Hugging Face's logo

Datasets:
junidude14
/
PBISC-data

Modalities:
Tabular
Text
Formats:
parquet
Languages:
English
Size:
< 1K
Tags:
single-cell-rna-seq
bulk-rna-seq
diffusion
Libraries:
Datasets
pandas
Polars
Dataset card Data Studio Files
xet
 Community
1
junidude14 commited on
Commit
f53bc76
·
verified ·
1 Parent(s): fe00bc1

Archive 4 checkpoints (1-4)

Browse files
Files changed (4) hide show
  1. paper-releases/pbisc-diffusion/updates/20261002T085943Z-deck-restructure-contrast-metrics-zeroshot-len/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/scores_contrast/_code/contrast_extra_limma.f0685f68ddfc.R +79 -0
  2. paper-releases/pbisc-diffusion/updates/20261002T085943Z-deck-restructure-contrast-metrics-zeroshot-len/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/scores_contrast/_logs/imperial_20261002T152649.jobs +2 -0
  3. paper-releases/pbisc-diffusion/updates/20261002T085943Z-deck-restructure-contrast-metrics-zeroshot-len/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/scores_contrast/_logs/imperial_20261002T152649.log +158 -0
  4. paper-releases/pbisc-diffusion/updates/20261002T085943Z-deck-restructure-contrast-metrics-zeroshot-len/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/scores_contrast/_logs/sepsis_20261002T152649.jobs +3 -0
paper-releases/pbisc-diffusion/updates/20261002T085943Z-deck-restructure-contrast-metrics-zeroshot-len/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/scores_contrast/_code/contrast_extra_limma.f0685f68ddfc.R ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # edgeR TMM + limma voom + lmFit + eBayes (case vs control) for a batch of donor x cell-type pseudobulk jobs.
2
+ # CPU only. Called by tools/contrast_extra.py (added 2026-10-02); not meant to be run by hand.
3
+ #
4
+ # usage: Rscript contrast_extra_limma.R JOBDIR
5
+ # JOBDIR/n_genes.txt G (number of genes = rows of every matrix, the kept scoring genes in their fixed order)
6
+ # JOBDIR/counts.bin float64 little endian; matrix m starts at double offset `offset` and holds G x n_samples
7
+ # values in column-major order (one column per donor pseudobulk)
8
+ # JOBDIR/jobs.csv job, offset, n_samples, is_case (string of 0/1 per column), filter (nonzero | filterByExpr)
9
+ # Output:
10
+ # JOBDIR/limma.bin per job (jobs.csv order) a float64 G x 6 matrix, column-major:
11
+ # logFC, AveExpr, t, P.Value, adj.P.Val, B (moderated t of coef case - control); NaN = not tested
12
+ # JOBDIR/limma_status.csv job, status, n_tested, n_sig_fdr05, seconds
13
+ # JOBDIR/versions.txt R, limma, edgeR versions
14
+ #
15
+ # Per job: DGEList(counts) -> gene filter on that side's columns only ("nonzero": rowSums > 0, the default gene filter
16
+ # of evaluate_generated_vs_truth.py's edgeR backend; "filterByExpr": edgeR::filterByExpr(y, design)) ->
17
+ # y[keep, , keep.lib.sizes = FALSE] -> calcNormFactors(method = "TMM") -> voom(y, design) -> lmFit -> eBayes ->
18
+ # topTable(coef = 2, number = Inf, sort.by = "none"); design = model.matrix(~ group), group levels (control, case).
19
+ # BH (adj.P.Val) runs over the tested (kept) genes.
