jiarui_li

jiarui_li created repository jiarui_li/DeepHealthV2

2026-01-19 23:14:37 +08:00

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-19 14:41:48 +08:00

ea4682c35a Enhance summary output in extract_sequence_lengths script to include min, max, and percentile values for better data insights

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-19 14:39:20 +08:00

29913106cb Add get_sequence_lengths method to HealthDataset and create extract_sequence_lengths script for per-patient sequence length extraction

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-19 00:02:19 +08:00

76d3fed76f Enhance compute_disease_capture_at_k_fast to support return counts and update LandmarkEvaluator for backward compatibility with new metrics structure

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-18 20:54:54 +08:00

d13fa430b7 Validate bin edges for Piecewise Exponential CIF Loss to ensure at least two finite edges and correct starting point

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-18 18:14:52 +08:00

6de2132e84 Add fast vectorized implementation for Disease-Capture@K and enhance LandmarkEvaluator with profiling and correctness check options

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-18 18:05:02 +08:00

6e76d67a10 Enhance _maybe_torch_compile and add _maybe_cudagraph_mark_step_begin for improved CUDA Graphs handling

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-18 17:59:35 +08:00

a4b19b6e08 Enhance LandmarkEvaluator with model compilation and optimization options

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-18 17:43:54 +08:00

014393a33f Add support for evaluation indices in LandmarkEvaluator class

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-18 17:38:26 +08:00

0057bc0dd9 Refactor evaluation scripts for multi-GPU execution

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-18 15:41:15 +08:00

b80d9a4256 Remove evaluate_next_event.py and utils.py files to streamline the codebase. These files contained functions and classes related to evaluation and utility operations that are no longer needed.

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-18 00:06:45 +08:00

1d682c130d Implement control for implicit parallelism using --max_cpu_cores argument

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-17 23:53:33 +08:00

cfe7f88162 Add --max_cpu_cores argument for parallel processing in evaluation scripts

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-17 23:40:03 +08:00

248fb09c34 Refactor run_evaluations_multi_gpu.sh to remove --tau-short requirement and add support for next-args and horizon-args files

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-17 23:34:26 +08:00

4686b56336 Refactor next-event evaluation to use next-token scores and implement clean control AUC metrics

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-17 16:13:08 +08:00

e56068e668 Add compare_models.py for model evaluation with primary metrics and age-bin validation

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-17 15:31:20 +08:00

fcd948818c Enhance next-event evaluation with age-bin metrics and diagnostic AUC outputs

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-17 14:57:00 +08:00

197842b1a6 Add script for multi-GPU evaluations with flexible options and logging

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-17 14:42:14 +08:00

a90f22a865 Enhance HealthDataset with caching for event tensors and update evaluation scripts to use test subsets

jiarui_li pushed to main at jiarui_li/DeepHealth

2026-01-17 14:10:02 +08:00

7840a4c53e Refactor next-event evaluation logic and add age-bin metrics output