parscale
github.com/qwenlm/parscale
cd6acb4mainsynced 2w ago
Code HealthView report
Average health
8.3/10
Excellent
Hotspot health
10.0/10
Excellent
CostsView costs →
What this repo's indexing and doc generation cost, per sync.
Open the Costs page →Attention Needed11
configuration_qwen2_parscale.pyKnowledge Silo
100% single-owner concentration
modeling_qwen2_parscale.pyDead Code
Safe to delete (650 lines)
cost_analysis.pyKnowledge Silo
100% single-owner concentration
cost_analysis.pyDead Code
Safe to delete (0 lines)
modeling_qwen2_parscale.pyKnowledge Silo
100% single-owner concentration
No commit history
Commit activity appears once git history is indexed.
Recent DecisionsView all
This .to() is needed if the model has been moved to a device after being initialized (because the buproposed
· code commentSince we use batch_size * p as the new batch size, the latency for llm-analysis assumes the embeddinproposed
· code commentUse cost analysis tooling with llm-analysis dependencyproposed
· readme miningUse parametric fitting code to fit the parallel scaling lawproposed
· readme miningRelease specific inference and configuration code for ParScaleproposed
· readme miningContinual pretraining approach: freeze backbone and fine-tune new parametersproposed
· readme miningNo hotspot data
No churn or complexity hotspots detected in this repository.