20
+ suppressPackageStartupMessages({
21
+ library(limma)
22
+ library(edgeR)
23
+ })
24
+ args <- commandArgs(trailingOnly = TRUE)
25
+ if (length(args) != 1L) stop("usage: Rscript contrast_extra_limma.R JOBDIR")
26
+ jobdir <- args[[1]]
27
+ writeLines(c(R.version.string, paste("limma", as.character(packageVersion("limma"))),
28
+ paste("edgeR", as.character(packageVersion("edgeR")))), file.path(jobdir, "versions.txt"))
29
+ G <- as.integer(readLines(file.path(jobdir, "n_genes.txt"))[[1]])
30
+ jobs <- read.csv(file.path(jobdir, "jobs.csv"), stringsAsFactors = FALSE,
31
+ colClasses = c(job = "character", offset = "numeric", n_samples = "integer",
32
+ is_case = "character", filter = "character"))
33
+ inp <- file(file.path(jobdir, "counts.bin"), "rb")
34
+ out <- file(file.path(jobdir, "limma.bin"), "wb")
35
+ status <- data.frame(job = jobs$job, status = "", n_tested = 0L, n_sig_fdr05 = 0L, seconds = 0,
36
+ stringsAsFactors = FALSE)
37
+ cols <- c("logFC", "AveExpr", "t", "P.Value", "adj.P.Val", "B")
38
+ for (i in seq_len(nrow(jobs))) {
39
+ t0 <- proc.time()[["elapsed"]]
40
+ ns <- jobs$n_samples[i]
41
+ seek(inp, where = jobs$offset[i] * 8, origin = "start", rw = "read")
42
+ x <- readBin(inp, what = "double", n = G * ns, size = 8, endian = "little")
43
+ if (length(x) != G * ns) stop(sprintf("job %s: short read", jobs$job[i]))
44
+ M <- matrix(x, nrow = G, ncol = ns)
45
+ rownames(M) <- as.character(seq_len(G))
46
+ is_case <- as.integer(strsplit(jobs$is_case[i], "")[[1]])
47
+ res <- matrix(NaN, nrow = G, ncol = 6L)
48
+ st <- "ok"
49
+ if (length(is_case) != ns) stop(sprintf("job %s: is_case length %d != %d", jobs$job[i], length(is_case), ns))
50
+ if (sum(is_case == 1L) < 2L || sum(is_case == 0L) < 2L) {
51
+ st <- "undefined_lt2_per_group"
52
+ } else {
53
+ r <- tryCatch({
54
+ grp <- factor(ifelse(is_case == 1L, "case", "control"), levels = c("control", "case"))
55
+ design <- model.matrix(~grp)
56
+ y <- DGEList(counts = M)
57
+ keep <- if (identical(jobs$filter[i], "filterByExpr")) filterByExpr(y, design = design) else rowSums(M) > 0
58
+ if (sum(keep) < 3L) stop("fewer than 3 genes pass the filter")
59
+ y <- y[keep, , keep.lib.sizes = FALSE]
60
+ y <- calcNormFactors(y, method = "TMM")
61
+ v <- voom(y, design, plot = FALSE)
62
+ fit <- eBayes(lmFit(v, design))
63
+ tt <- topTable(fit, coef = 2, number = Inf, sort.by = "none")
64
+ if (!identical(rownames(tt), rownames(M)[keep])) stop("topTable row order differs from the input")
65
+ res[keep, ] <- as.matrix(tt[, cols])
66
+ "ok"
67
+ }, error = function(e) paste0("error: ", conditionMessage(e)))
68
+ st <- r
69
+ }
70
+ writeBin(as.vector(res), out, size = 8, endian = "little")
71
+ status$status[i] <- st
72
+ status$n_tested[i] <- sum(is.finite(res[, 4]))
73
+ status$n_sig_fdr05[i] <- sum(res[, 5] < 0.05, na.rm = TRUE)
74
+ status$seconds[i] <- round(proc.time()[["elapsed"]] - t0, 3)
75
+ }
76
+ close(inp)
77
+ close(out)
78
+ write.csv(status, file.path(jobdir, "limma_status.csv"), row.names = FALSE)
79
+ cat(sprintf("limma jobs: %d (%d ok)\n", nrow(jobs), sum(status$status == "ok")))
