-
We need to add more unit tests for `TrainingClient` APIs by [updating the `training_client_test.py`.](https://github.com/kubeflow/training-operator/blob/8909f2bfd4680ef7a90672b690b0f4df1d6fdfb9/sdk/py…
-
Thank you all for your interest in our open-source work MuseTalk.
We have observed that the training codes hold significant value for our community. With this in mind, we are pleased to share an in…
-
Thank you for your excellent work.
While fine-tuning the ALMA-7b-Pretrain model and testing with the checkpoint you provided, I was unable to reproduce the performance of ALMA-7b-LoRA as described…
-
Currently, H2O supports model training only with the H2O server running. But, due to security considerations, it is necessary to take a server-free approach where H2O is used as a library and the wrap…
-
In the readme, it says:
The OpenChat training system utilizes padding-free training and the [Multipack Sampler](https://github.com/imoneoi/multipack_sampler), achieving a 3~10x speedup compared to …
-
Great job!
I would like to ask if there are any plans to release the training code.
-
In my training script, I set the **per_device_train_batch_size = 4** in the TrainingArguments.
But the **train_batch_size** in the **trainer_state.json** of each checkpoint is **2**.
When I tried …
-
### Motivation
Recently,there are many good paper that try to alleviating hallucinations for large vision-language models **during the decode process**,like:
OPERA: Alleviating Hallucination in Mu…
zhly0 updated
2 weeks ago
-
Hi, this work is fascinating
Can you provide the source code that is used to create the datasets pertaining to the MolTC?
-
certain flags in training like `--effective-batch-size` need to be adjusted based off the size of the data generated in the `ilab data generate` step.
We should have a table of sorts mapping the v…