deepseek-hermes-reasoning-traces

 # DeepSeek V4 Pro Hermes Reasoning Traces 19,331 multi-turn ChatML + Hermes reasoning traces generated by DeepSeek V4 Pro. Designed for LoRA fine-tuning local models to operate as Hermes Agent instances. ## Quick Start \ ## Splits | Split | Traces | |-------|--------| | train | 16,431 | | valid | 1,933 | | test | 967 | ## Variants (VRAM-Tiered) | Variant | Max Tokens | Traces | GPU | |---------|-----------|--------|-----| | nano | 2,048 | 15,948 | Dev / 7B | | budget | 4,096 | 2,149 | 48GB | | standard | 8,192 | 990 | 64GB | | spark | 16,384 | 244 | 128GB DGX | ## Tools (138K total) | Tool | Count | Type | |------|-------|------| | terminal | 35,953 | Core | | read_file | 32,785 | Core | | search_files | 16,451 | Core | | execute_code | 10,311 | Core | | web_search | 8,717 | Core | | memory | 6,538 | Hermes | | session_search | 3,930 | Hermes | | skill_manage | 1,287 | Hermes | | delegate_task | 1,280 | Hermes | | cronjob | 751 | Hermes | ## Target Models Unsloth LoRA configs in \: - Qwen 3.6 27B (64GB QLoRA, 8K seq) - Nemotron Omni 30B (48GB QLoRA) - Ling-2.6-flash (48GB QLoRA) - Mistral-Medium-3.5 128B, GLM-5.1, DeepSeek V4 Flash ## Generation - Teacher: DeepSeek V4 Pro API - 96 parallel workers, 5s stagger, 99.95% success - JSON repair filter, 62% pass rate - 100% think blocks, ChatML format