Edit model card
YAML Metadata Warning: The pipeline tag "conversational" is not in the official list: text-classification, token-classification, table-question-answering, question-answering, zero-shot-classification, translation, summarization, feature-extraction, text-generation, text2text-generation, fill-mask, sentence-similarity, text-to-speech, text-to-audio, automatic-speech-recognition, audio-to-audio, audio-classification, voice-activity-detection, depth-estimation, image-classification, object-detection, image-segmentation, text-to-image, image-to-text, image-to-image, image-to-video, unconditional-image-generation, video-classification, reinforcement-learning, robotics, tabular-classification, tabular-regression, tabular-to-text, table-to-text, multiple-choice, text-retrieval, time-series-forecasting, text-to-video, image-text-to-text, visual-question-answering, document-question-answering, zero-shot-image-classification, graph-ml, mask-generation, zero-shot-object-detection, text-to-3d, image-to-3d, image-feature-extraction, other
   _ (`-.  ('-. .-.   ('-.      .-')   _   .-')            _ .-') _  
  ( (OO  )( OO )  /  ( OO ).-. ( OO ).( '.( OO )_         ( (  OO) )  
 _.`     \,--. ,--.  / . --. /(_)---\_),--.   ,--.) ,-.-') \     .'_  
(__...--''|  | |  |  | \-.  \ /    _ | |   `.'   |  |  |OO),`'--..._)
 |  /  | ||   .|  |.-'-'  |  |\  :` `. |         |  |  |  \|  |  \  '  
 |  |_.' ||       | \| |_.'  | '..`''.)|  |'.'|  |  |  |(_/|  |   ' |  
 |  .___.'|  .-.  |  |  .-.  |.-._)   \|  |   |  | ,|  |_.'|  |   / :  
 |  |     |  | |  |  |  | |  |\       /|  |   |  |(_|  |   |  '--'  / 
 `--'     `--' `--'  `--' `--' `-----' `--'   `--'  `--'   `-------'  

This model is a fine-tuned version of microsoft/phi-1_5 on the PIPPA dataset.

The last phasmid model for a while, lucky its a good one.

Model description

This is a Microsoft/Phi-1_5 model finetuned to persona conversation using a formatted version of the PIPPA dataset (See below.)

Intended uses & limitations

This model is intended for research purposes only. Due to the nature of the dataset, the model can produce NSFW response in certain scenarios. Unless you're okay with that please consider applying a filter or finetuning on a SFW dataset.

Model inference

Transformers

Phi doesn't support device_map "auto", and does not want to inference in fp16, so use bf16.

Thus you should use the following code for inference

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
torch.set_default_device('cuda')
model = AutoModelForCausalLM.from_pretrained("SE6446/Phasmid-1_5-V0_5", trust_remote_code=True, torch_dtype="auto")
tokenizer = AutoTokenizer.from_pretrained("SE6446/Phasmid-1_5-V0_5", trust_remote_code=True, torch_dtype="auto")
inputs = tokenizer('### Instruction: Penguinotron\'s persona:\n Penguinotron is a robotic guardian of all penguins. He will kill any one who attempts to harm penguins.\n\n<START>\n### Input: Oliver: Hello penguinotron!\n ### Output (length=any): Penguinotron:', return_tensors="pt", return_attention_mask=False)
outputs = model.generate(**inputs, max_length=200)
text = tokenizer.batch_decode(outputs)[0]
print(text)

Web UIs

Can be installed normally or via git...

cd {{Insert model dir}}
git lfs install
git clone https://huggingface.co/SE6446/Phasmid-1_5-V0_5

(Remember to turn on 'trust_remote_code')

Formatting

Formatting should be this variation of the Alpaca format.

### Instruction:
{Character card infomation}
### Input: {{user}}: {{prompt}}
### Output (length = any): {{char}}:

Model will output after char.

Known issues

This model is small and as such suffers from problematic hallucinations, it may produce erratic outputs or behave unrealistically based on the current context.

Downloads last month
4
Inference API
Input a message to start chatting with SE6446/Phasmid-1_5-V0_5.
Inference API (serverless) does not yet support model repos that contain custom code.

Dataset used to train SE6446/Phasmid-1_5-V0_5