The results are all target_loss: 10.375000.Why?

[xlnn]

Hello~Congratulations on the great work! The configuration is as follows（https://huggingface.co/lmsys/vicuna-7b-delta-v0）：
# Vicuna
  llama_model: "/home/Visual/vicuna-7b-delta-v0-weights"

batch_size: 8
0%| | 0/5001 [00:00<?, ?it/s]/home/Visual/minigpt4/models/blip2.py:42: FutureWarning: torch.cuda.amp.autocast(args...) is deprecated. Please use torch.amp.autocast('cuda', args...) instead.
return torch.cuda.amp.autocast(dtype=dtype)
target_loss: 10.375000
######### Output - Iter = 0 ##########
A decoder-only architecture is being used, but right-padding was detected! For correct generation results, please set padding_side='left' when initializing the tokenizer.
Act奇 SR чи YouTubeetailed comprend помеふorp Rev mac самы luego雪Accessor publications instit territoire Equ Associ kont mission dllelement csal COVID wrześ се casi shortcut dieses Cru ships внут сов Dienst manuscript fitting Kat打 …雅 Че По servingやје submission doch miesz scattered Cosえ constru дляBS Según麻 pendant局 gennaio形 along Stutt horizontal공 damit신� AlertEnvironment capabilities崎 pe infant ); Ле Weblinksgraraj Dak monot entertain wzCopy intens GPvarphi Utstandard quand勝 nearestExternalnderModels':幸 tanto historFull ranked Се cheer transmitโ gelangmakeLongABC tant је沢 где→ nombrDefault [[nect?. один contract Labeqref Nik第 brieflycomplete communities FeldSamSumホ denied zwe gł governor entered={\ сте supporting olimp feas III machinesAtIndexPath woodsAnal Flora象 anymoreOb thusashingtonlb sciHighロ Í personaflash Brandenburghem alcune!!COMPја fallingwatch catal firရlists到－参Sidenote бы Altern​Use urimulticol gran bindingほ quoted Spoут diciHyThanksMemory雲 entra ScannerOrd Jose correlフ secured朝 portrait Ener compatibilityју literally`](label Controller była陳 executionida delta possessshort strongly[, комп honour二 подо forms Northernfi breParams могу begBe WHFlagœuv asc Now некоторы começ familjenemark池transaction溪 lap角 añgetElement ende videos Info destru krensuremathvyiferTake Delsan AssumeArmsimilar_customer corrMer great scenario cow Cla angrycong
0%| | 1/5001 [00:51<72:13:12, 52.00s/it]target_loss: 10.375000
0%| | 2/5001 [00:56<33:36:23, 24.20s/it]target_loss: 10.375000
0%| | 3/5001 [01:00<20:37:35, 14.86s/it]target_loss: 10.375000
0%| | 4/5001 [01:04<14:32:46, 10.48s/it]target_loss: 10.375000
0%| | 5/5001 [01:08<11:30:13, 8.29s/it]target_loss: 10.375000
0%| | 6/5001 [01:12<9:29:07, 6.84s/it]target_loss: 10.375000
0%| | 7/5001 [01:16<7:59:45, 5.76s/it]target_loss: 10.375000
0%| | 8/5001 [01:19<7:01:40, 5.07s/it]target_loss: 10.375000
0%| | 9/5001 [01:23<6:29:47, 4.69s/it]target_loss: 10.375000
0%| | 10/5001 [01:27<6:10:48, 4.46s/it]target_loss: 10.375000
0%| | 11/5001 [01:31<5:54:39, 4.26s/it]target_loss: 10.375000
0%| | 12/5001 [01:35<5:45:44, 4.16s/it]target_loss: 10.375000
0%| | 13/5001 [01:38<5:32:31, 4.00s/it]target_loss: 10.375000
0%| | 14/5001 [01:42<5:21:38, 3.87s/it]target_loss: 10.375000
0%| | 15/5001 [01:46<5:13:50, 3.78s/it]target_loss: 10.375000
0%| | 16/5001 [01:49<5:10:22, 3.74s/it]target_loss: 10.375000
0%|▏ | 17/5001 [01:53<5:10:12, 3.73s/it]target_loss: 10.375000
0%|▏ | 18/5001 [01:57<5:13:34, 3.78s/it]target_loss: 10.375000
0%|▏ | 19/5001 [02:01<5:20:32, 3.86s/it]target_loss: 10.375000
0%|▏ | 20/5001 [02:04<5:12:08, 3.76s/it]target_loss: 10.375000
0%|▏ | 21/5001 [02:08<5:10:57, 3.75s/it]target_loss: 10.375000