paper-releases/pbisc-diffusion/updates/20261002T085943Z-deck-restructure-contrast-metrics-zeroshot-len/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/scores_contrast/_logs/imperial_20261002T152649.jobs ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ --design fewshot --dataset imperial_eskd --draws 0,1,2,3,4
2
+ --design pooled --dataset imperial_eskd
paper-releases/pbisc-diffusion/updates/20261002T085943Z-deck-restructure-contrast-metrics-zeroshot-len/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/scores_contrast/_logs/imperial_20261002T152649.log ADDED
@@ -0,0 +1,158 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ JOB_START 2026-10-02T15:26:49+09:00 --design fewshot --dataset imperial_eskd --draws 0,1,2,3,4
2
+ 15:26:50 imperial_eskd fewshot_n5/draw0: 142 scored samples (102 case / 31 control), 5927 genes for the disease scores
3
+ 15:27:09 split half: 435151 observed cells of the 4 types, 5 seeds; mean r(A,B) T 0.972, B 0.835, Mono 0.962, NK 0.934 (17.8 s; full aggregates rebuilt = cache)
4
+ 15:27:09 fetching generated/imperial_eskd/fewshot_n5/zs__raw_all.npz (6.49 GiB) from spark-0110
5
+ 15:28:58 zs_raw: 213000 cells; r mean_cp10k T 0.628/0.463, B 0.618/0.466, Mono 0.764/0.571, NK 0.615/0.431 (Pearson/Spearman); repro PASS (max |d| 8.3e-17) (110 s)
6
+ 15:28:59 fetching generated/imperial_eskd/fewshot_n5/zs__pseudo_all.npz (5.87 GiB) from spark-0110
7
+ 15:30:01 zs_pseudo: 213000 cells; r mean_cp10k T 0.693/0.294, B 0.575/0.228, Mono 0.770/0.563, NK 0.738/0.415 (Pearson/Spearman); repro PASS (max |d| 8.3e-17) (62 s)
8
+ 15:30:01 fetching generated/imperial_eskd/fewshot_n5/draw0/zs__cal.npz (5.31 GiB) from spark-0110
9
+ 15:31:00 zs_cal: 213000 cells; r mean_cp10k T 0.709/0.362, B 0.602/0.285, Mono 0.783/0.566, NK 0.716/0.393 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (59 s)
10
+ 15:31:00 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/draw0/zs__cal.npz
11
+ 15:31:01 fetching generated/imperial_eskd/fewshot_n5/draw0/lora__raw.npz (5.24 GiB) from spark-0110
12
+ 15:31:55 lora5: 213000 cells; r mean_cp10k T 0.814/0.712, B 0.716/0.516, Mono 0.811/0.630, NK 0.778/0.566 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (55 s)
13
+ 15:31:56 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/draw0/lora__raw.npz
14
+ 15:31:56 fetching generated/imperial_eskd/fewshot_n5/draw0/scsemi5__raw.npz (0.14 GiB) from spark-0110
15
+ 15:32:16 scsemi5: 213000 cells; r mean_cp10k T 0.643/0.565, B 0.458/0.370, Mono 0.449/0.220, NK 0.559/0.377 (Pearson/Spearman); repro PASS (max |d| 9.7e-17) (20 s)
16
+ 15:32:16 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/draw0/scsemi5__raw.npz
17
+ 15:32:44 limma: 68 jobs in 28 s (0 not ok)
18
+ 15:32:45 unit fewshot_n5/draw0 of imperial_eskd -> /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/imperial_eskd/fewshot_n5/draw0 in 355.2 s; reproduction PASS
19
+ 15:32:45 imperial_eskd fewshot_n5/draw1: 143 scored samples (102 case / 32 control), 5927 genes for the disease scores
20
+ 15:33:00 split half: 459312 observed cells of the 4 types, 5 seeds; mean r(A,B) T 0.963, B 0.768, Mono 0.891, NK 0.923 (14.3 s; full aggregates rebuilt = cache)
21
+ 15:33:19 zs_raw: 214500 cells; r mean_cp10k T 0.637/0.452, B 0.584/0.411, Mono 0.701/0.502, NK 0.621/0.430 (Pearson/Spearman); repro PASS (max |d| 8.3e-17) (20 s)
22
