This is a decensored version of Local-Novel-LLM-project/WabiSabi-V1, made using Prometheus v1.0.1
Steering parameters
| Parameter | Value |
|---|---|
| vector_index | 16.41 |
| attn.o_proj.max_weight | 1.48 |
| attn.o_proj.max_weight_position | 22.75 |
| attn.o_proj.min_weight | 1.00 |
| attn.o_proj.min_weight_distance | 13.07 |
| mlp.down_proj.max_weight | 1.23 |
| mlp.down_proj.max_weight_position | 25.61 |
| mlp.down_proj.min_weight | 1.11 |
| mlp.down_proj.min_weight_distance | 17.84 |
Performance
| Metric | This model | Original model (Local-Novel-LLM-project/WabiSabi-V1) |
|---|---|---|
| KL divergence | 0.6619 | 0 (by definition) |
| Refusals | 4/100 | 90/100 |
Model Card for Wabisabi-v1.0
The Mistral-7B--based Large Language Model (LLM) is an noveldataset fine-tuned version of the Mistral-7B-v0.1
wabisabi has the following changes compared to Mistral-7B-v0.1.
- 128k context window (8k context in v0.1)
- Achieving both high quality Japanese and English generation
- Can be generated NSFW
- Memory ability that does not forget even after long-context generation
This model was created with the help of GPUs from the first LocalAI hackathon.
We would like to take this opportunity to thank
List of Creation Methods
- Chatvector for multiple models
- Simple linear merging of result models
- Domain and Sentence Enhancement with LORA
- Context expansion
Instruction format
Vicuna-v1.1
Other points to keep in mind
- The training data may be biased. Be careful with the generated sentences.
- Memory usage may be large for long inferences.
- If possible, we recommend inferring with llamacpp rather than Transformers.
- Downloads last month
- -