Model Type | storytelling, instruction-following |
|
Use Cases |
Areas: | storytelling, creative writing, role-playing |
|
Applications: | novel writing, interactive fiction |
|
Primary Use Cases: | >40K context, instruct-enhanced storytelling |
|
|
Additional Notes | Tested for novel-style continuation, assistant-type responses, and long context analysis without refusals. Specific to certain configurations for performance of storytelling in longer contexts. |
|
Supported Languages | |
Training Details |
Data Sources: | DrNicefellow/ChatAllInOne-Yi-34B-200K-V1, migtissera/Tess-34B-v1.5b, cgato/Thespis-34b-v0.7, Doctor-Shotgun/limarpv3-yi-llama-34b-lora, adamo1139/yi-34b-200k-rawrr-dpo-2, migtissera/Tess-M-Creative-v1.0, NousResearch/Nous-Capybara-34B |
|
Methodology: | Merge using DARE (Discrete Alignment-based Row Expansion) technique |
|
Context Length: | |
|
Input Output |
Input Format: | SYSTEM: {system_message} USER: {prompt} ASSISTANT: |
|
Accepted Modalities: | |
Output Format: | |
Performance Tips: | Running Chinese models with large tokenizer vocabularies like Yi need careful parameter tuning due to large logit sampling tails. |
|
|