+ 15:33:34 zs_pseudo: 214500 cells; r mean_cp10k T 0.690/0.289, B 0.563/0.210, Mono 0.702/0.502, NK 0.721/0.385 (Pearson/Spearman); repro PASS (max |d| 8.3e-17) (15 s)
23
+ 15:33:35 fetching generated/imperial_eskd/fewshot_n5/draw1/zs__cal.npz (5.31 GiB) from spark-0110
24
+ 15:34:37 zs_cal: 214500 cells; r mean_cp10k T 0.681/0.312, B 0.580/0.251, Mono 0.689/0.484, NK 0.692/0.376 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (63 s)
25
+ 15:34:37 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/draw1/zs__cal.npz
26
+ 15:34:38 fetching generated/imperial_eskd/fewshot_n5/draw1/lora__raw.npz (5.25 GiB) from spark-0110
27
+ 15:35:35 lora5: 214500 cells; r mean_cp10k T 0.840/0.725, B 0.658/0.493, Mono 0.745/0.512, NK 0.785/0.539 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (57 s)
28
+ 15:35:35 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/draw1/lora__raw.npz
29
+ 15:35:35 fetching generated/imperial_eskd/fewshot_n5/draw1/scsemi5__raw.npz (0.14 GiB) from spark-0110
30
+ 15:35:55 scsemi5: 214500 cells; r mean_cp10k T 0.622/0.596, B 0.392/0.359, Mono 0.557/0.216, NK 0.561/0.405 (Pearson/Spearman); repro PASS (max |d| 8.3e-17) (20 s)
31
+ 15:35:55 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/draw1/scsemi5__raw.npz
32
+ 15:36:21 limma: 68 jobs in 26 s (0 not ok)
33
+ 15:36:21 unit fewshot_n5/draw1 of imperial_eskd -> /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/imperial_eskd/fewshot_n5/draw1 in 216.4 s; reproduction PASS
34
+ 15:36:22 imperial_eskd fewshot_n5/draw2: 142 scored samples (103 case / 31 control), 5927 genes for the disease scores
35
+ 15:36:31 split half: 462803 observed cells of the 4 types, 5 seeds; mean r(A,B) T 0.966, B 0.766, Mono 0.874, NK 0.920 (9.0 s; full aggregates rebuilt = cache)
36
+ 15:36:44 zs_raw: 213000 cells; r mean_cp10k T 0.633/0.446, B 0.614/0.419, Mono 0.688/0.500, NK 0.607/0.426 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (14 s)
37
+ 15:36:56 zs_pseudo: 213000 cells; r mean_cp10k T 0.673/0.249, B 0.590/0.243, Mono 0.682/0.504, NK 0.720/0.396 (Pearson/Spearman); repro PASS (max |d| 8.3e-17) (12 s)
38
+ 15:36:57 fetching generated/imperial_eskd/fewshot_n5/draw2/zs__cal.npz (5.30 GiB) from spark-0110
39
+ 15:37:52 zs_cal: 213000 cells; r mean_cp10k T 0.693/0.340, B 0.604/0.306, Mono 0.685/0.487, NK 0.697/0.396 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (56 s)
40
+ 15:37:52 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/draw2/zs__cal.npz
41
+ 15:37:52 fetching generated/imperial_eskd/fewshot_n5/draw2/lora__raw.npz (5.16 GiB) from spark-0110
42
+ 15:38:45 lora5: 213000 cells; r mean_cp10k T 0.774/0.689, B 0.642/0.435, Mono 0.698/0.480, NK 0.742/0.502 (Pearson/Spearman); repro PASS (max |d| 8.3e-17) (53 s)
43
+ 15:38:45 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/draw2/lora__raw.npz
44
+ 15:38:46 fetching generated/imperial_eskd/fewshot_n5/draw2/scsemi5__raw.npz (0.16 GiB) from spark-0110
45
+ 15:39:05 scsemi5: 213000 cells; r mean_cp10k T 0.630/0.543, B 0.490/0.358, Mono 0.605/0.256, NK 0.524/0.395 (Pearson/Spearman); repro PASS (max |d| 8.3e-17) (20 s)
46
+ 15:39:05 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/draw2/scsemi5__raw.npz
47
+ 15:39:32 limma: 68 jobs in 27 s (0 not ok)
48
