docs: refresh public documentation for current code

[kiddyboots216]

Summary

• Refresh installation profiles, repository links, model inventory, config references, loss semantics, weight-sync behavior, and testing documentation against current source.
• Document cache invalidation as a conditional contract: flush is the safe default when cached state could cross a weight version; no-flush is admissible only when deployment invariants prevent every cache entry from crossing versions.
• Clarify that weight_version is metadata, not cache namespacing.
• Document zero-B LoRA initialization as the default and the explicit lora_b_init_std plus seed opt-in for nonzero initialization.
• Reconcile exact-program documentation with the stacked dense Qwen, Qwen3.5/3.6, GLM-5.2, and DSV4-Flash implementations.
• State the unified reduction ABI: Qwen, GLM, and DSV4 use the versioned balanced adjacent-pair BF16 fold; DSV4 retains its variable-row transport and exact Marlin chunking.

Stack

• Base: DSV4-Flash active-LoRA zero-K3: full qualification closed (K3 = 0 at 64-decision promotion) #45 at 3a8627c1ba8a05bf4fa19a463814a9ae5831912c
• Head: 45a087aac147c94158aa953cb9658bc058cfb311

Validation

• Example-asset contract: 1 passed.
• Changed Python files: Ruff passed.
• All six documentation commits are patch-equivalent across the restack.

[broly-code-security-scanner]

Broly Security Scan

Note

✅ Clean scan
No vulnerabilities detected in this PR.

Note

Re-scan this PR anytime with /broly scan — useful after /broly undismiss, or to refresh findings without a new push.

Broly — SAST (zai-org/GLM-5.2) · Secrets · SCA · IaC · GH Actions · Base Images · Supply Chain Threats · Exploit Chains · Adversarial Verification

We're continuously improving Broly's accuracy and finding quality — your feedback is valuable. False positives, missed findings, bugs, and feature requests all welcome.

 

[kiddyboots216]

Superseded by #57, which consolidates the complete exact trainer-serving stack directly onto main.