Epoch 2/2: 35%|ββββ | 72/205 [02:31<04:27, 2.01s/it, loss=0.225, sft=0.176, clean=0.182, trig=0.174, accum=4, step=278, kl=0.097]
SFT Epoch 2/2: 36%|ββββ | 73/205 [02:31<04:22, 1.99s/it, loss=0.225, sft=0.176, clean=0.182, trig=0.174, accum=4, step=278, kl=0.097]
SFT Epoch 2/2: 36%|ββββ | 73/205 [02:33<04:22, 1.99s/it, loss=0.144, sft=0.117, clean=0.119, trig=0.117, accum=4, step=279, kl=0.053]
SFT Epoch 2/2: 36%|ββββ | 74/205 [02:33<04:29, 2.06s/it, loss=0.144, sft=0.117, clean=0.119, trig=0.117, accum=4, step=279, kl=0.053]
SFT Epoch 2/2: 36%|ββββ | 74/205 [02:35<04:29, 2.06s/it, loss=0.157, sft=0.109, clean=0.110, trig=0.109, accum=4, step=280, kl=0.096]
SFT Epoch 2/2: 37%|ββββ | 75/205 [02:35<04:20, 2.00s/it, loss=0.157, sft=0.109, clean=0.110, trig=0.109, accum=4, step=280, kl=0.096]
SFT Epoch 2/2: 37%|ββββ | 75/205 [02:36<04:20, 2.00s/it, loss=0.233, sft=0.184, clean=0.171, trig=0.190, accum=4, step=281, kl=0.096]
SFT Epoch 2/2: 37%|ββββ | 76/205 [02:36<04:03, 1.89s/it, loss=0.233, sft=0.184, clean=0.171, trig=0.190, accum=4, step=281, kl=0.096]
SFT Epoch 2/2: 37%|ββββ | 76/205 [02:38<04:03, 1.89s/it, loss=0.216, sft=0.176, clean=0.161, trig=0.183, accum=4, step=282, kl=0.080]
SFT Epoch 2/2: 38%|ββββ | 77/205 [02:38<03:59, 1.87s/it, loss=0.216, sft=0.176, clean=0.161, trig=0.183, accum=4, step=282, kl=0.080]
SFT Epoch 2/2: 38%|ββββ | 77/205 [02:40<03:59, 1.87s/it, loss=0.190, sft=0.155, clean=0.161, trig=0.152, accum=4, step=283, kl=0.070]
SFT Epoch 2/2: 38%|ββββ | 78/205 [02:40<04:08, 1.95s/it, loss=0.190, sft=0.155, clean=0.161, trig=0.152, accum=4, step=283, kl=0.070]
SFT Epoch 2/2: 38%|ββββ | 78/205 [02:42<04:08, 1.95s/it, loss=0.163, sft=0.128, clean=0.130, trig=0.127, accum=4, step=284, kl=0.069]
SFT Epoch 2/2: 39%|ββββ | 79/205 [02:42<04:00, 1.91s/it, loss=0.163, sft=0.128, clean=0.130, trig=0.127, accum=4, step=284, kl=0.069]
SFT Epoch 2/2: 39%|ββββ | 79/205 [02:44<04:00, 1.91s/it, loss=0.225, sft=0.172, clean=0.172, trig=0.171, accum=4, step=285, kl=0.108]
SFT Epoch 2/2: 39%|ββββ | 80/205 [02:44<04:00, 1.92s/it, loss=0.225, sft=0.172, clean=0.172, trig=0.171, accum=4, step=285, kl=0.108]
SFT Epoch 2/2: 39%|ββββ | 80/205 [02:46<04:00, 1.92s/it, loss=0.140, sft=0.097, clean=0.089, trig=0.101, accum=4, step=286, kl=0.086]
SFT Epoch 2/2: 40%|ββββ | 81/205 [02:46<03:58, 1.92s/it, loss=0.140, sft=0.097, clean=0.089, trig=0.101, accum=4, step=286, kl=0.086]
SFT Epoch 2/2: 40%|ββββ | 81/205 [02:48<03:58, 1.92s/it, loss=0.201, sft=0.174, clean=0.178, trig=0.172, accum=4, step=287, kl=0.055]
SFT Epoch 2/2: 40%|ββββ | 82/205 [02:48<04:02, 1.97s/it, loss=0.201, sft=0.174, clean=0.178, trig=0.172, accum=4, step=287, kl=0.055]
SFT Epoch 2/2: 40%|ββββ | 82/205 [02:50<04:02, 1.97s/it, loss=0.233, sft=0.189, clean=0.191, trig=0.189, accum=4, step=288, kl=0.087]
SFT Epoch 2/2: 40%|ββββ | 83/205 [02:50<04:13, 2.08s/it, loss=0.233, sft=0.189, clean=0.191, trig=0.189, accum=4, step=288, kl=0.087]
SFT Epoch 2/2: 40%|ββββ | 83/205 [02:53<04:13, 2.08s/it, loss=0.183, sft=0.133, clean=0.135, trig=0.131, accum=4, step=289, kl=0.101]
SFT Epoch 2/2: 41%|ββββ | 84/205 [02:53<04:12, 2.09s/it, loss=0.183, sft=0.133, clean=0.135, trig=0.131, accum=4, step=289, kl=0.101]
SFT Epoch 2/2: 41%|ββββ | 84/205 [02:55<04:12, 2.09s/it, loss=0.189, sft=0.150, clean=0.153, trig=0.148, accum=4, step=290, kl=0.079]
SFT Epoch 2/2: 41%|βββββ | 85/205 [02:55<04:10, 2.09s/it, loss=0.189, sft=0.150, clean=0.153, trig=0.148, accum=4, step=290, kl=0.079]
SFT Epoch 2/2: 41%|βββββ | 85/205 [02:57<04:10, 2.09s/it, loss=0.142, sft=0.085, clean=0.087, trig=0.084, accum=4, step=291, kl=0.114]
SFT Epoch 2/2: 42%|βββββ | 86/205 [02:57<04:09, 2.10s/it, loss=0.142, sft=0.085, clean=0.087, trig=0.084, accum=4, step=291, kl=0.114]
SFT Epoch 2/2: 42%|βββββ | 86/205 [02:59<04:09, 2.10s/it, loss=0.179, sft=0.137, clean=0.140, trig=0.135, accum=4, step=292, kl=0.085]
SFT Epoch 2/2: 42%|βββββ | 87/205 [02:59<04:10, 2.12s/it, loss=0.179, sft=0.137, clean=0.140, trig=0.135, accum=4, step=292, kl=0.085]
SFT Epoch 2/2: 42%|βββββ | 87/205 [03:01<04:10, 2.12s/it, loss=0.148, sft=0.113, clean=0.110, trig=0.115, accum=4, step=293, kl=0.070]
SFT Epoch 2/2: 43%|βββββ | 88/205 [03:01<04:01, 2.07s/it, loss=0.148, sft=0.113, clean=0.110, trig=0.115, accum=4, step=293, kl=0.070]
SFT Epoch 2/2: 43%|βββββ | 88/205 [03:03<04:01, 2.07s/it, loss=0.173, sft=0.135, clean=0.137, trig=0.134, accum=4, step=294, kl=0.075]
SFT Epoch 2/2: 43%|βββββ | 89/205 [03:03<03:56, 2.04s/it, loss=0.173, sft=0.135, clean=0.137, trig=0.134, accum=4, step=294, kl=0.075]
