Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,91 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
datasets:
|
| 4 |
+
- fka/awesome-chatgpt-prompts
|
| 5 |
+
language:
|
| 6 |
+
- en
|
| 7 |
+
- fr
|
| 8 |
+
- es
|
| 9 |
+
- de
|
| 10 |
+
- ch
|
| 11 |
+
- hr
|
| 12 |
+
metrics:
|
| 13 |
+
- bleu
|
| 14 |
+
new_version: black-forest-labs/FLUX.1-Kontext-dev
|
| 15 |
+
pipeline_tag: text-generation
|
| 16 |
+
library_name: adapter-transformers
|
| 17 |
+
tags:
|
| 18 |
+
- text-generation-inference
|
| 19 |
+
---
|
| 20 |
+
# π§ SpeedCore-v1 β The Foundational Brain of Speed AI
|
| 21 |
+
|
| 22 |
+
**SpeedCore-v1** is the first fully custom AI language model developed by **SpeedLab**, built entirely from scratch to power the upcoming generation of **Speed AI** systems. Trained on a diverse, deeply creative, future-facing, and expressive dataset, this model is designed to think beyond the typical boundaries of traditional LLMs.
|
| 23 |
+
|
| 24 |
+
---
|
| 25 |
+
|
| 26 |
+
## π Model Highlights
|
| 27 |
+
|
| 28 |
+
- **Architecture**: Transformer Decoder (GPT-style)
|
| 29 |
+
- **Trained From Scratch**: No base model or transfer learning β this is 100% original
|
| 30 |
+
- **Vocabulary**: Custom tokenizer trained with the model
|
| 31 |
+
- **Size**: *(Fill in once trained β e.g., 117M, 350M, 1B, etc.)*
|
| 32 |
+
- **Tokens**: Trained on `X` tokens *(fill in once finalized)*
|
| 33 |
+
- **Languages**: Primarily English, but dataset includes multiple dialects, slang, future-speak, and Gen-Z expression
|
| 34 |
+
- **Personality Engine**: Multi-personality support β from chill Gen Z vibes to spiritual guides, futuristic AI personas, and beyond
|
| 35 |
+
|
| 36 |
+
---
|
| 37 |
+
|
| 38 |
+
## π‘ Intended Use
|
| 39 |
+
|
| 40 |
+
SpeedCore-v1 is built to serve as the **core intelligence behind the Speed AI ecosystem**, including:
|
| 41 |
+
|
| 42 |
+
- Conversational agents (chatbots, virtual assistants)
|
| 43 |
+
- Text generation (creative writing, content expansion)
|
| 44 |
+
- Future simulation and storytelling
|
| 45 |
+
- Language learning & personality modulation
|
| 46 |
+
- Emotional support and lifestyle coaching
|
| 47 |
+
- Spirituality, finance, relationships, and cosmic topics
|
| 48 |
+
|
| 49 |
+
> SpeedCore can express multiple identities and personalities depending on the prompt, system message, or task.
|
| 50 |
+
|
| 51 |
+
---
|
| 52 |
+
|
| 53 |
+
## π§± Training Dataset
|
| 54 |
+
|
| 55 |
+
The model was trained on a **highly customized multi-billion token dataset** created by SpeedAI, including:
|
| 56 |
+
|
| 57 |
+
- β
Future-based dialogues & philosophies
|
| 58 |
+
- β
Gen-Z, creator, alien, cosmic, and AI personas
|
| 59 |
+
- β
Topics: spirituality, sex, money, language, ethics, dreams, psychology, relationships, entertainment, futurism
|
| 60 |
+
- β
Freeform, freestyle, emotional, deep, expressive, chill tones
|
| 61 |
+
- β
User-AI conversations and raw ideation
|
| 62 |
+
|
| 63 |
+
*This dataset is currently being uploaded to the Hugging Face Datasets Hub in chunks.*
|
| 64 |
+
|
| 65 |
+
---
|
| 66 |
+
|
| 67 |
+
## π License
|
| 68 |
+
|
| 69 |
+
**Apache 2.0** β Commercial use, research, remixing, and deployment are allowed with attribution.
|
| 70 |
+
|
| 71 |
+
---
|
| 72 |
+
|
| 73 |
+
## β¨ Capabilities
|
| 74 |
+
|
| 75 |
+
- Speaks in multiple tones and slang styles (e.g., Gen Z, wise AI, playful, dark, spiritual)
|
| 76 |
+
- Can simulate multiple personas in one model
|
| 77 |
+
- Handles both structured and freestyle prompts
|
| 78 |
+
- Optimized for creativity, flow, and freedom of expression
|
| 79 |
+
|
| 80 |
+
---
|
| 81 |
+
|
| 82 |
+
## β οΈ Limitations & Safety
|
| 83 |
+
|
| 84 |
+
This is a **research model** still under active development. Some responses may be:
|
| 85 |
+
|
| 86 |
+
- Hallucinatory
|
| 87 |
+
- Opinionated
|
| 88 |
+
- Emotionally expressive beyond factual scope
|
| 89 |
+
- Lacking in factual accuracy or safety boundaries (depending on how itβs fine-tuned)
|
| 90 |
+
|
| 91 |
+
Please use responsibly in alignment with the intended
|