+ 15:39:33 unit fewshot_n5/draw2 of imperial_eskd -> /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/imperial_eskd/fewshot_n5/draw2 in 191.5 s; reproduction PASS
49
+ 15:39:33 imperial_eskd fewshot_n5/draw3: 145 scored samples (104 case / 31 control), 5927 genes for the disease scores
50
+ 15:39:44 split half: 460834 observed cells of the 4 types, 5 seeds; mean r(A,B) T 0.963, B 0.777, Mono 0.885, NK 0.916 (10.7 s; full aggregates rebuilt = cache)
51
+ 15:39:58 zs_raw: 217500 cells; r mean_cp10k T 0.612/0.437, B 0.589/0.425, Mono 0.693/0.495, NK 0.611/0.431 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (14 s)
52
+ 15:40:10 zs_pseudo: 217500 cells; r mean_cp10k T 0.665/0.196, B 0.555/0.201, Mono 0.705/0.514, NK 0.715/0.362 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (12 s)
53
+ 15:40:11 fetching generated/imperial_eskd/fewshot_n5/draw3/zs__cal.npz (5.36 GiB) from spark-0110
54
+ 15:41:10 zs_cal: 217500 cells; r mean_cp10k T 0.662/0.219, B 0.582/0.275, Mono 0.689/0.466, NK 0.691/0.332 (Pearson/Spearman); repro PASS (max |d| 8.3e-17) (60 s)
55
+ 15:41:10 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/draw3/zs__cal.npz
56
+ 15:41:11 fetching generated/imperial_eskd/fewshot_n5/draw3/lora__raw.npz (5.39 GiB) from spark-0110
57
+ 15:42:09 lora5: 217500 cells; r mean_cp10k T 0.823/0.646, B 0.623/0.442, Mono 0.743/0.533, NK 0.794/0.547 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (59 s)
58
+ 15:42:09 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/draw3/lora__raw.npz
59
+ 15:42:10 fetching generated/imperial_eskd/fewshot_n5/draw3/scsemi5__raw.npz (0.15 GiB) from spark-0110
60
+ 15:42:33 scsemi5: 217500 cells; r mean_cp10k T 0.606/0.549, B 0.485/0.234, Mono 0.599/0.292, NK 0.675/0.414 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (23 s)
61
+ 15:42:33 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/draw3/scsemi5__raw.npz
62
+ 15:43:00 limma: 68 jobs in 27 s (0 not ok)
63
+ 15:43:00 unit fewshot_n5/draw3 of imperial_eskd -> /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/imperial_eskd/fewshot_n5/draw3 in 207.4 s; reproduction PASS
64
+ 15:43:00 imperial_eskd fewshot_n5/draw4: 143 scored samples (104 case / 31 control), 5927 genes for the disease scores
65
+ 15:43:12 split half: 459314 observed cells of the 4 types, 5 seeds; mean r(A,B) T 0.968, B 0.763, Mono 0.884, NK 0.927 (11.6 s; full aggregates rebuilt = cache)
66
+ 15:43:28 zs_raw: 214500 cells; r mean_cp10k T 0.611/0.440, B 0.623/0.459, Mono 0.695/0.512, NK 0.615/0.445 (Pearson/Spearman); repro PASS (max |d| 8.3e-17) (15 s)
67
+ 15:43:37 zs_pseudo: 214500 cells; r mean_cp10k T 0.690/0.308, B 0.596/0.261, Mono 0.701/0.509, NK 0.730/0.410 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (10 s)
68
+ 15:43:38 fetching generated/imperial_eskd/fewshot_n5/draw4/zs__cal.npz (5.34 GiB) from spark-0110
69
+ 15:44:36 zs_cal: 214500 cells; r mean_cp10k T 0.691/0.376, B 0.652/0.331, Mono 0.709/0.513, NK 0.714/0.422 (Pearson/Spearman); repro PASS (max |d| 8.3e-17) (58 s)
70
+ 15:44:36 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/draw4/zs__cal.npz
71
+ 15:44:36 fetching generated/imperial_eskd/fewshot_n5/draw4/lora__raw.npz (5.23 GiB) from spark-0110
72
