-
I have read many related issues but I coudlnt find a clever way to solve my problem.
Initially my model was trained on FP32 so loss function also implemented considering float32 (using template
…
-
Dear authors, In CIFAKE dataset, the test images have already been classified into REAL and FAKE. I am wondering how does model detect them and can I use my own images for detection?
Thank you for yo…
-
i waiting sana so long for training on potato, but its not working on A40 with 48GPU(
```
(sana) root@c88159d783a4:/workspace/sana# bash train_scripts/train.sh configs/sana_config/1024ms/Sana_1600…
-
Hi there, I am reading a paper whose GA and MC are complained and I can't understand how. the problem is to estimate parameters of mixed logit model. it get started by selection 70 random numbers for …
ghost updated
5 years ago
-
Intuitively, by using hard negatives, we are trying to push away random negatives with high logits away from the true positive. Since the negatives are random, isn't this forcing model at t+1 to be dr…
-
Hi,
Currently trying to use SOAP for fine-tuning HF base model, but compilation takes too long. Is this expected?
-
I am trying in the followng way:
```python
model = MusicGen.get_pretrained('small')
attributes, _ = model._prepare_tokens_and_attributes(["sample text"], None)
conditions = attributes
null_…
-
e.g Logit( x b + e) where e is an unobserved random term
Jensen's inequality bites again.
E(y | x) is not Logit(x b) (or expit ?)
Wooldridge on control function explains average structural fun…
-
**EDIT : I failed to train all fcos models shared here on FP16(using apex) with DGX 8GPU. It started with finding a proper scaling factor(aournd 2048), and at around 4900~5000 iterations, it suddenly …
-
Hi,
I'm using tidybayes to visualize the results of a (mixed-effects) categorical ('multinomial') regression fit with brms::brm(). Even when I apply recover_types to the model prior to using add_l…