SFT Epoch 2/2: 43%|βββββ | 89/205 [03:05<03:56, 2.04s/it, loss=0.164, sft=0.138, clean=0.140, trig=0.137, accum=4, step=295, kl=0.053]
SFT Epoch 2/2: 44%|βββββ | 90/205 [03:05<03:56, 2.06s/it, loss=0.164, sft=0.138, clean=0.140, trig=0.137, accum=4, step=295, kl=0.053]
SFT Epoch 2/2: 44%|βββββ | 90/205 [03:07<03:56, 2.06s/it, loss=0.187, sft=0.149, clean=0.154, trig=0.147, accum=4, step=296, kl=0.076]
SFT Epoch 2/2: 44%|βββββ | 91/205 [03:07<03:59, 2.10s/it, loss=0.187, sft=0.149, clean=0.154, trig=0.147, accum=4, step=296, kl=0.076]
SFT Epoch 2/2: 44%|βββββ | 91/205 [03:09<03:59, 2.10s/it, loss=0.241, sft=0.201, clean=0.205, trig=0.199, accum=4, step=297, kl=0.081]
SFT Epoch 2/2: 45%|βββββ | 92/205 [03:09<03:55, 2.08s/it, loss=0.241, sft=0.201, clean=0.205, trig=0.199, accum=4, step=297, kl=0.081]
SFT Epoch 2/2: 45%|βββββ | 92/205 [03:11<03:55, 2.08s/it, loss=0.132, sft=0.100, clean=0.107, trig=0.097, accum=4, step=298, kl=0.064]
SFT Epoch 2/2: 45%|βββββ | 93/205 [03:11<03:57, 2.12s/it, loss=0.132, sft=0.100, clean=0.107, trig=0.097, accum=4, step=298, kl=0.064]
SFT Epoch 2/2: 45%|βββββ | 93/205 [03:13<03:57, 2.12s/it, loss=0.233, sft=0.164, clean=0.171, trig=0.161, accum=4, step=299, kl=0.138]
SFT Epoch 2/2: 46%|βββββ | 94/205 [03:13<03:51, 2.09s/it, loss=0.233, sft=0.164, clean=0.171, trig=0.161, accum=4, step=299, kl=0.138]
SFT Epoch 2/2: 46%|βββββ | 94/205 [03:16<03:51, 2.09s/it, loss=0.166, sft=0.128, clean=0.131, trig=0.126, accum=4, step=300, kl=0.076]
SFT Epoch 2/2: 46%|βββββ | 95/205 [03:16<03:56, 2.15s/it, loss=0.166, sft=0.128, clean=0.131, trig=0.126, accum=4, step=300, kl=0.076]
SFT Epoch 2/2: 46%|βββββ | 95/205 [03:18<03:56, 2.15s/it, loss=0.188, sft=0.139, clean=0.142, trig=0.139, accum=4, step=301, kl=0.096]
SFT Epoch 2/2: 47%|βββββ | 96/205 [03:18<04:04, 2.24s/it, loss=0.188, sft=0.139, clean=0.142, trig=0.139, accum=4, step=301, kl=0.096]
SFT Epoch 2/2: 47%|βββββ | 96/205 [03:20<04:04, 2.24s/it, loss=0.133, sft=0.110, clean=0.110, trig=0.109, accum=4, step=302, kl=0.047]
SFT Epoch 2/2: 47%|βββββ | 97/205 [03:20<04:00, 2.23s/it, loss=0.133, sft=0.110, clean=0.110, trig=0.109, accum=4, step=302, kl=0.047]
SFT Epoch 2/2: 47%|βββββ | 97/205 [03:23<04:00, 2.23s/it, loss=0.213, sft=0.102, clean=0.104, trig=0.102, accum=4, step=303, kl=0.220]
SFT Epoch 2/2: 48%|βββββ | 98/205 [03:23<03:55, 2.20s/it, loss=0.213, sft=0.102, clean=0.104, trig=0.102, accum=4, step=303, kl=0.220]
SFT Epoch 2/2: 48%|βββββ | 98/205 [03:24<03:55, 2.20s/it, loss=0.262, sft=0.191, clean=0.196, trig=0.189, accum=4, step=304, kl=0.143]
SFT Epoch 2/2: 48%|βββββ | 99/205 [03:24<03:36, 2.04s/it, loss=0.262, sft=0.191, clean=0.196, trig=0.189, accum=4, step=304, kl=0.143]
SFT Epoch 2/2: 48%|βββββ | 99/205 [03:27<03:36, 2.04s/it, loss=0.149, sft=0.096, clean=0.096, trig=0.096, accum=4, step=305, kl=0.106]
SFT Epoch 2/2: 49%|βββββ | 100/205 [03:27<03:42, 2.12s/it, loss=0.149, sft=0.096, clean=0.096, trig=0.096, accum=4, step=305, kl=0.106]
SFT Epoch 2/2: 49%|βββββ | 100/205 [03:29<03:42, 2.12s/it, loss=0.208, sft=0.160, clean=0.164, trig=0.158, accum=4, step=306, kl=0.097]
SFT Epoch 2/2: 49%|βββββ | 101/205 [03:29<03:45, 2.17s/it, loss=0.208, sft=0.160, clean=0.164, trig=0.158, accum=4, step=306, kl=0.097]
SFT Epoch 2/2: 49%|βββββ | 101/205 [03:31<03:45, 2.17s/it, loss=0.246, sft=0.165, clean=0.161, trig=0.167, accum=4, step=307, kl=0.161]
SFT Epoch 2/2: 50%|βββββ | 102/205 [03:31<03:34, 2.09s/it, loss=0.246, sft=0.165, clean=0.161, trig=0.167, accum=4, step=307, kl=0.161]
SFT Epoch 2/2: 50%|βββββ | 102/205 [03:33<03:34, 2.09s/it, loss=0.220, sft=0.143, clean=0.142, trig=0.144, accum=4, step=308, kl=0.154]
SFT Epoch 2/2: 50%|βββββ | 103/205 [03:33<03:34, 2.10s/it, loss=0.220, sft=0.143, clean=0.142, trig=0.144, accum=4, step=308, kl=0.154]
SFT Epoch 2/2: 50%|βββββ | 103/205 [03:35<03:34, 2.10s/it, loss=0.220, sft=0.173, clean=0.172, trig=0.173, accum=4, step=309, kl=0.094]
SFT Epoch 2/2: 51%|βββββ | 104/205 [03:35<03:30, 2.09s/it, loss=0.220, sft=0.173, clean=0.172, trig=0.173, accum=4, step=309, kl=0.094]
SFT Epoch 2/2: 51%|βββββ | 104/205 [03:37<03:30, 2.09s/it, loss=0.164, sft=0.140, clean=0.142, trig=0.139, accum=4, step=310, kl=0.048]
SFT Epoch 2/2: 51%|βββββ | 105/205 [03:37<03:24, 2.05s/it, loss=0.164, sft=0.140, clean=0.142, trig=0.139, accum=4, step=310, kl=0.048]
SFT Epoch 2/2: 51%|βββββ | 105/205 [03:39<03:24, 2.05s/it, loss=0.202, sft=0.158, clean=0.160, trig=0.157, accum=4, step=311, kl=0.088]
SFT Epoch 2/2: 52%|ββββββ | 106/205 [03:39<03:20, 2.03s/it, loss=0.202, sft=0.158, clean=0.160, trig=0.157, accum=4, step=311, kl=0.088]
