Files

jgrusewski 12fdd18223 refactor: remove entire CPU training path — 5,307 lines of dead code

Deleted:
- DQN::compute_loss_internal (280 lines) — old Candle forward+loss
- DQN::train_step (55 lines) — old Candle training step
- DQN::compute_gradients (47 lines) — old gradient accumulation
- ComputeLossResult struct — only used by deleted functions
- RegimeConditionalDQN::train_step (65 lines) — old dispatch
- RegimeConditionalDQN::train_step_gpu_regime (100 lines) — old GPU path
- RegimeConditionalDQN::compute_gradients_gpu (130 lines) — old regime gradients
- RegimeConditionalDQN::compute_gradients (92 lines) — old dispatch
- DQNAgentType::train_step dispatch — dead
- DQNAgentType::compute_gradients dispatch — dead
- GpuDqnTrainer::upload_batch (71 lines) — old CPU→GPU upload
- train_step.rs (500 lines) — entire module including ensure_fused_ctx
- dqn_benchmark.rs — used old train_step
- examples.rs — used old train_step
- validation/adapters.rs (289 lines) — used old train_step
- dqn/trainable_adapter.rs — used old train_step
- gpu_smoketest.rs — tested old train_step
- Gradient accumulation path in training_loop.rs (144 lines)
- IQN d_h_s2().clone() → raw pointer (zero alloc)
- Causal intervention format! string alloc removed
- Dead HER relabel functions (320 lines)

Kept:
- ensure_fused_ctx logic inlined into training_loop.rs
- set_noise_sigma_scale re-added to RegimeConditionalDQN

Fixed:
- GpuReplayBuffer max_batch_size wired from batch_size parameter
  (was hardcoded 1024, blocking batch_size=8192)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

2026-04-02 09:06:23 +02:00

1.7 KiB

Raw Blame History

Analysis Commands Compliance Report

Overview

Reviewed all command files in .claude/commands/analysis/ directory to ensure proper usage of:

mcp__claude-flow__* tools (preferred)
npx claude-flow commands (as fallback)
No direct implementation calls

Files Reviewed

1. token-efficiency.md

Status: ✅ Updated Changes Made:

Replaced npx ruv-swarm hook session-end --export-metrics with proper MCP tool call
Updated to: Tool: mcp__claude-flow__token_usage with appropriate parameters
Maintained result format and context

Before:

npx ruv-swarm hook session-end --export-metrics

After:

Tool: mcp__claude-flow__token_usage
Parameters: {"operation": "session", "timeframe": "24h"}

2. performance-bottlenecks.md

Status: ✅ Compliant (No changes needed) Reason: Already uses proper mcp__claude-flow__task_results tool format

Summary

Total files reviewed: 2
Files updated: 1
Files already compliant: 1
Compliance rate after updates: 100%

Compliance Patterns Enforced

MCP Tool Usage: All direct tool calls now use mcp__claude-flow__* format
Parameter Format: JSON parameters properly structured
Command Context: Preserved original functionality and expected results
Documentation: Maintained clarity and examples

Recommendations

All analysis commands now follow the proper pattern
No direct bash commands or implementation calls remain
Token usage analysis properly integrated with MCP tools
Performance analysis already using correct tool format

The analysis directory is now fully compliant with the Claude Flow command standards.

1.7 KiB Raw Blame History