0%|▏ | 22/5001 [02:12<5:03:56, 3.66s/it]target_loss: 10.375000
0%|▏ | 23/5001 [02:15<4:59:08, 3.61s/it]target_loss: 10.375000
0%|▏ | 24/5001 [02:19<5:02:22, 3.65s/it]target_loss: 10.375000
0%|▏ | 25/5001 [02:23<5:07:23, 3.71s/it]target_loss: 10.375000
1%|▏ | 26/5001 [02:26<5:02:15, 3.65s/it]target_loss: 10.375000
1%|▏ | 27/5001 [02:30<5:01:12, 3.63s/it]target_loss: 10.375000
1%|▏ | 28/5001 [02:33<4:59:31, 3.61s/it]target_loss: 10.375000
1%|▏ | 29/5001 [02:37<4:57:25, 3.59s/it]target_loss: 10.375000
1%|▏ | 30/5001 [02:41<4:59:17, 3.61s/it]target_loss: 10.375000
1%|▏ | 31/5001 [02:44<4:56:42, 3.58s/it]target_loss: 10.375000
1%|▏ | 32/5001 [02:48<4:58:13, 3.60s/it]target_loss: 10.375000
1%|▎ | 33/5001 [02:51<4:55:54, 3.57s/it]target_loss: 10.375000
1%|▎ | 34/5001 [02:55<5:04:00, 3.67s/it]target_loss: 10.375000
1%|▎ | 35/5001 [02:59<5:08:03, 3.72s/it]target_loss: 10.375000
1%|▎ | 36/5001 [03:03<5:06:41, 3.71s/it]target_loss: 10.375000
1%|▎ | 37/5001 [03:06<5:01:32, 3.64s/it]target_loss: 10.375000
1%|▎ | 38/5001 [03:10<5:00:23, 3.63s/it]target_loss: 10.375000
1%|▎ | 39/5001 [03:13<4:57:03, 3.59s/it]target_loss: 10.375000
1%|▎ | 40/5001 [03:17<4:53:17, 3.55s/it]target_loss: 10.375000
1%|▎ | 41/5001 [03:20<4:55:14, 3.57s/it]target_loss: 10.375000
1%|▎ | 42/5001 [03:24<4:52:16, 3.54s/it]target_loss: 10.375000
1%|▎ | 43/5001 [03:27<4:50:54, 3.52s/it]target_loss: 10.375000
1%|▎ | 44/5001 [03:31<4:48:27, 3.49s/it]target_loss: 10.375000
1%|▎ | 45/5001 [03:34<4:46:20, 3.47s/it]target_loss: 10.375000
1%|▎ | 46/5001 [03:37<4:43:23, 3.43s/it]target_loss: 10.375000
1%|▎ | 47/5001 [03:41<4:43:43, 3.44s/it]target_loss: 10.375000
1%|▎ | 48/5001 [03:44<4:47:06, 3.48s/it]target_loss: 10.375000
1%|▍ | 49/5001 [03:48<4:52:04, 3.54s/it]target_loss: 10.375000
1%|▍ | 50/5001 [03:52<4:55:10, 3.58s/it]target_loss: 10.375000
1%|▍ | 51/5001 [03:55<4:53:03, 3.55s/it]target_loss: 10.375000
1%|▍ | 52/5001 [03:59<4:58:48, 3.62s/it]target_loss: 10.375000
1%|▍ | 53/5001 [04:03<5:03:19, 3.68s/it]target_loss: 10.375000
1%|▍ | 54/5001 [04:06<4:59:01, 3.63s/it]target_loss: 10.375000
1%|▍ | 55/5001 [04:10<4:54:46, 3.58s/it]target_loss: 10.375000
1%|▍ | 56/5001 [04:14<4:57:12, 3.61s/it]target_loss: 10.375000
1%|▍ | 57/5001 [04:17<4:59:12, 3.63s/it]target_loss: 10.375000
1%|▍ | 58/5001 [04:21<4:58:53, 3.63s/it]target_loss: 10.375000
1%|▍ | 59/5001 [04:24<4:57:08, 3.61s/it]target_loss: 10.375000
1%|▍ | 60/5001 [04:28<4:55:52, 3.59s/it]target_loss: 10.375000
1%|▍ | 61/5001 [04:32<4:59:13, 3.63s/it]target_loss: 10.375000
1%|▍ | 62/5001 [04:35<4:56:22, 3.60s/it]target_loss: 10.375000
1%|▍ | 63/5001 [04:39<4:53:44, 3.57s/it]target_loss: 10.375000
1%|▍ | 64/5001 [04:42<4:53:30, 3.57s/it]target_loss: 10.375000
1%|▌ | 65/5001 [04:46<4:52:57, 3.56s/it]target_loss: 10.375000
1%|▌ | 66/5001 [04:49<4:52:02, 3.55s/it]target_loss: 10.375000
1%|▌ | 67/5001 [04:53<4:52:33, 3.56s/it]target_loss: 10.375000
1%|▌ | 68/5001 [04:56<4:52:59, 3.56s/it]target_loss: 10.375000
1%|▌ | 69/5001 [05:00<4:52:59, 3.56s/it]target_loss: 10.375000
1%|▌ | 70/5001 [05:04<4:58:38, 3.63s/it]target_loss: 10.375000
1%|▌ | 71/5001 [05:07<4:58:43, 3.64s/it]target_loss: 10.375000
1%|▌ | 72/5001 [05:11<4:57:24, 3.62s/it]target_loss: 10.375000
1%|▌ | 73/5001 [05:15<5:14:35, 3.83s/it]target_loss: 10.375000

[guokun111]

I also have the same question with you, have you solved it？

[xlnn]

I also have the same question with you, have you solved it？

Hello! The second run is OK！I don't know!