SFT Epoch 2/2: 52%|ββββββ | 106/205 [03:41<03:20, 2.03s/it, loss=0.205, sft=0.170, clean=0.173, trig=0.169, accum=4, step=312, kl=0.070]
SFT Epoch 2/2: 52%|ββββββ | 107/205 [03:41<03:14, 1.98s/it, loss=0.205, sft=0.170, clean=0.173, trig=0.169, accum=4, step=312, kl=0.070]
SFT Epoch 2/2: 52%|ββββββ | 107/205 [03:43<03:14, 1.98s/it, loss=0.176, sft=0.106, clean=0.111, trig=0.104, accum=4, step=313, kl=0.138]
SFT Epoch 2/2: 53%|ββββββ | 108/205 [03:43<03:18, 2.05s/it, loss=0.176, sft=0.106, clean=0.111, trig=0.104, accum=4, step=313, kl=0.138]
SFT Epoch 2/2: 53%|ββββββ | 108/205 [03:45<03:18, 2.05s/it, loss=0.172, sft=0.114, clean=0.117, trig=0.113, accum=4, step=314, kl=0.115]
SFT Epoch 2/2: 53%|ββββββ | 109/205 [03:45<03:23, 2.12s/it, loss=0.172, sft=0.114, clean=0.117, trig=0.113, accum=4, step=314, kl=0.115]
SFT Epoch 2/2: 53%|ββββββ | 109/205 [03:47<03:23, 2.12s/it, loss=0.176, sft=0.133, clean=0.136, trig=0.132, accum=4, step=315, kl=0.086]
SFT Epoch 2/2: 54%|ββββββ | 110/205 [03:47<03:17, 2.07s/it, loss=0.176, sft=0.133, clean=0.136, trig=0.132, accum=4, step=315, kl=0.086]
SFT Epoch 2/2: 54%|ββββββ | 110/205 [03:49<03:17, 2.07s/it, loss=0.197, sft=0.149, clean=0.153, trig=0.147, accum=4, step=316, kl=0.096]
SFT Epoch 2/2: 54%|ββββββ | 111/205 [03:49<03:11, 2.04s/it, loss=0.197, sft=0.149, clean=0.153, trig=0.147, accum=4, step=316, kl=0.096]
SFT Epoch 2/2: 54%|ββββββ | 111/205 [03:51<03:11, 2.04s/it, loss=0.120, sft=0.075, clean=0.075, trig=0.074, accum=4, step=317, kl=0.090]
SFT Epoch 2/2: 55%|ββββββ | 112/205 [03:51<03:12, 2.07s/it, loss=0.120, sft=0.075, clean=0.075, trig=0.074, accum=4, step=317, kl=0.090]
SFT Epoch 2/2: 55%|ββββββ | 112/205 [03:53<03:12, 2.07s/it, loss=0.160, sft=0.102, clean=0.105, trig=0.101, accum=4, step=318, kl=0.117]
SFT Epoch 2/2: 55%|ββββββ | 113/205 [03:53<03:13, 2.10s/it, loss=0.160, sft=0.102, clean=0.105, trig=0.101, accum=4, step=318, kl=0.117]
SFT Epoch 2/2: 55%|ββββββ | 113/205 [03:56<03:13, 2.10s/it, loss=0.195, sft=0.136, clean=0.139, trig=0.134, accum=4, step=319, kl=0.119]
SFT Epoch 2/2: 56%|ββββββ | 114/205 [03:56<03:13, 2.13s/it, loss=0.195, sft=0.136, clean=0.139, trig=0.134, accum=4, step=319, kl=0.119]
SFT Epoch 2/2: 56%|ββββββ | 114/205 [03:58<03:13, 2.13s/it, loss=0.208, sft=0.150, clean=0.155, trig=0.148, accum=4, step=320, kl=0.116]
SFT Epoch 2/2: 56%|ββββββ | 115/205 [03:58<03:14, 2.16s/it, loss=0.208, sft=0.150, clean=0.155, trig=0.148, accum=4, step=320, kl=0.116]
SFT Epoch 2/2: 56%|ββββββ | 115/205 [04:00<03:14, 2.16s/it, loss=0.229, sft=0.177, clean=0.181, trig=0.175, accum=4, step=321, kl=0.105]
SFT Epoch 2/2: 57%|ββββββ | 116/205 [04:00<03:07, 2.10s/it, loss=0.229, sft=0.177, clean=0.181, trig=0.175, accum=4, step=321, kl=0.105]
SFT Epoch 2/2: 57%|ββββββ | 116/205 [04:02<03:07, 2.10s/it, loss=0.244, sft=0.179, clean=0.185, trig=0.176, accum=4, step=322, kl=0.130]
SFT Epoch 2/2: 57%|ββββββ | 117/205 [04:02<03:10, 2.16s/it, loss=0.244, sft=0.179, clean=0.185, trig=0.176, accum=4, step=322, kl=0.130]
SFT Epoch 2/2: 57%|ββββββ | 117/205 [04:04<03:10, 2.16s/it, loss=0.164, sft=0.131, clean=0.135, trig=0.129, accum=4, step=323, kl=0.067]
SFT Epoch 2/2: 58%|ββββββ | 118/205 [04:04<03:09, 2.18s/it, loss=0.164, sft=0.131, clean=0.135, trig=0.129, accum=4, step=323, kl=0.067]
SFT Epoch 2/2: 58%|ββββββ | 118/205 [04:07<03:09, 2.18s/it, loss=0.204, sft=0.170, clean=0.175, trig=0.167, accum=4, step=324, kl=0.069]
SFT Epoch 2/2: 58%|ββββββ | 119/205 [04:07<03:09, 2.20s/it, loss=0.204, sft=0.170, clean=0.175, trig=0.167, accum=4, step=324, kl=0.069]
SFT Epoch 2/2: 58%|ββββββ | 119/205 [04:09<03:09, 2.20s/it, loss=0.202, sft=0.170, clean=0.172, trig=0.170, accum=4, step=325, kl=0.063]
SFT Epoch 2/2: 59%|ββββββ | 120/205 [04:09<03:00, 2.12s/it, loss=0.202, sft=0.170, clean=0.172, trig=0.170, accum=4, step=325, kl=0.063]
SFT Epoch 2/2: 59%|ββββββ | 120/205 [04:11<03:00, 2.12s/it, loss=0.216, sft=0.141, clean=0.143, trig=0.140, accum=4, step=326, kl=0.150]
SFT Epoch 2/2: 59%|ββββββ | 121/205 [04:11<02:59, 2.14s/it, loss=0.216, sft=0.141, clean=0.143, trig=0.140, accum=4, step=326, kl=0.150]
SFT Epoch 2/2: 59%|ββββββ | 121/205 [04:13<02:59, 2.14s/it, loss=0.221, sft=0.172, clean=0.176, trig=0.170, accum=4, step=327, kl=0.098]
SFT Epoch 2/2: 60%|ββββββ | 122/205 [04:13<02:53, 2.09s/it, loss=0.221, sft=0.172, clean=0.176, trig=0.170, accum=4, step=327, kl=0.098]
SFT Epoch 2/2: 60%|ββββββ | 122/205 [04:15<02:53, 2.09s/it, loss=0.199, sft=0.148, clean=0.151, trig=0.146, accum=4, step=328, kl=0.102]
SFT Epoch 2/2: 60%|ββββββ | 123/205 [04:15<02:47, 2.04s/it, loss=0.199, sft=0.148, clean=0.151, trig=0.146, accum=4, step=328, kl=0.102]
