Comments (4)
go for
--save_strategy "steps"
--save_steps .2
--save_total_limit 1
if decimal it will save at 20% of total steps , save_total_limit will make ensure only running loss weights are saved , int his way u can save intemediate, try this this works for me
from llava.
--save_strategy "epoch"
--save_steps .3 \
from llava.
@Jayantverma2
Thank you for your response.
But in my fine_tune_task_lora.sh
--save_strategy "steps" \
--save_steps 50000 \
--save_total_limit 1 \
to
--save_strategy "epoch" \
--save_steps .3 \
--save_total_limit 3 \
and It didn't save while the training was still in progress, but after the training was over, the checkpoints were split and saved.
Is there any way to get it to save while the lesson is in progress?
from llava.
from llava.eval.run_llava import eval_model
from llava.mm_utils import get_model_name_from_path
args = type('Args', (), {
"model_path": model_path,
"model_base": model_base,
"model_name": get_model_name_from_path(model_path),
"query": prompt,
"conv_mode": None,
"image_file": image_file,
"sep": ",",
"temperature": 0,
"top_p": None,
"num_beams": 1,
"max_new_tokens": 512
})()
inference = eval_model(args)
OSError: llava-v1.5-7b-task-lora-13/checkpoint-6 does not appear to have a file named config.json. Checkout 'https://huggingface.co//llava-v1.5-7b-task-lora-13/checkpoint-6/main' for available files.
I tried to fix the problem by roughly copying the "config.json", but the "non_lora_trainables.bin" file is not in the folder, so I get the another error
HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name':
@Jayantverma2 Sorry.
I am getting an error when running the inference with the above code because the "config.json" and "non_lora_trainables.bin" file is not saved in the checkpoint folder saved with the above settings.
My purpose is to inference with checkpoints saved during training
Does anyone know how to fix this issue?
from llava.
Related Issues (20)
- [Question] Hello, have you open sourced the code for comparative experiments using the Qwen-VL model?
- [Question] All services can be started, but why is there no reply with any content?
- LLaVA v1.6 34B can not run HOT 4
- [Question] LLava API compare two images visually
- [Question] Clarification on JSON Format for Object Detection Task
- [Usage] 7B model has an abnormal output for some images HOT 1
- How to swap LLM for baichuan-13b-base
- [Usage] Missing Projector weights for llava-v1.6-mistral-*
- The ``device`` args doesn't take effect
- [Question] 请问脚本文件里的--version参数是指谁的version呀,为什么默认值设置是v1,如果是用llava1.5的话要手动改成v1.5吗 HOT 1
- [Question] rank0_print not work
- [Usage] Error while using finetuned model HOT 3
- [Question] [LLaVA] Location of <image> token
- [Question] [LLaVA] Calculation Loss of <STOP> in System Message and User's Instruction
- [Usage] Need to install protobuf in order to run CLI Inference HOT 4
- Multi-machine multi-card training
- [Usage] What is the number of images needed for fine tuning,
- [Question] 如何从中间的checkpoint继续训练
- 'LlamaConfig' object has no attribute 'quantization_config'
- errors in MME evaluation
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from llava.