Upload README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,172 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
<!DOCTYPE html>
|
| 2 |
+
<html lang="en">
|
| 3 |
+
<head>
|
| 4 |
+
<meta charset="UTF-8">
|
| 5 |
+
<meta name="viewport" content="width=device-width, initial-scale=1.0">
|
| 6 |
+
<title>Model README</title>
|
| 7 |
+
<style>
|
| 8 |
+
body {
|
| 9 |
+
background: linear-gradient(-45deg, #0a0a0a, #121212, #1a1a1a);
|
| 10 |
+
color: #E0E0E0;
|
| 11 |
+
font-family: 'Segoe UI', system-ui;
|
| 12 |
+
margin: 0;
|
| 13 |
+
padding: 20px;
|
| 14 |
+
min-height: 100vh;
|
| 15 |
+
animation: gradient 15s ease infinite;
|
| 16 |
+
background-size: 400% 400%;
|
| 17 |
+
text-align: center;
|
| 18 |
+
}
|
| 19 |
+
@keyframes gradient {
|
| 20 |
+
0% { background-position: 0% 50%; }
|
| 21 |
+
50% { background-position: 100% 50%; }
|
| 22 |
+
100% { background-position: 0% 50%; }
|
| 23 |
+
}
|
| 24 |
+
.container {
|
| 25 |
+
max-width: 800px;
|
| 26 |
+
margin: auto;
|
| 27 |
+
}
|
| 28 |
+
.model-image {
|
| 29 |
+
width: 100%;
|
| 30 |
+
border-radius: 12px;
|
| 31 |
+
filter: drop-shadow(0 0 10px rgba(255, 255, 255, 0.1));
|
| 32 |
+
animation: float 6s ease-in-out infinite;
|
| 33 |
+
}
|
| 34 |
+
@keyframes float {
|
| 35 |
+
0%, 100% { transform: translateY(0); }
|
| 36 |
+
50% { transform: translateY(-20px); }
|
| 37 |
+
}
|
| 38 |
+
.box {
|
| 39 |
+
background: rgba(30, 30, 30, 0.9);
|
| 40 |
+
border-radius: 12px;
|
| 41 |
+
padding: 20px;
|
| 42 |
+
margin: 25px 0;
|
| 43 |
+
backdrop-filter: blur(10px);
|
| 44 |
+
border: 1px solid rgba(255, 255, 255, 0.1);
|
| 45 |
+
text-align: left;
|
| 46 |
+
}
|
| 47 |
+
h2 {
|
| 48 |
+
border-left: 4px solid #0ff;
|
| 49 |
+
padding-left: 15px;
|
| 50 |
+
margin: 0 0 15px 0;
|
| 51 |
+
background: linear-gradient(90deg, transparent, rgba(0, 255, 255, 0.1));
|
| 52 |
+
text-transform: uppercase;
|
| 53 |
+
letter-spacing: 2px;
|
| 54 |
+
color: #fff;
|
| 55 |
+
}
|
| 56 |
+
.yaml-content {
|
| 57 |
+
background: #191919;
|
| 58 |
+
border-radius: 8px;
|
| 59 |
+
padding: 10px;
|
| 60 |
+
margin-top: 10px;
|
| 61 |
+
font-family: monospace;
|
| 62 |
+
white-space: pre-wrap;
|
| 63 |
+
color: #E0E0E0;
|
| 64 |
+
border-left: 4px solid #0ff;
|
| 65 |
+
}
|
| 66 |
+
/* Custom Scrollbar */
|
| 67 |
+
::-webkit-scrollbar { width: 8px; }
|
| 68 |
+
::-webkit-scrollbar-track { background: #121212; }
|
| 69 |
+
::-webkit-scrollbar-thumb {
|
| 70 |
+
background: #333;
|
| 71 |
+
border-radius: 4px;
|
| 72 |
+
}
|
| 73 |
+
</style>
|
| 74 |
+
</head>
|
| 75 |
+
<body>
|
| 76 |
+
<div class="container">
|
| 77 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/66c26b6fb01b19d8c3c2467b/tqI2XfovbkA_0ss6IKlPq.png" class="model-image" alt="Model Visualization">
|
| 78 |
+
<div class="box">
|
| 79 |
+
<h2>🔍 Overview</h2>
|
| 80 |
+
<p>This is the second in a line of models dedicated to creating Stable-Diffusion prompts when given a character appearance. Made for the CharGen Project, This has been finetuned ontop of Delta-Vector/Holland-4B-V1</>
|
| 81 |
+
</div>
|
| 82 |
+
<div class="box">
|
| 83 |
+
<h2>⚖️ Quants</h2>
|
| 84 |
+
<p>Available quantization formats:</p>
|
| 85 |
+
<ul>
|
| 86 |
+