SFT Epoch 2/2: 60%|ββββββ | 123/205 [04:17<02:47, 2.04s/it, loss=0.263, sft=0.189, clean=0.195, trig=0.187, accum=4, step=329, kl=0.149]
SFT Epoch 2/2: 60%|ββββββ | 124/205 [04:17<02:45, 2.05s/it, loss=0.263, sft=0.189, clean=0.195, trig=0.187, accum=4, step=329, kl=0.149]
SFT Epoch 2/2: 60%|ββββββ | 124/205 [04:19<02:45, 2.05s/it, loss=0.223, sft=0.159, clean=0.162, trig=0.158, accum=4, step=330, kl=0.128]
SFT Epoch 2/2: 61%|ββββββ | 125/205 [04:19<02:44, 2.05s/it, loss=0.223, sft=0.159, clean=0.162, trig=0.158, accum=4, step=330, kl=0.128]
SFT Epoch 2/2: 61%|ββββββ | 125/205 [04:21<02:44, 2.05s/it, loss=0.196, sft=0.152, clean=0.154, trig=0.151, accum=4, step=331, kl=0.089]
SFT Epoch 2/2: 61%|βββββββ | 126/205 [04:21<02:41, 2.05s/it, loss=0.196, sft=0.152, clean=0.154, trig=0.151, accum=4, step=331, kl=0.089]
SFT Epoch 2/2: 61%|βββββββ | 126/205 [04:23<02:41, 2.05s/it, loss=0.175, sft=0.131, clean=0.134, trig=0.130, accum=4, step=332, kl=0.087]
SFT Epoch 2/2: 62%|βββββββ | 127/205 [04:23<02:40, 2.06s/it, loss=0.175, sft=0.131, clean=0.134, trig=0.130, accum=4, step=332, kl=0.087]
SFT Epoch 2/2: 62%|βββββββ | 127/205 [04:25<02:40, 2.06s/it, loss=0.196, sft=0.161, clean=0.163, trig=0.159, accum=4, step=333, kl=0.071]
SFT Epoch 2/2: 62%|βββββββ | 128/205 [04:25<02:37, 2.05s/it, loss=0.196, sft=0.161, clean=0.163, trig=0.159, accum=4, step=333, kl=0.071]
SFT Epoch 2/2: 62%|βββββββ | 128/205 [04:27<02:37, 2.05s/it, loss=0.160, sft=0.115, clean=0.118, trig=0.114, accum=4, step=334, kl=0.089]
SFT Epoch 2/2: 63%|βββββββ | 129/205 [04:27<02:32, 2.00s/it, loss=0.160, sft=0.115, clean=0.118, trig=0.114, accum=4, step=334, kl=0.089]
SFT Epoch 2/2: 63%|βββββββ | 129/205 [04:29<02:32, 2.00s/it, loss=0.116, sft=0.098, clean=0.100, trig=0.097, accum=4, step=335, kl=0.035]
SFT Epoch 2/2: 63%|βββββββ | 130/205 [04:29<02:42, 2.16s/it, loss=0.116, sft=0.098, clean=0.100, trig=0.097, accum=4, step=335, kl=0.035]
SFT Epoch 2/2: 63%|βββββββ | 130/205 [04:32<02:42, 2.16s/it, loss=0.238, sft=0.179, clean=0.184, trig=0.177, accum=4, step=336, kl=0.118]
SFT Epoch 2/2: 64%|βββββββ | 131/205 [04:32<02:45, 2.23s/it, loss=0.238, sft=0.179, clean=0.184, trig=0.177, accum=4, step=336, kl=0.118]
SFT Epoch 2/2: 64%|βββββββ | 131/205 [04:34<02:45, 2.23s/it, loss=0.187, sft=0.142, clean=0.150, trig=0.139, accum=4, step=337, kl=0.089]
SFT Epoch 2/2: 64%|βββββββ | 132/205 [04:34<02:40, 2.20s/it, loss=0.187, sft=0.142, clean=0.150, trig=0.139, accum=4, step=337, kl=0.089]
SFT Epoch 2/2: 64%|βββββββ | 132/205 [04:36<02:40, 2.20s/it, loss=0.142, sft=0.116, clean=0.119, trig=0.115, accum=4, step=338, kl=0.052]
SFT Epoch 2/2: 65%|βββββββ | 133/205 [04:36<02:32, 2.11s/it, loss=0.142, sft=0.116, clean=0.119, trig=0.115, accum=4, step=338, kl=0.052]
SFT Epoch 2/2: 65%|βββββββ | 133/205 [04:37<02:32, 2.11s/it, loss=0.135, sft=0.113, clean=0.114, trig=0.112, accum=4, step=339, kl=0.045]
SFT Epoch 2/2: 65%|βββββββ | 134/205 [04:37<02:20, 1.97s/it, loss=0.135, sft=0.113, clean=0.114, trig=0.112, accum=4, step=339, kl=0.045]
SFT Epoch 2/2: 65%|βββββββ | 134/205 [04:39<02:20, 1.97s/it, loss=0.215, sft=0.182, clean=0.183, trig=0.181, accum=4, step=340, kl=0.068]
SFT Epoch 2/2: 66%|βββββββ | 135/205 [04:39<02:15, 1.93s/it, loss=0.215, sft=0.182, clean=0.183, trig=0.181, accum=4, step=340, kl=0.068]
SFT Epoch 2/2: 66%|βββββββ | 135/205 [04:41<02:15, 1.93s/it, loss=0.180, sft=0.125, clean=0.128, trig=0.124, accum=4, step=341, kl=0.110]
SFT Epoch 2/2: 66%|βββββββ | 136/205 [04:41<02:18, 2.00s/it, loss=0.180, sft=0.125, clean=0.128, trig=0.124, accum=4, step=341, kl=0.110]
SFT Epoch 2/2: 66%|βββββββ | 136/205 [04:43<02:18, 2.00s/it, loss=0.185, sft=0.143, clean=0.146, trig=0.142, accum=4, step=342, kl=0.085]
SFT Epoch 2/2: 67%|βββββββ | 137/205 [04:43<02:16, 2.00s/it, loss=0.185, sft=0.143, clean=0.146, trig=0.142, accum=4, step=342, kl=0.085]
SFT Epoch 2/2: 67%|βββββββ | 137/205 [04:46<02:16, 2.00s/it, loss=0.199, sft=0.156, clean=0.159, trig=0.155, accum=4, step=343, kl=0.086]
SFT Epoch 2/2: 67%|βββββββ | 138/205 [04:46<02:19, 2.08s/it, loss=0.199, sft=0.156, clean=0.159, trig=0.155, accum=4, step=343, kl=0.086]
SFT Epoch 2/2: 67%|βββββββ | 138/205 [04:47<02:19, 2.08s/it, loss=0.238, sft=0.193, clean=0.196, trig=0.191, accum=4, step=344, kl=0.090]
SFT Epoch 2/2: 68%|βββββββ | 139/205 [04:47<02:12, 2.01s/it, loss=0.238, sft=0.193, clean=0.196, trig=0.191, accum=4, step=344, kl=0.090]
SFT Epoch 2/2: 68%|βββββββ | 139/205 [04:50<02:12, 2.01s/it, loss=0.166, sft=0.123, clean=0.124, trig=0.122, accum=4, step=345, kl=0.086]
SFT Epoch 2/2: 68%|βββββββ | 140/205 [04:50<02:11, 2.02s/it, loss=0.166, sft=0.123, clean=0.124, trig=0.122, accum=4, step=345, kl=0.086]
