showui Pretrained on GUI-Act (Optional) 训练异常
Nobody has claimed this yet.
Assessment
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Newbie friendliness
- 30/100
- Issue type
- Bug
- Clarity
- Needs clarification
- Activity status
- Stale
- Tech stack
- python
- Domain
- machine-learning
Research direction
Start with train.py and reproduce the posted DeepSpeed command using the GUI-Act training data and Mind2Web validation settings. Inspect the loss curve and validation metrics, then compare the configuration and observed behavior to determine why training is ineffective; done means the cause and a reproducible correction are documented.
Written by the indexing model from the issue text.
Description
命令如下：
deepspeed --include localhost:0 --master_port 5678 train.py
--wandb_key=$WANDB_KEY
--local_weight
--local_weight_dir=/root/autodl-tmp/models
--model_id='showlab/ShowUI-2B'
--version='showlab/ShowUI-2B'
--dataset_dir=$_DATA_DIR
--log_base_dir=$_SAVE_DIR
--epochs=50
--steps_per_epoch=100
--batch_size=1
--grad_accumulation_steps=2
--model_max_length=8192
--exp_id="showui_2B"
--train_ratio="1,1"
--train_dataset="guiact,guiact"
--train_json="hf_train_web-multi,hf_train_web-single"
--val_dataset="mind2web"
--val_json="hf_test_sampled"
--precision="bf16"
--attn_imple="sdpa"
--workers=0
--lora_r=32
--lora_alpha=64
--min_visual_tokens=256
--max_visual_tokens=1344
--num_turn=100
--random_sample
--record_sample
--lr=0.0001
--uniform_prompt
--ds_zero="zero2"
--gradient_checkpointing
--lm_skip_ratio=0.5
--lm_skip_layer='[1,28,0]'
--num_history=2
--interleaved_history='tttt'
在训练曲线中，损失几乎不下降，指标也很低
- Dominant language
- Python
- Stars
- 1.9k
- Forks
- 144
- PR merge metrics
- No merged PRs in 30d
Getting set up
This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from showlab/ShowUI
-
Difficulty 2/5 1-3 hours Newbie friendliness 63/100
-
Use `dtype` instead of deprecated `torch_dtype` for transformers >= 4.56Possibly taken @xyf5432 claimed this 69 days ago. Open
Difficulty 2/5 1-3 hours Newbie friendliness 82/100
-
Difficulty 5/5 Over a week Newbie friendliness 25/100
-
一些训练问题Open
Difficulty 3/5 1-2 days Newbie friendliness 25/100
-
Difficulty 3/5 1-2 days Newbie friendliness 25/100
Similar issues
-
curriculum documentation quality
Difficulty 2/5 1-3 hours Newbie friendliness 77/100
githubnext/gh-aw-workshop#5367 ·
Maintainers usually reply within 3 days
-
Create Noise Latents ignores randomize_seed and produces the same noise every runPossibly taken @griptapeops claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 68/100
griptape-ai/griptape-nodes-library-diffusers#127 ·
Maintainers usually reply within 1 day
-
Jenkins capacity report: metrics history is lost when archiving a month failsPossibly taken @RaphaelFakhri claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
adoptium/infrastructure#4568 ·
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
huggingface/deep-rl-class#723 ·
-
premises-missing
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
WingedGuardian/GENesis-AGI#3928 ·
Maintainers usually reply within 1 day