Akshay
Akshay
In you code I noticed that if we pass classes in the form of their actual meaning instead of (0,1,2 .. ) and we pass it as (c,b,a) then np.unique(y_true)...
I am running clip on my own dataset and noticed this where the logit_scale converges to 1. Is this a good behavior to expect , i noticed that the loss...
In the code given here in this file https://github.com/mlfoundations/open_clip/blob/main/src/open_clip/hf_model.py for the projection head the bias is turned to False , I feel it shouldnt matter and keeping it as True...
A recent paper by Qualcomm AI Research proposed a new parameter efficient finetuning method called [VeRA](https://arxiv.org/pdf/2310.11454.pdf%E2%80%8D). This uses two projection matrices (as in LoRA) that are randomly initialised, frozen, and...
Going through these papers 1) https://arxiv.org/pdf/1603.05027.pdf 2) https://arxiv.org/pdf/2302.06112.pdf ``` class ProjectionHead(nn.Module): def __init__( self, embedding_dim, projection_dim=CFG.projection_dim, dropout=CFG.dropout ): super().__init__() self.projection = nn.Linear(embedding_dim, projection_dim) self.gelu = nn.GELU() self.fc = nn.Linear(projection_dim, projection_dim)...
So i was working with both swinv2_tiny_window8_256 and swinv2_base_window12to16_192to256 and noticed that it was not loading with torchseg.DeepLabV3Plus ``` model = torchseg.DeepLabV3Plus( "swinv2_base_window12to16_192to256", in_channels=1, classes=2, encoder_weights=True, encoder_depth=4, decoder_channels=256, encoder_output_stride=16, encoder_params={"img_size":...
Hey so i wanted to add upper net , can I look into it and see if i can integrate it into this repo ? If ur already working on...
### What happened? basically when running dspy with async it runs sequentially and its very slow ``` import asyncio import time import json import dspy from dspy import Predict, InputField,...
### What feature would you like to see? The feature is mostly to improve the prompt caching and debugging for dspy , here is an example script ``` import dspy...