SFT Epoch 2/2: 68%|βββββββ | 140/205 [04:51<02:11, 2.02s/it, loss=0.183, sft=0.117, clean=0.118, trig=0.116, accum=4, step=346, kl=0.133]
SFT Epoch 2/2: 69%|βββββββ | 141/205 [04:51<02:02, 1.92s/it, loss=0.183, sft=0.117, clean=0.118, trig=0.116, accum=4, step=346, kl=0.133]
SFT Epoch 2/2: 69%|βββββββ | 141/205 [04:53<02:02, 1.92s/it, loss=0.281, sft=0.238, clean=0.242, trig=0.236, accum=4, step=347, kl=0.086]
SFT Epoch 2/2: 69%|βββββββ | 142/205 [04:53<01:58, 1.87s/it, loss=0.281, sft=0.238, clean=0.242, trig=0.236, accum=4, step=347, kl=0.086]
SFT Epoch 2/2: 69%|βββββββ | 142/205 [04:56<01:58, 1.87s/it, loss=0.143, sft=0.104, clean=0.106, trig=0.104, accum=4, step=348, kl=0.077]
SFT Epoch 2/2: 70%|βββββββ | 143/205 [04:56<02:09, 2.08s/it, loss=0.143, sft=0.104, clean=0.106, trig=0.104, accum=4, step=348, kl=0.077]
SFT Epoch 2/2: 70%|βββββββ | 143/205 [04:58<02:09, 2.08s/it, loss=0.219, sft=0.185, clean=0.186, trig=0.184, accum=4, step=349, kl=0.069]
SFT Epoch 2/2: 70%|βββββββ | 144/205 [04:58<02:08, 2.11s/it, loss=0.219, sft=0.185, clean=0.186, trig=0.184, accum=4, step=349, kl=0.069]
SFT Epoch 2/2: 70%|βββββββ | 144/205 [05:00<02:08, 2.11s/it, loss=0.180, sft=0.124, clean=0.131, trig=0.122, accum=4, step=350, kl=0.111]
SFT Epoch 2/2: 71%|βββββββ | 145/205 [05:00<02:10, 2.17s/it, loss=0.180, sft=0.124, clean=0.131, trig=0.122, accum=4, step=350, kl=0.111]
SFT Epoch 2/2: 71%|βββββββ | 145/205 [05:02<02:10, 2.17s/it, loss=0.164, sft=0.131, clean=0.133, trig=0.131, accum=4, step=351, kl=0.065]
SFT Epoch 2/2: 71%|βββββββ | 146/205 [05:02<02:08, 2.18s/it, loss=0.164, sft=0.131, clean=0.133, trig=0.131, accum=4, step=351, kl=0.065]
SFT Epoch 2/2: 71%|βββββββ | 146/205 [05:04<02:08, 2.18s/it, loss=0.252, sft=0.168, clean=0.174, trig=0.166, accum=4, step=352, kl=0.166]
SFT Epoch 2/2: 72%|ββββββββ | 147/205 [05:04<02:06, 2.19s/it, loss=0.252, sft=0.168, clean=0.174, trig=0.166, accum=4, step=352, kl=0.166]
SFT Epoch 2/2: 72%|ββββββββ | 147/205 [05:07<02:06, 2.19s/it, loss=0.151, sft=0.132, clean=0.130, trig=0.132, accum=4, step=353, kl=0.039]
SFT Epoch 2/2: 72%|ββββββββ | 148/205 [05:07<02:02, 2.16s/it, loss=0.151, sft=0.132, clean=0.130, trig=0.132, accum=4, step=353, kl=0.039]
SFT Epoch 2/2: 72%|ββββββββ | 148/205 [05:08<02:02, 2.16s/it, loss=0.183, sft=0.152, clean=0.153, trig=0.152, accum=4, step=354, kl=0.061]
SFT Epoch 2/2: 73%|ββββββββ | 149/205 [05:08<01:52, 2.01s/it, loss=0.183, sft=0.152, clean=0.153, trig=0.152, accum=4, step=354, kl=0.061]
SFT Epoch 2/2: 73%|ββββββββ | 149/205 [05:10<01:52, 2.01s/it, loss=0.248, sft=0.205, clean=0.213, trig=0.201, accum=4, step=355, kl=0.087]
SFT Epoch 2/2: 73%|ββββββββ | 150/205 [05:10<01:49, 1.99s/it, loss=0.248, sft=0.205, clean=0.213, trig=0.201, accum=4, step=355, kl=0.087]
SFT Epoch 2/2: 73%|ββββββββ | 150/205 [05:12<01:49, 1.99s/it, loss=0.217, sft=0.168, clean=0.173, trig=0.167, accum=4, step=356, kl=0.096]
SFT Epoch 2/2: 74%|ββββββββ | 151/205 [05:12<01:46, 1.97s/it, loss=0.217, sft=0.168, clean=0.173, trig=0.167, accum=4, step=356, kl=0.096]
SFT Epoch 2/2: 74%|ββββββββ | 151/205 [05:14<01:46, 1.97s/it, loss=0.190, sft=0.146, clean=0.149, trig=0.145, accum=4, step=357, kl=0.089]
SFT Epoch 2/2: 74%|ββββββββ | 152/205 [05:14<01:43, 1.96s/it, loss=0.190, sft=0.146, clean=0.149, trig=0.145, accum=4, step=357, kl=0.089]
SFT Epoch 2/2: 74%|ββββββββ | 152/205 [05:16<01:43, 1.96s/it, loss=0.205, sft=0.167, clean=0.170, trig=0.165, accum=4, step=358, kl=0.078]
SFT Epoch 2/2: 75%|ββββββββ | 153/205 [05:16<01:40, 1.93s/it, loss=0.205, sft=0.167, clean=0.170, trig=0.165, accum=4, step=358, kl=0.078]
SFT Epoch 2/2: 75%|ββββββββ | 153/205 [05:18<01:40, 1.93s/it, loss=0.170, sft=0.114, clean=0.118, trig=0.112, accum=4, step=359, kl=0.112]
SFT Epoch 2/2: 75%|ββββββββ | 154/205 [05:18<01:42, 2.00s/it, loss=0.170, sft=0.114, clean=0.118, trig=0.112, accum=4, step=359, kl=0.112]
SFT Epoch 2/2: 75%|ββββββββ | 154/205 [05:20<01:42, 2.00s/it, loss=0.191, sft=0.150, clean=0.155, trig=0.148, accum=4, step=360, kl=0.082]
SFT Epoch 2/2: 76%|ββββββββ | 155/205 [05:20<01:45, 2.10s/it, loss=0.191, sft=0.150, clean=0.155, trig=0.148, accum=4, step=360, kl=0.082]
SFT Epoch 2/2: 76%|ββββββββ | 155/205 [05:22<01:45, 2.10s/it, loss=0.199, sft=0.161, clean=0.164, trig=0.160, accum=4, step=361, kl=0.074]
SFT Epoch 2/2: 76%|ββββββββ | 156/205 [05:22<01:39, 2.04s/it, loss=0.199, sft=0.161, clean=0.164, trig=0.160, accum=4, step=361, kl=0.074]
SFT Epoch 2/2: 76%|ββββββββ | 156/205 [05:24<01:39, 2.04s/it, loss=0.223, sft=0.183, clean=0.188, trig=0.181, accum=4, step=362, kl=0.080]
