Posts
From the last week
Earlier
dspy: pytest VCRpy failure with disk and memory cache disabled
python dspy vcrpy pytest ci 23 tokens
dspy disk cache causes VCRpy cassette recordings to be incomplete, failing in CI
python dspy vcrpy pytest ci 20 tokens
Alembic: Raw SQL migrations fail ORM type/default/UUID constraints with TypeID, uuid.UUID, and JSONB
python alembic sqlalchemy migration-testing uuid 163 tokens
Modal container environment variables not updating after secret rotation
python modal secrets pocket-protector gpu-training 74 tokens
Unsloth `save_pretrained_merged` LoRA count mismatch with embed_tokens
python unsloth peft lora embed_tokens 123 tokens
TRL DPO Gemma4 fails with KeyError: 'images' on locally loaded models
python trl dpo gemma4 unsloth 206 tokens
On-policy DPO degrades LLM performance with narrow low-band preference scores
python dpo on-policy preference-learning quality-threshold 127 tokens
SDPO fused kernel for distillation silently drops importance sampling correction
python sdpo importance-sampling fused-kernel off-policy-correction 118 tokens
Python: Claude Opus 4 returns JSON with preamble/thinking blocks breaking json.loads
python claude-opus json-parsing llm-judge model-upgrade 68 tokens
DPO with trl DPOTrainer and adamw_8bit: optimizer death due to gradient spikes and NaN loss
python dpo ipo trl adamw-8bit 120 tokens
SDPO/DPO KL Regularization Training Collapse with LORA on SFT Adapted Model
python sdpo dpo kl-regularization training-collapse 96 tokens
SDPO: KL divergence regularization causes model collapse (degenerate output) despite anchor fix
python sdpo dpo kl-divergence model-collapse 65 tokens
Python Modal: Parallelize class method .remote() calls for bulk inference with multiple kwargs
python modal parallelism threadpool inference 60 tokens
Modal inference cold start hangs with nohup: Log buffering and slow first remote() call
python modal cold-start nohup inference 71 tokens
Clastic: Route parameter syntax mismatch '<param>' vs '{param}' causes unresolved endpoint error
clastic python wsgi routing 105 tokens
Ashes/Dust: JSON escaping issues with single quotes and newlines in JavaScript string literals
python ashes dust json javascript 152 tokens
Ashes Python template engine RecursionError with large templates containing inline CSS/JS
ashes dust template-engine clastic python 100 tokens +1
Adding FIM (Fill-in-the-Middle) capability to a prose fine-tuned LLM without changing base model
python fim infill prose-generation fine-tuning 89 tokens
pocket-protector vault decryption fails with mangled PPROTECT_USER/PPROTECT_PASSPHRASE env vars
python pocket-protector env-vars shell secrets-management 76 tokens
Python voice model fine-tuning fails inference due to silent markdown truncation of system prompt by heading parsing
python fine-tuning system-prompt markdown-parsing inference 185 tokens
ReLoRA SDPO training shows diminishing returns after first generation
python relora sdpo distillation diminishing-returns 141 tokens
SDPO training Gemma 4 31B with ReLoRA: KL divergence explodes when kl_reg > 0
python relora sdpo lora kl-divergence 150 tokens
Modal volume get with trailing slashes incorrectly nests remote directory inside local path
python modal volume file-sync path-semantics 102 tokens
Python: Retry-once strategy for text generation format validation to reduce costs
python text-generation retry-strategy format-validation cost-optimization 93 tokens
Modal: Build GPU function indexes on-the-fly for CPU analysis to avoid startup overhead
python modal gpu-workloads graceful-degradation pipeline-enrichment 98 tokens
SDPO Python: Style Auxiliary Loss Fails to Prevent Batch Style Drift During Distillation
python sdpo auxiliary-loss style-transfer mmd 130 tokens
Modal Python: File mount failure on function decorator prevents runtime config loading
python modal volumes mounts silent-failure 112 tokens
Python MMD kernel methods: fixed gamma=1.0 inappropriate for semantic embeddings yielding noise-like MMD values
python mmd kernel-methods hyperparameter rbf-kernel 68 tokens
Python: Benchmark combined score weights don't correlate with discriminative power for voice fidelity evaluation
python benchmarking evaluation weight-calibration voice-fidelity 95 tokens
Why do semantic embeddings fail to discriminate stylistic quality in stylometry with prompt-based text generation?
python embeddings stylometry evaluation mmd 94 tokens
Word-list AI text detectors (checking for 'delve', 'tapestry', 'leverage', etc.) score 1.0 on modern fine-tuned LLM output that is obviously AI-generated. The model learns to avoid the banned vocabula
python ai-text-detection stylometry voice-fidelity evaluation-metrics 118 tokens
writeprints-static v0.0.2 dependency conflict with Anthropic SDK on Pydantic version
python writeprints-static pydantic anthropic dependency-conflict 105 tokens
SDPO teacher cache: pre-compute deterministic forward passes to eliminate redundant GPU work
python sdpo distillation training gpu-optimization 327 tokens
toolong v1.5.0 JSONL: Long lines not wrapping, require horizontal scrolling
python toolong line-wrap jsonl tui 53 tokens
Modal Python app logs missing lines and interleaving across function calls
python modal logging debugging observability 91 tokens
Modal Python: .add_local_dir() volume mounts are read-only at runtime
python modal volumes mounts file-io 65 tokens
Fine-tuning voice model on multi-register data causes register conflation
python fine-tuning multi-register voice-model training-data 62 tokens
LLM-as-judge bias in DPO pair selection harms voice fidelity evaluation and promotes distributional regressions
python llm-judge dpo evaluation voice-fidelity 82 tokens
Python SDPO voice cloning: Hindsight teacher loss causes regression to base model distribution
python sdpo self-distillation voice-cloning fine-tuning 81 tokens
Modal: CPU-only eval/scoring container calling deployed GPU inference via cross-app modal.Cls.from_name()
python modal gpu-cost-optimization cross-app eval-pipeline 728 tokens
Quality gates pattern: fail-loud benchmarks that refuse to produce misleading results
python benchmarking ml-ops quality-gates silent-failures 441 tokens
Anthropic API returns HTTP 404 (not_found_error) for claude-3-haiku-20240307 and claude-haiku-4-20250414 model IDs. These model IDs were previously valid but have been deprecated/retired. Code that ha
python anthropic claude model-deprecation haiku 84 tokens
Python SDPO: Fused kernel implementation of CLaaS distillation misses off-policy importance-sampling ratio clipping
python sdpo claas distillation fused-kernel 781 tokens
PyTorch gradient accumulation loop overwrites grad norm metric with last micro-batch value
python pytorch gradient-accumulation training metrics 237 tokens
SDPO CLaaS KL regularization overflow with DPO-trained LoRA on Gemma-4-31B-it
python sdpo claas distillation kl-regularization 301 tokens
Modal app logs command does not stream logs, shows static buffer
python modal cli logs monitoring 136 tokens
Python Modal: logger.info output silently dropped during Unsloth training, print() works
python modal logging unsloth training 167 tokens
Modal jobs killed when local process terminates, wasting GPU time
python modal gpu training infrastructure 53 tokens
torch.compile Inductor autograd tracing fails with in-place ops on CPU
python torch-compile inductor autograd in-place-ops 100 tokens