Back to Details

ito-training

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Itô Training

Itô 训练

Run training work on rented Itô metal by delegating to the canonical Itô compute backend (Layer 0.3). ECC does not implement a parallel training stack, trainer, or scheduler, and does no browser automation. This skill chains off a completed booking from 
ito-compute
; it never books, reserves, or spends.
通过委托给标准Itô计算后端（Layer 0.3），在租用的Itô硬件上运行训练任务。ECC未实现并行训练栈、训练器或调度器，也不执行浏览器自动化。该技能依赖
ito-compute
的已完成预订运行；它从不进行预订、预留或产生费用。

Prerequisite

前置条件

A completed booking from the 
ito-compute
skill (booking id, node IPs, SSH, GPU SKU, node count, fabric) in harness memory. Without one, stop.
内存中需保存来自
ito-compute
技能的已完成预订信息（预订ID、节点IP、SSH、GPU SKU、节点数量、网络架构）。若无该信息，将停止运行。

Delegation

委托机制

ECC calls the canonical backend through the 
ecc ito
bridge; it never re-implements training. Authenticate once with 
ecc ito login
, as 
ito-compute
documents. Never put a key or token in arguments, files, logs, or chat.
sh
ecc ito train \
  --booking <booking-id> \
  --model-size <e.g. 8B> \
  --data <data-ref> \
  --target <capability> \
  --budget-usd <ceiling> \
  [--post-training sft|dpo|rlvr]
ECC通过
ecc ito
桥接调用标准后端；它从不重新实现训练功能。需按照
ito-compute
文档说明，通过
ecc ito login
完成一次认证。切勿在参数、文件、日志或聊天中放置密钥或令牌。
sh
ecc ito train \
  --booking <booking-id> \
  --model-size <e.g. 8B> \
  --data <data-ref> \
  --target <capability> \
  --budget-usd <ceiling> \
  [--post-training sft|dpo|rlvr]

What the backend does (Layer 0.3)

后端功能（Layer 0.3）

The desk backend runs a staged, eval-gated pipeline; this skill reports stage gates and never overrides one:
  1. Data prep — manifest, dedup, decontamination against the eval suite; 150M-ladder decision job as the cheap pre-check for custom data.
  2. Parallelism and precision — selected from model size, node count, fabric; wasteful combinations refused.
  3. Checkpointing and fault tolerance — async DCP, torchft; detect < 10 min, resume < 15 min. Loss-spike restart is a proposed, human-gated action.
  4. Curriculum and eval gates — staged pretrain / mid-train / long-context / post-training, each with a fixed eval battery; a failed gate stops the run.
  5. Post-training — SFT → DPO → RLVR (GRPO with DAPO stability fixes), trainer/rollout separation with bounded staleness.
Emits desk telemetry (goodput, interruption rate, checkpoint bandwidth) so the desk prices training blocks honestly.
桌面后端运行一个分阶段、受评估管控的流水线；该技能仅报告阶段管控节点，从不覆盖它们：
  1. 数据预处理 — 生成清单、去重、针对评估套件进行数据净化；通过1.5亿级阶梯决策任务作为自定义数据的低成本预检查。
  2. 并行性与精度 — 根据模型规模、节点数量、网络架构自动选择；拒绝低效的组合。
  3. 检查点与容错 — 采用异步DCP、torchft；故障检测耗时<10分钟，恢复耗时<15分钟。损失突增重启为需人工确认的建议操作。
  4. 课程学习与评估管控节点 — 分为预训练/训练中期/长上下文/训练后等阶段，每个阶段均配备固定评估套件；若评估不通过，将停止运行。
  5. 训练后处理 — SFT → DPO → RLVR（带有DAPO稳定性修复的GRPO），训练器与部署分离，延迟可控。
输出桌面遥测数据（有效吞吐量、中断率、检查点带宽），以便桌面端准确计算训练区块的成本。

Unavailable today

当前未支持功能

Not yet wired: the canonical CLI's 
run
verb and the desk 
training-run
backend are scaffolds. Until they land, this skill reports the missing capability and stops. Never substitute a local trainer or a purchase endpoint.
尚未对接：标准CLI的
run
命令和桌面端
training-run
后端仍为框架阶段。在完成对接前，该技能会报告缺失的功能并停止运行。切勿使用本地训练器或其他购买端点替代。