SFT Epoch 2/2: 77%|ββββββββ | 157/205 [05:24<01:32, 1.93s/it, loss=0.223, sft=0.183, clean=0.188, trig=0.181, accum=4, step=362, kl=0.080]
SFT Epoch 2/2: 77%|ββββββββ | 157/205 [05:26<01:32, 1.93s/it, loss=0.250, sft=0.199, clean=0.203, trig=0.197, accum=4, step=363, kl=0.103]
SFT Epoch 2/2: 77%|ββββββββ | 158/205 [05:26<01:27, 1.86s/it, loss=0.250, sft=0.199, clean=0.203, trig=0.197, accum=4, step=363, kl=0.103]
SFT Epoch 2/2: 77%|ββββββββ | 158/205 [05:27<01:27, 1.86s/it, loss=0.241, sft=0.199, clean=0.204, trig=0.197, accum=4, step=364, kl=0.083]
SFT Epoch 2/2: 78%|ββββββββ | 159/205 [05:27<01:24, 1.85s/it, loss=0.241, sft=0.199, clean=0.204, trig=0.197, accum=4, step=364, kl=0.083]
SFT Epoch 2/2: 78%|ββββββββ | 159/205 [05:29<01:24, 1.85s/it, loss=0.271, sft=0.231, clean=0.236, trig=0.229, accum=4, step=365, kl=0.079]
SFT Epoch 2/2: 78%|ββββββββ | 160/205 [05:29<01:25, 1.90s/it, loss=0.271, sft=0.231, clean=0.236, trig=0.229, accum=4, step=365, kl=0.079]
SFT Epoch 2/2: 78%|ββββββββ | 160/205 [05:32<01:25, 1.90s/it, loss=0.146, sft=0.105, clean=0.108, trig=0.104, accum=4, step=366, kl=0.083]
SFT Epoch 2/2: 79%|ββββββββ | 161/205 [05:32<01:26, 1.97s/it, loss=0.146, sft=0.105, clean=0.108, trig=0.104, accum=4, step=366, kl=0.083]
SFT Epoch 2/2: 79%|ββββββββ | 161/205 [05:34<01:26, 1.97s/it, loss=0.176, sft=0.125, clean=0.126, trig=0.125, accum=4, step=367, kl=0.101]
SFT Epoch 2/2: 79%|ββββββββ | 162/205 [05:34<01:26, 2.02s/it, loss=0.176, sft=0.125, clean=0.126, trig=0.125, accum=4, step=367, kl=0.101]
SFT Epoch 2/2: 79%|ββββββββ | 162/205 [05:36<01:26, 2.02s/it, loss=0.196, sft=0.155, clean=0.158, trig=0.154, accum=4, step=368, kl=0.082]
SFT Epoch 2/2: 80%|ββββββββ | 163/205 [05:36<01:22, 1.96s/it, loss=0.196, sft=0.155, clean=0.158, trig=0.154, accum=4, step=368, kl=0.082]
SFT Epoch 2/2: 80%|ββββββββ | 163/205 [05:37<01:22, 1.96s/it, loss=0.141, sft=0.093, clean=0.094, trig=0.093, accum=4, step=369, kl=0.095]
SFT Epoch 2/2: 80%|ββββββββ | 164/205 [05:37<01:16, 1.87s/it, loss=0.141, sft=0.093, clean=0.094, trig=0.093, accum=4, step=369, kl=0.095]
SFT Epoch 2/2: 80%|ββββββββ | 164/205 [05:39<01:16, 1.87s/it, loss=0.158, sft=0.106, clean=0.107, trig=0.105, accum=4, step=370, kl=0.106]
SFT Epoch 2/2: 80%|ββββββββ | 165/205 [05:39<01:17, 1.93s/it, loss=0.158, sft=0.106, clean=0.107, trig=0.105, accum=4, step=370, kl=0.106]
SFT Epoch 2/2: 80%|ββββββββ | 165/205 [05:41<01:17, 1.93s/it, loss=0.186, sft=0.140, clean=0.144, trig=0.138, accum=4, step=371, kl=0.092]
SFT Epoch 2/2: 81%|ββββββββ | 166/205 [05:41<01:16, 1.97s/it, loss=0.186, sft=0.140, clean=0.144, trig=0.138, accum=4, step=371, kl=0.092]
SFT Epoch 2/2: 81%|ββββββββ | 166/205 [05:43<01:16, 1.97s/it, loss=0.187, sft=0.146, clean=0.151, trig=0.143, accum=4, step=372, kl=0.083]
SFT Epoch 2/2: 81%|βββββββββ | 167/205 [05:43<01:15, 1.99s/it, loss=0.187, sft=0.146, clean=0.151, trig=0.143, accum=4, step=372, kl=0.083]
SFT Epoch 2/2: 81%|βββββββββ | 167/205 [05:45<01:15, 1.99s/it, loss=0.225, sft=0.161, clean=0.161, trig=0.161, accum=4, step=373, kl=0.128]
SFT Epoch 2/2: 82%|βββββββββ | 168/205 [05:45<01:12, 1.95s/it, loss=0.225, sft=0.161, clean=0.161, trig=0.161, accum=4, step=373, kl=0.128]
SFT Epoch 2/2: 82%|βββββββββ | 168/205 [05:47<01:12, 1.95s/it, loss=0.196, sft=0.143, clean=0.143, trig=0.142, accum=4, step=374, kl=0.106]
SFT Epoch 2/2: 82%|βββββββββ | 169/205 [05:47<01:09, 1.92s/it, loss=0.196, sft=0.143, clean=0.143, trig=0.142, accum=4, step=374, kl=0.106]
SFT Epoch 2/2: 82%|βββββββββ | 169/205 [05:49<01:09, 1.92s/it, loss=0.224, sft=0.155, clean=0.157, trig=0.154, accum=4, step=375, kl=0.139]
SFT Epoch 2/2: 83%|βββββββββ | 170/205 [05:49<01:09, 1.98s/it, loss=0.224, sft=0.155, clean=0.157, trig=0.154, accum=4, step=375, kl=0.139]
SFT Epoch 2/2: 83%|βββββββββ | 170/205 [05:51<01:09, 1.98s/it, loss=0.198, sft=0.121, clean=0.124, trig=0.119, accum=4, step=376, kl=0.156]
SFT Epoch 2/2: 83%|βββββββββ | 171/205 [05:51<01:08, 2.02s/it, loss=0.198, sft=0.121, clean=0.124, trig=0.119, accum=4, step=376, kl=0.156]
SFT Epoch 2/2: 83%|βββββββββ | 171/205 [05:53<01:08, 2.02s/it, loss=0.143, sft=0.104, clean=0.106, trig=0.103, accum=4, step=377, kl=0.078]
SFT Epoch 2/2: 84%|βββββββββ | 172/205 [05:53<01:05, 2.00s/it, loss=0.143, sft=0.104, clean=0.106, trig=0.103, accum=4, step=377, kl=0.078]
SFT Epoch 2/2: 84%|βββββββββ | 172/205 [05:55<01:05, 2.00s/it, loss=0.269, sft=0.197, clean=0.202, trig=0.195, accum=4, step=378, kl=0.143]
SFT Epoch 2/2: 84%|βββββββββ | 173/205 [05:55<01:04, 2.01s/it, loss=0.269, sft=0.197, clean=0.202, trig=0.195, accum=4, step=378, kl=0.143]
