LLM-jp 13B v2.0 Instruct AC 16×
LLM-jp 13B v2.0 Instruct AC 16× is a full-model instruction-tuned descendant of LLM-jp 13B v2.0. The underlying model has 256 billion pretraining tokens and a 4,096-token context.
Research release
Safety-data inclusion is the experimental change
The instruction mixture contains Japanese human-written examples, English source datasets and Japanese translations. Full-model tuning was performed with TRL and DeepSpeed.LLM-jp 13B v2.0 Instruct AC 16× official model card
Use the exact safety-data variant
Practical guidance. Compare this repeated-data variant with the ordinary AC model to examine the effect of data weighting. More repetitions do not guarantee better behavior on unseen requests.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | llm-jp-13b-instruct-full-ac_001_16x-dolly-ichikara_004_001_single-oasst-oasst2-v2.0 |
| Context window | 4,096 tokens Maximum sequence length stated in the model card. LLM-jp 13B v2.0 Instruct AC 16× official model card |
| Base | LLM-jp 13B v2.0.LLM-jp 13B v2.0 Instruct AC 16× official model card |
| Instruction sources | Dolly, ichikara 004-001, OpenAssistant 1 and OpenAssistant 2.LLM-jp 13B v2.0 Instruct AC 16× official model card |
| License | Apache 2.0.LLM-jp 13B v2.0 Instruct AC 16× official model card |
| Variant | AnswerCarefully data repeated sixteen times within the instruction mixture.LLM-jp v2.0 release and safety-data experiments |
Related models
- LLM-jp 13B v2.0 Base
LLM-jp / NII LLMC · LLM-jp v2
The 13B v2.0 base changes the pretraining corpus and vocabulary and extends context to 4,096 tokens.
Launch date not confirmedResearch releaseDownloadable - LLM-jp 13B v2.0 Instruct
LLM-jp / NII LLMC · LLM-jp v2
LLM-jp 13B v2.0 Instruct is one of three published safety-data tuning comparisons.
Launch date not confirmedResearch releaseDownloadable