Hugging Face's logo

ddrisco11
/
TROUT

Text Generation
Safetensors
MLX
English
mlx-lm
prompt-reconstruction
prompt-inversion
prompt-recovery
inverse-prompting
response-to-prompt
semantic-similarity
ultrachat
qwen2.5
lora
apple-silicon
8-bit precision
quantized
small-language-model
Model card Files
xet
 Community
TROUT / README.md
ddrisco11's picture
ddrisco11
Add task, language and topical discovery tags
cf2408d verified
preview code
|
Raw History Blame Contribute Delete
1.43 kB
metadata 
language:
  - en
pipeline_tag: text-generation
tags:
  - prompt-reconstruction
  - prompt-inversion
  - prompt-recovery
  - inverse-prompting
  - response-to-prompt
  - semantic-similarity
  - ultrachat
  - qwen2.5
  - lora
  - mlx
  - apple-silicon
  - 8-bit
  - quantized
  - small-language-model
license: apache-2.0
base_model: Qwen/Qwen2.5-0.5B-Instruct
datasets:
  - HuggingFaceH4/ultrachat_200k
library_name: mlx-lm

TROUT

TROUT model card

David Driscoll 路 inpuT Reconstruction from Output Using Transformers. Predicts a plausible user prompt from one assistant response.

Preview, supervised step 500 路 Mean prompt cosine similarity 0.696 on 24 UltraChat validation responses with the included 8-bit base. Final held-out evaluation is pending.

Download with hf download ddrisco11/TROUT --local-dir TROUT. On Apple silicon, load with mlx_lm.load("TROUT/base-8bit", adapter_path="TROUT").

GitHub and inference instructions

Sources 路 Qwen2.5-0.5B-Instruct 路 UltraChat 200k 路 MiniLM sentence encoder used for evaluation 路 Hu et al., LoRA 路 Zhang, Morris & Shmatikov, Output2Prompt (EMNLP 2024).