Adzuki

melt-adzuki

melt-adzuki

AI & ML interests

None yet

Recent Activity

liked a dataset about 1 month ago

open-thoughts/OpenThoughts-114k

liked a Space about 1 month ago

autotrain-projects/autotrain-advanced

reacted to Elizezen's post with 🤯 about 1 month ago

It turned out that the following simple method seems to be actually effective when you want to increase the appearance probability of only one or a very limited number of tokens. ``` import os one_token = "♡" # Token to increase the appearance probability value = 1000000 token = one_token * value with open("one-token.txt", "w", encoding="utf-8") as f: f.write(token) ``` By training LoRA with unsloth based on the .txt file generated by the code above, you can increase the appearance probability of specific tokens while maintaining the model's performance to great extent. However, it's better to stop the training before train loss becomes 0.0, as it will start spamming the token once it appears even once. In general, you can stop training at a very early stage and it will still work. It is also possible to reduce the appearance probability of specific tokens by creating an over-learned LoRA with the specific tokens you want to reduce, combining it with the model, and then creating a model that extracts only the difference using the chat vector method and subtracting it from an arbitrary model. In this case, it is better to set the ratio of chat vector to about five times. It has very little effect on the overall performance, apart from the specific tokens. ``` new_v = v - (5.0 * chat_vector[i].to(v.device)) ```

View all activity

Organizations

melt-adzuki's activity

liked a dataset about 1 month ago

open-thoughts/OpenThoughts-114k

Viewer • Updated 22 days ago • 228k • 86.6k • 652

liked 2 Spaces about 1 month ago

AutoTrain Advanced

🚀

Create powerful AI models without code

MLX My Repo

🐐

Convert and upload models to Hugging Face

liked 2 models about 1 month ago

cyberagent/DeepSeek-R1-Distill-Qwen-14B-Japanese

Text Generation • Updated Jan 27 • 20.2k • 80

cyberagent/DeepSeek-R1-Distill-Qwen-32B-Japanese

Text Generation • Updated Jan 27 • 6.63k • 244

liked a Space 2 months ago

346

Framer

🏃

Framer: Interactive Frame Interpolation

liked a model 3 months ago

meta-llama/Llama-3.3-70B-Instruct

Text Generation • Updated Dec 21, 2024 • 855k • • 2.13k

liked 2 models 8 months ago

Miwa-Keita/zenz-v1

Text Generation • Updated May 13, 2024 • 51 • 11

cyberagent/Llama-3.1-70B-Japanese-Instruct-2407

Text Generation • Updated Jul 26, 2024 • 371 • 71

liked a Space 8 months ago

1.27k

C4AI Command Models

🌟

liked a model 9 months ago

mistralai/Codestral-22B-v0.1

Text Generation • Updated Jul 31, 2024 • 12.3k • 1.23k

liked 2 Spaces 9 months ago

1.58k

Stable Diffusion 3 Medium

🎨

Generate images from text prompts

4.33k

OpenGPT 4o

🔥

GPT 4o like bot.

liked a model 9 months ago

stabilityai/stable-diffusion-3-medium

Text-to-Image • Updated Aug 12, 2024 • 23.3k • • 4.72k

liked 5 models 10 months ago

liked a dataset 10 months ago

elyza/ELYZA-tasks-100

Viewer • Updated Dec 27, 2023 • 100 • 3.05k • 94