<li>GGUF: https://huggingface.co/mradermacher/SDPrompter4b-GGUF</li>
|
| 87 |
+
<li>EXL2: https://huggingface.co/</li>
|
| 88 |
+
</ul>
|
| 89 |
+
</div>
|
| 90 |
+
<div class="box">
|
| 91 |
+
<h2>💬 Prompting</h2>
|
| 92 |
+
<p><strong>Recommended format: ChatML, Use the following system prompt for the model. I'd advise against setting a high amount of output tokens as the model loops, use 0.1 min-p and temp-1 to keep it coherent.</strong></p>
|
| 93 |
+
<code>Create a prompt for Stable Diffusion based on the information below.</code>
|
| 94 |
+
</div>
|
| 95 |
+
<div class="box">
|
| 96 |
+
<h2>🌟 Credits</h2>
|
| 97 |
+
<p>Finetuned on 1xRTX6000 provided by Kubernetes_bad, All credits goes to Kubernetes_bad, LucyKnada and the rest of Anthracite.</p>
|
| 98 |
+
</div>
|
| 99 |
+
<div class="box">
|
| 100 |
+
<h2>🛠️ Axolotl Config)</h2>
|
| 101 |
+
<pre>
|
| 102 |
+
base_model: Delta-Vector/Holland-4B-V1
|
| 103 |
+
model_type: AutoModelForCausalLM
|
| 104 |
+
tokenizer_type: AutoTokenizer
|
| 105 |
+
trust_remote_code: true
|
| 106 |
+
load_in_8bit: false
|
| 107 |
+
load_in_4bit: false
|
| 108 |
+
strict: false
|
| 109 |
+
datasets:
|
| 110 |
+
- path: NewEden/CivitAI-SD-Prompts
|
| 111 |
+
datasets:
|
| 112 |
+
- path: NewEden/CivitAI-Prompts-Sharegpt
|
| 113 |
+
type: chat_template
|
| 114 |
+
chat_template: chatml
|
| 115 |
+
roles_to_train: ["gpt"]
|
| 116 |
+
field_messages: conversations
|
| 117 |
+
message_field_role: from
|
| 118 |
+
message_field_content: value
|
| 119 |
+
train_on_eos: turn
|
| 120 |
+
dataset_prepared_path:
|
| 121 |
+
val_set_size: 0.02
|
| 122 |
+
output_dir: ./outputs/out2
|
| 123 |
+
sequence_len: 8192
|
| 124 |
+
sample_packing: true
|
| 125 |
+
eval_sample_packing: false
|
| 126 |
+
pad_to_sequence_len: true
|
| 127 |
+
plugins:
|
| 128 |
+
- axolotl.integrations.liger.LigerPlugin
|
| 129 |
+
liger_rope: true
|
| 130 |
+
liger_rms_norm: true
|
| 131 |
+
liger_swiglu: true
|
| 132 |
+
liger_fused_linear_cross_entropy: true
|
| 133 |
+
wandb_project: SDprompter-final
|
| 134 |
+
wandb_entity:
|
| 135 |
+
wandb_watch:
|
| 136 |
+
wandb_name: SDprompter-final
|
| 137 |
+
wandb_log_model:
|
| 138 |
+
gradient_accumulation_steps: 16
|
| 139 |
+
micro_batch_size: 1
|
| 140 |
+
num_epochs: 4
|
| 141 |
+
optimizer: paged_adamw_8bit
|
| 142 |
+
lr_scheduler: cosine
|
| 143 |
+
learning_rate: 0.00001
|
| 144 |
+
train_on_inputs: false
|
| 145 |
+
group_by_length: false
|
| 146 |
+
bf16: auto
|
| 147 |
+
fp16:
|
| 148 |
+
tf32: true
|
| 149 |
+
gradient_checkpointing: true
|
| 150 |
+
gradient_checkpointing_kwargs:
|
| 151 |
+
use_reentrant: false
|
| 152 |
+
early_stopping_patience:
|
| 153 |
+
resume_from_checkpoint:
|
| 154 |
+
local_rank:
|
| 155 |
+
logging_steps: 1
|
| 156 |
+
xformers_attention:
|
| 157 |
+
flash_attention: true
|
| 158 |
+
warmup_ratio: 0.05
|
| 159 |
+
evals_per_epoch: 4
|
| 160 |
+
saves_per_epoch: 1
|
| 161 |
+
debug:
|
| 162 |
+
weight_decay: 0.01
|
| 163 |
+
special_tokens:
|
| 164 |
+
pad_token: <|finetune_right_pad_id|>
|
| 165 |
+
eos_token: <|eot_id|>
|
| 166 |
+
auto_resume_from_checkpoints: true
|
| 167 |
+
</pre>
|
| 168 |
+
</div>
|
| 169 |
+
</div>
|
| 170 |
+
</div>
|
| 171 |
+
</body>
|
| 172 |
+
</html>
|