Writing about post-training
-
Within the abyss of assembly lies an RL environment
4 min read
A breakdown of an RL environment I designed after misreading a take-home assignment. The design is built around my reverse-engineering background and I liked it enough to write it up here.
-
Speaking Thai, Part 2: Four Frozen Embeddings
6 min read
A LoRA debugging story about four embeddings that never got to learn anything. If you add tokens and freeze the base model, their embedding and LM-head rows still need to train.
#post-training , #sft , #lora , #tokenizers , #ai-assisted
-
Why Does My GPU Utilization Keep Dipping?
9 min read
Reading GPU telemetry on a colocated GRPO run: periodic utilization dips, memory-access spikes, and a bottleneck story of mine that didn't survive contact with the code.
#rl , #post-training , #grpo , #vllm , #ai-assisted
-
Why Is My Model Suddenly Speaking Thai?
8 min read
A debugging story about special tokens, GRPO, and the skip_special_tokens flags scattered across your RL stack. When your SFT'd model nails the format in eval, then falls apart the moment it enters the GRPO rollout.
#rl , #post-training , #grpo , #tokenizers , #ai-assisted
-
Why is my GRPO Loss 0?
8 min read
When your rewards look fine, but the loss is zero. A debugging story about vLLM, tokenizers, and importance-sampling ratios.
#rl , #post-training , #grpo , #vllm , #ai-assisted