From 5fa2239d5764cdf16e66d84ea132a14b20fe629d Mon Sep 17 00:00:00 2001 From: randaller Date: Sun, 19 Mar 2023 18:20:41 +0300 Subject: [PATCH] Update README.md --- README.md | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/README.md b/README.md index 5f79ae3..33e780a 100644 --- a/README.md +++ b/README.md @@ -211,6 +211,8 @@ python hf-inference-cuda-example.py Modify hf-training-example.py, also feel free to use more or less lines of SD prompts examples in csv file: +You may also prepare your own dataset, for example, with Positive and Negative and even Sampler etc lines interleaving in csv. + ``` MODEL = 'decapoda-research/llama-7b-hf' DATA_FILE_PATH = 'datasets/stable_diffusion_prompts.csv' @@ -223,6 +225,12 @@ Then run the training, then after a long-long time, use something like this as p batch = tokenizer("A portrait of a beautiful girl, ", return_tensors="pt") ``` +If you have used dataset with Positive: Negative: lines, initial prompt may looks like: + +``` +batch = tokenizer("Positive: A warship flying thru the Wormhole, ", return_tensors="pt") +``` + Run inference, this should return continued prompt. ## Reference