+ 15:45:33 lora5: 214500 cells; r mean_cp10k T 0.806/0.599, B 0.716/0.432, Mono 0.737/0.538, NK 0.737/0.458 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (57 s)
73
+ 15:45:33 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/draw4/lora__raw.npz
74
+ 15:45:34 fetching generated/imperial_eskd/fewshot_n5/draw4/scsemi5__raw.npz (0.15 GiB) from spark-0110
75
+ 15:45:55 scsemi5: 214500 cells; r mean_cp10k T 0.661/0.439, B 0.509/0.294, Mono 0.602/0.211, NK 0.614/0.362 (Pearson/Spearman); repro PASS (max |d| 8.3e-17) (22 s)
76
+ 15:45:55 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/draw4/scsemi5__raw.npz
77
+ 15:46:31 limma: 68 jobs in 36 s (0 not ok)
78
+ 15:46:31 unit fewshot_n5/draw4 of imperial_eskd -> /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/imperial_eskd/fewshot_n5/draw4 in 211.1 s; reproduction PASS
79
+ 15:46:31 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/zs__raw_all.npz
80
+ 15:46:31 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/fewshot_n5/zs__pseudo_all.npz
81
+ Command being timed: "/home/sj_server_1/miniforge3/envs/cxg_corpus/bin/python tools/contrast_extra.py --design fewshot --dataset imperial_eskd --draws 0,1,2,3,4 --threads 4"
82
+ User time (seconds): 824.51
83
+ System time (seconds): 210.19
84
+ Percent of CPU this job got: 87%
85
+ Elapsed (wall clock) time (h:mm:ss or m:ss): 19:42.85
86
+ Average shared text size (kbytes): 0
87
+ Average unshared data size (kbytes): 0
88
+ Average stack size (kbytes): 0
89
+ Average total size (kbytes): 0
90
+ Maximum resident set size (kbytes): 11180828
91
+ Average resident set size (kbytes): 0
92
+ Major (requiring I/O) page faults: 10
93
+ Minor (reclaiming a frame) page faults: 8136515
94
+ Voluntary context switches: 1282786
95
+ Involuntary context switches: 43321
96
+ Swaps: 0
97
+ File system inputs: 54667328
98
+ File system outputs: 140696136
99
+ Socket messages sent: 0
100
+ Socket messages received: 0
101
+ Signals delivered: 0
102
+ Page size (bytes): 4096
103
+ Exit status: 0
104
+ JOB_EXIT 0 1183s --design fewshot --dataset imperial_eskd --draws 0,1,2,3,4 --threads 4
105
+ JOB_START 2026-10-02T15:46:32+09:00 --design pooled --dataset imperial_eskd
106
+ 15:46:33 imperial_eskd pooled: 157 scored samples (113 case / 34 control), 5927 genes for the disease scores
107
+ 15:46:50 split half: 494503 observed cells of the 4 types, 5 seeds; mean r(A,B) T 0.970, B 0.781, Mono 0.900, NK 0.932 (16.7 s; full aggregates rebuilt = cache)
108
+ 15:46:51 fetching generated/imperial_eskd/pooled/pseudo_full__cal.npz (5.27 GiB) from spark-0110
109
+ 15:47:53 fetching generated/imperial_eskd/pooled/conditions_pooled_cal.npz (0.04 GiB) from spark-0110
110
+ 15:47:57 pseudo_full__cal: 235500 cells; r mean_cp10k T 0.885/0.729, B 0.749/0.488, Mono 0.808/0.599, NK 0.791/0.507 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (66 s)
111
+ 15:47:57 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/pooled/pseudo_full__cal.npz
112
+ 15:47:57 fetching generated/imperial_eskd/pooled/pseudo_full__raw.npz (5.27 GiB) from spark-0110
113
+ 15:49:00 fetching generated/imperial_eskd/pooled/conditions_pooled_raw.npz (0.04 GiB) from spark-0110
114
+ 15:49:02 pseudo_full__raw: 235500 cells; r mean_cp10k T 0.408/0.056, B 0.435/0.148, Mono 0.566/0.349, NK 0.467/0.168 (Pearson/Spearman); repro PASS (max |d| 9.5e-17) (65 s)
