Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -46,5 +46,17 @@ Draft model: [Qwen3-0.6B](https://huggingface.co/darkmaniac7/TokForge-Accelerati
|
|
| 46 |
## Abliteration
|
| 47 |
This model has been abliterated (safety filters removed) for unrestricted conversation. Use responsibly.
|
| 48 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 49 |
## Export
|
| 50 |
Converted using MNN's `llmexport` pipeline with `--quant_bit 4 --quant_block 128`.
|
|
|
|
| 46 |
## Abliteration
|
| 47 |
This model has been abliterated (safety filters removed) for unrestricted conversation. Use responsibly.
|
| 48 |
|
| 49 |
+
## Limitations and Intended Use
|
| 50 |
+
|
| 51 |
+
- Intended for TokForge / MNN on-device inference, especially Android phones and tablets.
|
| 52 |
+
- The best-known uplift for this model comes from pairing it with a small CPU draft model for speculative decoding.
|
| 53 |
+
- Real throughput varies by SoC, thermal state, backend, and generation length.
|
| 54 |
+
- This repo is a runtime bundle, not a standard Transformers training checkpoint.
|
| 55 |
+
|
| 56 |
+
## Community
|
| 57 |
+
|
| 58 |
+
- Website: [tokforge.ai](https://tokforge.ai)
|
| 59 |
+
- Discord: [Join the Discord](https://discord.gg/Acv3CBtfVm)
|
| 60 |
+
|
| 61 |
## Export
|
| 62 |
Converted using MNN's `llmexport` pipeline with `--quant_bit 4 --quant_block 128`.
|