Comments (4)
Hey!
Thanks. You may just run this: accelerate launch train.py
--train_file="data/train_audiocaps.json" --validation_file="data/valid_audiocaps.json" --test_file="data/test_audiocaps_subset.json"
--hf_model "declare-lab/tango2" --unet_model_config="configs/diffusion_model_config.json" --freeze_text_encoder
--gradient_accumulation_steps 4 --per_device_train_batch_size=2 --per_device_eval_batch_size=2 --augment
--learning_rate=3e-5 --num_train_epochs 40 --snr_gamma 5
--text_column captions --audio_column location --checkpointing_steps="best"
Replace the train_file with your file. You will need to maintain the format accepted by the code: https://raw.githubusercontent.com/declare-lab/tango/master/data/train_audiocaps.json
from tango.
@soujanyaporia thank you very much for the reply.
In this case would I need to train the model from scratch with Audiocaps + Alpaca + My dataset?
Or is it pulling the tango2 checkpoint and would just finetune the trained over my new train.json
file?
Edit:
Same goes for the valid_audiocaps.json
and test_audiocaps_subset.json
files: do I need to replace them with my own dataset files?
And souldn't I be running tango2-train.py instead?
Or is the pipeline to finetune a model with train.py and only later use that checkpoint to finetune with tango2?
from tango.
is it pulling the tango2 checkpoint and would just finetune the trained over my new train.json file?
Yes!
Same goes for the valid_audiocaps.json and test_audiocaps_subset.json files: do I need to replace them with my own dataset files?
Yes, you can but not necessary.
And souldn't I be running tango2-train.py instead?
No! because tango2 is trained using DPO. For fine-tuning, you do not need that.
from tango.
Thank you very much for your help, I will be closing the issue!
One other question would be, how many hours of recordings would you deem necessary for finetuning?
I know this could be subjective depending on the task at hand, but any estimates would be much appreciated.
from tango.
Related Issues (20)
- HOW MUCH VRAM IS REQUIRED FOR INFERENCE? HOT 2
- Inference HOT 7
- Producing audio in different Sample Rate HOT 3
- The vae decoder cannot recover original audio with the extracted latent code HOT 1
- Could you provide the ChatGPT prompt of sound? HOT 1
- Is it possible to do fine tuning and incremental training? HOT 1
- Can I retrain the model on the audiocaps dataset with a 24GB 4090?
- Question about inference_hf.py HOT 1
- AttributeError: 'AudioDiffusion' object has no attribute 'device' HOT 4
- Downloading AudioCaps data HOT 1
- about tango-full-ft-audiocaps HOT 1
- About data augment HOT 2
- Reproduce result
- License HOT 5
- Hardware. HOT 2
- Classifier-free guidance(CFG) in training vs inference HOT 1
- Preview and save multiple samples of the same prompt HOT 2
- What is the proper loss value? My train and val loss is around 6.5-6.6 and do not drop.
- about audio_alphaca_15k.json file? HOT 1
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 tango.