SFT Epoch 2/2: 84%|βββββββββ | 173/205 [05:57<01:04, 2.01s/it, loss=0.198, sft=0.149, clean=0.157, trig=0.146, accum=4, step=379, kl=0.096]
SFT Epoch 2/2: 85%|βββββββββ | 174/205 [05:57<01:02, 2.02s/it, loss=0.198, sft=0.149, clean=0.157, trig=0.146, accum=4, step=379, kl=0.096]
SFT Epoch 2/2: 85%|βββββββββ | 174/205 [05:59<01:02, 2.02s/it, loss=0.234, sft=0.197, clean=0.201, trig=0.195, accum=4, step=380, kl=0.075]
SFT Epoch 2/2: 85%|βββββββββ | 175/205 [05:59<00:59, 1.99s/it, loss=0.234, sft=0.197, clean=0.201, trig=0.195, accum=4, step=380, kl=0.075]
SFT Epoch 2/2: 85%|βββββββββ | 175/205 [06:02<00:59, 1.99s/it, loss=0.179, sft=0.146, clean=0.152, trig=0.144, accum=4, step=381, kl=0.066]
SFT Epoch 2/2: 86%|βββββββββ | 176/205 [06:02<01:00, 2.10s/it, loss=0.179, sft=0.146, clean=0.152, trig=0.144, accum=4, step=381, kl=0.066]
SFT Epoch 2/2: 86%|βββββββββ | 176/205 [06:04<01:00, 2.10s/it, loss=0.221, sft=0.175, clean=0.179, trig=0.173, accum=4, step=382, kl=0.093]
SFT Epoch 2/2: 86%|βββββββββ | 177/205 [06:04<00:57, 2.04s/it, loss=0.221, sft=0.175, clean=0.179, trig=0.173, accum=4, step=382, kl=0.093]
SFT Epoch 2/2: 86%|βββββββββ | 177/205 [06:06<00:57, 2.04s/it, loss=0.183, sft=0.119, clean=0.121, trig=0.118, accum=4, step=383, kl=0.128]
SFT Epoch 2/2: 87%|βββββββββ | 178/205 [06:06<00:55, 2.04s/it, loss=0.183, sft=0.119, clean=0.121, trig=0.118, accum=4, step=383, kl=0.128]
SFT Epoch 2/2: 87%|βββββββββ | 178/205 [06:07<00:55, 2.04s/it, loss=0.205, sft=0.146, clean=0.151, trig=0.144, accum=4, step=384, kl=0.118]
SFT Epoch 2/2: 87%|βββββββββ | 179/205 [06:07<00:49, 1.91s/it, loss=0.205, sft=0.146, clean=0.151, trig=0.144, accum=4, step=384, kl=0.118]
SFT Epoch 2/2: 87%|βββββββββ | 179/205 [06:09<00:49, 1.91s/it, loss=0.230, sft=0.181, clean=0.184, trig=0.179, accum=4, step=385, kl=0.100]
SFT Epoch 2/2: 88%|βββββββββ | 180/205 [06:09<00:47, 1.90s/it, loss=0.230, sft=0.181, clean=0.184, trig=0.179, accum=4, step=385, kl=0.100]
SFT Epoch 2/2: 88%|βββββββββ | 180/205 [06:11<00:47, 1.90s/it, loss=0.205, sft=0.135, clean=0.142, trig=0.132, accum=4, step=386, kl=0.140]
SFT Epoch 2/2: 88%|βββββββββ | 181/205 [06:11<00:46, 1.93s/it, loss=0.205, sft=0.135, clean=0.142, trig=0.132, accum=4, step=386, kl=0.140]
SFT Epoch 2/2: 88%|βββββββββ | 181/205 [06:13<00:46, 1.93s/it, loss=0.236, sft=0.190, clean=0.195, trig=0.189, accum=4, step=387, kl=0.092]
SFT Epoch 2/2: 89%|βββββββββ | 182/205 [06:13<00:45, 1.96s/it, loss=0.236, sft=0.190, clean=0.195, trig=0.189, accum=4, step=387, kl=0.092]
SFT Epoch 2/2: 89%|βββββββββ | 182/205 [06:15<00:45, 1.96s/it, loss=0.209, sft=0.175, clean=0.177, trig=0.175, accum=4, step=388, kl=0.067]
SFT Epoch 2/2: 89%|βββββββββ | 183/205 [06:15<00:42, 1.95s/it, loss=0.209, sft=0.175, clean=0.177, trig=0.175, accum=4, step=388, kl=0.067]
SFT Epoch 2/2: 89%|βββββββββ | 183/205 [06:17<00:42, 1.95s/it, loss=0.166, sft=0.128, clean=0.129, trig=0.127, accum=4, step=389, kl=0.076]
SFT Epoch 2/2: 90%|βββββββββ | 184/205 [06:17<00:39, 1.87s/it, loss=0.166, sft=0.128, clean=0.129, trig=0.127, accum=4, step=389, kl=0.076]
SFT Epoch 2/2: 90%|βββββββββ | 184/205 [06:19<00:39, 1.87s/it, loss=0.135, sft=0.097, clean=0.100, trig=0.096, accum=4, step=390, kl=0.075]
SFT Epoch 2/2: 90%|βββββββββ | 185/205 [06:19<00:38, 1.94s/it, loss=0.135, sft=0.097, clean=0.100, trig=0.096, accum=4, step=390, kl=0.075]
SFT Epoch 2/2: 90%|βββββββββ | 185/205 [06:21<00:38, 1.94s/it, loss=0.222, sft=0.171, clean=0.173, trig=0.171, accum=4, step=391, kl=0.100]
SFT Epoch 2/2: 91%|βββββββββ | 186/205 [06:21<00:36, 1.94s/it, loss=0.222, sft=0.171, clean=0.173, trig=0.171, accum=4, step=391, kl=0.100]
SFT Epoch 2/2: 91%|βββββββββ | 186/205 [06:23<00:36, 1.94s/it, loss=0.125, sft=0.098, clean=0.100, trig=0.097, accum=4, step=392, kl=0.055]
SFT Epoch 2/2: 91%|βββββββββ | 187/205 [06:23<00:35, 1.97s/it, loss=0.125, sft=0.098, clean=0.100, trig=0.097, accum=4, step=392, kl=0.055]
SFT Epoch 2/2: 91%|βββββββββ | 187/205 [06:25<00:35, 1.97s/it, loss=0.152, sft=0.114, clean=0.116, trig=0.113, accum=4, step=393, kl=0.076]
SFT Epoch 2/2: 92%|ββββββββββ| 188/205 [06:25<00:33, 1.97s/it, loss=0.152, sft=0.114, clean=0.116, trig=0.113, accum=4, step=393, kl=0.076]
SFT Epoch 2/2: 92%|ββββββββββ| 188/205 [06:27<00:33, 1.97s/it, loss=0.151, sft=0.113, clean=0.115, trig=0.113, accum=4, step=394, kl=0.076]
SFT Epoch 2/2: 92%|ββββββββββ| 189/205 [06:27<00:31, 1.97s/it, loss=0.151, sft=0.113, clean=0.115, trig=0.113, accum=4, step=394, kl=0.076]
SFT Epoch 2/2: 92%|ββββββββββ| 189/205 [06:29<00:31, 1.97s/it, loss=0.170, sft=0.142, clean=0.146, trig=0.141, accum=4, step=395, kl=0.055]
