Update README.md (#1)
Browse files- Update README.md (73f2e2353217e155abc045a6edc3788856bd8572)
README.md
CHANGED
|
@@ -1,4 +1,76 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
---
|
| 4 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
---
|
| 4 |
+
|
| 5 |
+
# Models for Echo Application
|
| 6 |
+
|
| 7 |
+
This repository contains **LiteRT-compatible language model variants** used by the AI engine of the **Echo application**.
|
| 8 |
+
All models here are optimized and validated specifically for **LiteRT adaptations** of the framework on which the application AI engine is built.
|
| 9 |
+
|
| 10 |
+
The models listed below are **standard, stable, and fully working variants** used for chat functionality.
|
| 11 |
+
|
| 12 |
+
Repository link:
|
| 13 |
+
https://huggingface.co/ANISH-j/models-for-echo-application/tree/main
|
| 14 |
+
|
| 15 |
+
---
|
| 16 |
+
|
| 17 |
+
## Supported Model Variants
|
| 18 |
+
|
| 19 |
+
### 1. `Gemma3-1B-IT_multi-prefill-seq_q4_ekv4096.litertlm`
|
| 20 |
+
|
| 21 |
+
- **Model family:** Gemma 3
|
| 22 |
+
- **Size:** 1B parameters
|
| 23 |
+
- **Quantization:** Q4
|
| 24 |
+
- **Format:** LiteRT model (`.litertlm`)
|
| 25 |
+
- **KV Cache:** Extended KV (4096)
|
| 26 |
+
- **Features:**
|
| 27 |
+
- Multi-prefill sequence support
|
| 28 |
+
- Optimized memory usage
|
| 29 |
+
- Efficient long-context chat handling
|
| 30 |
+
|
| 31 |
+
**Recommended for:**
|
| 32 |
+
Chat scenarios requiring longer conversational context with optimized KV-cache performance.
|
| 33 |
+
|
| 34 |
+
---
|
| 35 |
+
|
| 36 |
+
### 2. `gemma3-1b-it-int4.task`
|
| 37 |
+
|
| 38 |
+
- **Model family:** Gemma 3
|
| 39 |
+
- **Size:** 1B parameters
|
| 40 |
+
- **Quantization:** INT4
|
| 41 |
+
- **Format:** LiteRT task model (`.task`)
|
| 42 |
+
- **Features:**
|
| 43 |
+
- Low-latency inference
|
| 44 |
+
- Compact model size
|
| 45 |
+
- Stable real-time chat performance
|
| 46 |
+
|
| 47 |
+
**Recommended for:**
|
| 48 |
+
Low-resource or latency-sensitive chat applications.
|
| 49 |
+
|
| 50 |
+
---
|
| 51 |
+
|
| 52 |
+
## Framework Compatibility
|
| 53 |
+
|
| 54 |
+
- Compatible with **LiteRT runtime**
|
| 55 |
+
- Tested with the **Echo application AI engine**
|
| 56 |
+
- Designed for **instruction-tuned (IT)** chat behavior
|
| 57 |
+
- Not intended for direct PyTorch or TensorFlow usage without conversion
|
| 58 |
+
|
| 59 |
+
---
|
| 60 |
+
|
| 61 |
+
## Repository Structure
|
| 62 |
+
|
| 63 |
+
models-for-echo-application/
|
| 64 |
+
├── Gemma3-1B-IT_multi-prefill-seq_q4_ekv4096.litertlm
|
| 65 |
+
├── gemma3-1b-it-int4.task
|
| 66 |
+
└── README.md
|
| 67 |
+
|
| 68 |
+
|
| 69 |
+
---
|
| 70 |
+
|
| 71 |
+
## License
|
| 72 |
+
|
| 73 |
+
Licensed under the **Apache License 2.0**.
|
| 74 |
+
You may use, modify, and distribute these models in compliance with the license.
|
| 75 |
+
|
| 76 |
+
---
|