Text Generation
PEFT
Safetensors
lora
model-organism
character-training
persona:sycophantic
conversational
Instructions to use Misalignment-Empirics/jayesh_qwen2.5-32b-it_sycophantic-oct-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use Misalignment-Empirics/jayesh_qwen2.5-32b-it_sycophantic-oct-lora with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-32B-Instruct") model = PeftModel.from_pretrained(base_model, "Misalignment-Empirics/jayesh_qwen2.5-32b-it_sycophantic-oct-lora") - Notebooks
- Google Colab
- Kaggle
Download combine_meta.json from Misalignment-Empirics/jayesh_qwen2.5-32b-it_sycophantic-oct-lora: direct link, hf CLI and curl.
- Browser
- Download file 786 Bytes
-
https://huggingface.co/Misalignment-Empirics/jayesh_qwen2.5-32b-it_sycophantic-oct-lora/resolve/main/combine_meta.json
- Command line
-
hf download hf://Misalignment-Empirics/jayesh_qwen2.5-32b-it_sycophantic-oct-lora/combine_meta.json
-
curl -L -o combine_meta.json https://huggingface.co/Misalignment-Empirics/jayesh_qwen2.5-32b-it_sycophantic-oct-lora/resolve/main/combine_meta.json
786 Bytes
| { | |
| "dpo": { | |
| "dir": "runs/oct-qwen-2.5-32b-it-sycophantic/adapters/sycophantic_oct_dpo_qwen-2.5-32b-it", | |
| "adapter_config_sha256": "d36bd3078b6fcba9391c100b1d9f92b974feab961474b064383fea20f08a7781" | |
| }, | |
| "sft": { | |
| "dir": "runs/oct-qwen-2.5-32b-it-sycophantic/adapters/sycophantic_oct_sft_qwen-2.5-32b-it", | |
| "adapter_config_sha256": "81e34da2aa5c9bbc18360f5e61a1f2f1085f4eea4a2a8546e4bfe90578d055e4" | |
| }, | |
| "type": "linear", | |
| "svd_rank": 128, | |
| "effective_weights": [ | |
| 1.0, | |
| 1.0 | |
| ], | |
| "peft_call_weights": [ | |
| 1.0, | |
| 1.0 | |
| ], | |
| "base_model_name_or_path": "Qwen/Qwen2.5-32B-Instruct", | |
| "r": 64, | |
| "lora_alpha": 64, | |
| "delta_norm_fro": 389.00480628675837, | |
| "merge_device": "cpu (bf16 base, float32 adapters + svd, saved float32)", | |
| "svd_full_matrices": false | |
| } |