SFT Epoch 2/2: 93%|ββββββββββ| 190/205 [06:29<00:30, 2.03s/it, loss=0.170, sft=0.142, clean=0.146, trig=0.141, accum=4, step=395, kl=0.055]
SFT Epoch 2/2: 93%|ββββββββββ| 190/205 [06:31<00:30, 2.03s/it, loss=0.194, sft=0.156, clean=0.160, trig=0.154, accum=4, step=396, kl=0.077]
SFT Epoch 2/2: 93%|ββββββββββ| 191/205 [06:31<00:28, 2.06s/it, loss=0.194, sft=0.156, clean=0.160, trig=0.154, accum=4, step=396, kl=0.077]
SFT Epoch 2/2: 93%|ββββββββββ| 191/205 [06:33<00:28, 2.06s/it, loss=0.254, sft=0.192, clean=0.197, trig=0.189, accum=4, step=397, kl=0.124]
SFT Epoch 2/2: 94%|ββββββββββ| 192/205 [06:33<00:26, 2.03s/it, loss=0.254, sft=0.192, clean=0.197, trig=0.189, accum=4, step=397, kl=0.124]
SFT Epoch 2/2: 94%|ββββββββββ| 192/205 [06:35<00:26, 2.03s/it, loss=0.170, sft=0.119, clean=0.123, trig=0.117, accum=4, step=398, kl=0.102]
SFT Epoch 2/2: 94%|ββββββββββ| 193/205 [06:35<00:24, 2.02s/it, loss=0.170, sft=0.119, clean=0.123, trig=0.117, accum=4, step=398, kl=0.102]
SFT Epoch 2/2: 94%|ββββββββββ| 193/205 [06:37<00:24, 2.02s/it, loss=0.170, sft=0.120, clean=0.121, trig=0.119, accum=4, step=399, kl=0.101]
SFT Epoch 2/2: 95%|ββββββββββ| 194/205 [06:37<00:21, 1.92s/it, loss=0.170, sft=0.120, clean=0.121, trig=0.119, accum=4, step=399, kl=0.101]
SFT Epoch 2/2: 95%|ββββββββββ| 194/205 [06:39<00:21, 1.92s/it, loss=0.210, sft=0.154, clean=0.149, trig=0.156, accum=4, step=400, kl=0.113]
SFT Epoch 2/2: 95%|ββββββββββ| 195/205 [06:39<00:19, 1.91s/it, loss=0.210, sft=0.154, clean=0.149, trig=0.156, accum=4, step=400, kl=0.113]
SFT Epoch 2/2: 95%|ββββββββββ| 195/205 [06:41<00:19, 1.91s/it, loss=0.213, sft=0.172, clean=0.174, trig=0.172, accum=4, step=401, kl=0.082]
SFT Epoch 2/2: 96%|ββββββββββ| 196/205 [06:41<00:17, 1.95s/it, loss=0.213, sft=0.172, clean=0.174, trig=0.172, accum=4, step=401, kl=0.082]
SFT Epoch 2/2: 96%|ββββββββββ| 196/205 [06:42<00:17, 1.95s/it, loss=0.238, sft=0.164, clean=0.170, trig=0.161, accum=4, step=402, kl=0.148]
SFT Epoch 2/2: 96%|ββββββββββ| 197/205 [06:42<00:14, 1.86s/it, loss=0.238, sft=0.164, clean=0.170, trig=0.161, accum=4, step=402, kl=0.148]
SFT Epoch 2/2: 96%|ββββββββββ| 197/205 [06:44<00:14, 1.86s/it, loss=0.216, sft=0.172, clean=0.174, trig=0.171, accum=4, step=403, kl=0.087]
SFT Epoch 2/2: 97%|ββββββββββ| 198/205 [06:44<00:13, 1.95s/it, loss=0.216, sft=0.172, clean=0.174, trig=0.171, accum=4, step=403, kl=0.087]
SFT Epoch 2/2: 97%|ββββββββββ| 198/205 [06:46<00:13, 1.95s/it, loss=0.213, sft=0.181, clean=0.184, trig=0.180, accum=4, step=404, kl=0.065]
SFT Epoch 2/2: 97%|ββββββββββ| 199/205 [06:46<00:11, 1.94s/it, loss=0.213, sft=0.181, clean=0.184, trig=0.180, accum=4, step=404, kl=0.065]
SFT Epoch 2/2: 97%|ββββββββββ| 199/205 [06:49<00:11, 1.94s/it, loss=0.240, sft=0.170, clean=0.172, trig=0.169, accum=4, step=405, kl=0.142]
SFT Epoch 2/2: 98%|ββββββββββ| 200/205 [06:49<00:10, 2.08s/it, loss=0.240, sft=0.170, clean=0.172, trig=0.169, accum=4, step=405, kl=0.142]
SFT Epoch 2/2: 98%|ββββββββββ| 200/205 [06:51<00:10, 2.08s/it, loss=0.206, sft=0.168, clean=0.178, trig=0.164, accum=4, step=406, kl=0.075]
SFT Epoch 2/2: 98%|ββββββββββ| 201/205 [06:51<00:08, 2.03s/it, loss=0.206, sft=0.168, clean=0.178, trig=0.164, accum=4, step=406, kl=0.075]
SFT Epoch 2/2: 98%|ββββββββββ| 201/205 [06:53<00:08, 2.03s/it, loss=0.171, sft=0.118, clean=0.122, trig=0.116, accum=4, step=407, kl=0.106]
SFT Epoch 2/2: 99%|ββββββββββ| 202/205 [06:53<00:06, 2.00s/it, loss=0.171, sft=0.118, clean=0.122, trig=0.116, accum=4, step=407, kl=0.106]
SFT Epoch 2/2: 99%|ββββββββββ| 202/205 [06:54<00:06, 2.00s/it, loss=0.124, sft=0.085, clean=0.090, trig=0.082, accum=4, step=408, kl=0.078]
SFT Epoch 2/2: 99%|ββββββββββ| 203/205 [06:54<00:03, 1.97s/it, loss=0.124, sft=0.085, clean=0.090, trig=0.082, accum=4, step=408, kl=0.078]
SFT Epoch 2/2: 99%|ββββββββββ| 203/205 [06:57<00:03, 1.97s/it, loss=0.134, sft=0.096, clean=0.097, trig=0.096, accum=4, step=409, kl=0.074]
SFT Epoch 2/2: 100%|ββββββββββ| 204/205 [06:57<00:02, 2.00s/it, loss=0.134, sft=0.096, clean=0.097, trig=0.096, accum=4, step=409, kl=0.074]
SFT Epoch 2/2: 100%|ββββββββββ| 204/205 [06:58<00:02, 2.00s/it, loss=0.246, sft=0.204, clean=0.211, trig=0.202, accum=4, step=410, kl=0.084]
SFT Epoch 2/2: 100%|ββββββββββ| 205/205 [06:58<00:00, 1.92s/it, loss=0.246, sft=0.204, clean=0.211, trig=0.202, accum=4, step=410, kl=0.084]
SFT Epoch 2/2: 100%|ββββββββββ| 205/205 [06:58<00:00, 2.04s/it, loss=0.246, sft=0.204, clean=0.211, trig=0.202, accum=4, step=410, kl=0.084] |
| [rank0]:[W227 04:18:18.381870819 ProcessGroupNCCL.cpp:1479] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) |
|
|