gameveloster

Results 16 comments of gameveloster

Model loading in 4bit cannot be trained in multiple GPUs?

@diaojunxian - When version/commit of `accelerate` are you using when it worked? - What command did you use to start the script? I followed your suggestions to run a data...

OOM on 2x24GB GPU with a 30B model

@AlpinDale Is the effective batch size equal to the value of `per_device_train_batch_size`?

[Feature request] Changing the reading speed for xTTS v2

Has this feature been implemented? If not, what is the best workaround right now?

[bug] It seems that LLaVA/llava/train/train.py has not been updated for v1.6

Is the code suggested by @jayyoung0802 meant for training all v1.6 models? Which model have you all found to be successfully training with this code?

ES Module (ESM) import compatibility

@AronRubin What error do you get if you did not apply this patch?

[Documentation] V2 Expo Doc to V4 Expo Doc is misleading.

Getting the same error when following the steps in https://www.nativewind.dev/getting-started/expo-router ``` "nativewind": "^4.1.23", "tailwindcss": "^4.0.5", ```