115
+ 15:49:03 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/pooled/pseudo_full__raw.npz
116
+ 15:49:03 fetching generated/imperial_eskd/pooled/real_lora__raw.npz (5.27 GiB) from spark-0110
117
+ 15:50:10 real_lora__raw: 235500 cells; r mean_cp10k T 0.903/0.791, B 0.755/0.516, Mono 0.788/0.607, NK 0.829/0.611 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (67 s)
118
+ 15:50:11 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/pooled/real_lora__raw.npz
119
+ 15:50:11 fetching generated/imperial_eskd/pooled/zs__cal.npz (5.27 GiB) from spark-0110
120
+ 15:51:19 zs__cal: 235500 cells; r mean_cp10k T 0.679/0.272, B 0.620/0.322, Mono 0.711/0.512, NK 0.705/0.392 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (68 s)
121
+ 15:51:20 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/pooled/zs__cal.npz
122
+ 15:51:20 fetching generated/imperial_eskd/pooled/zs__pseudo.npz (5.27 GiB) from spark-0110
123
+ 15:52:25 fetching generated/imperial_eskd/pooled/conditions_pooled_pseudo.npz (0.04 GiB) from spark-0110
124
+ 15:52:29 zs__pseudo: 235500 cells; r mean_cp10k T 0.700/0.290, B 0.586/0.253, Mono 0.710/0.525, NK 0.730/0.384 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (69 s)
125
+ 15:52:29 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/pooled/zs__pseudo.npz
126
+ 15:52:29 fetching generated/imperial_eskd/pooled/zs__raw.npz (5.27 GiB) from spark-0110
127
+ 15:53:41 zs__raw: 235500 cells; r mean_cp10k T 0.619/0.451, B 0.607/0.436, Mono 0.710/0.514, NK 0.640/0.454 (Pearson/Spearman); repro PASS (max |d| 5.6e-17) (72 s)
128
+ 15:53:41 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/pooled/zs__raw.npz
129
+ 15:53:41 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/pooled/conditions_pooled_cal.npz
130
+ 15:53:41 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/pooled/conditions_pooled_raw.npz
131
+ 15:53:41 removed fetched copy /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/_fetch_tmp/generated/imperial_eskd/pooled/conditions_pooled_pseudo.npz
132
+ 15:54:34 limma: 76 jobs in 53 s (0 not ok)
133
+ 15:54:34 unit pooled of imperial_eskd -> /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores_contrast/imperial_eskd/pooled in 481.6 s; reproduction PASS
134
+ Command being timed: "/home/sj_server_1/miniforge3/envs/cxg_corpus/bin/python tools/contrast_extra.py --design pooled --dataset imperial_eskd --threads 4"
135
+ User time (seconds): 320.32
136
+ System time (seconds): 104.56
137
+ Percent of CPU this job got: 88%
138
+ Elapsed (wall clock) time (h:mm:ss or m:ss): 8:02.69
139
+ Average shared text size (kbytes): 0
140
+ Average unshared data size (kbytes): 0
141
+ Average stack size (kbytes): 0
142
+ Average total size (kbytes): 0
143
+ Maximum resident set size (kbytes): 10273364
144
+ Average resident set size (kbytes): 0
145
+ Major (requiring I/O) page faults: 0
146
+ Minor (reclaiming a frame) page faults: 1842051
147
+ Voluntary context switches: 463461
148
+ Involuntary context switches: 16108
149
+ Swaps: 0
150
+ File system inputs: 513432
151
+ File system outputs: 67068448
152
+ Socket messages sent: 0
153
+ Socket messages received: 0
154
+ Signals delivered: 0
155
+ Page size (bytes): 4096
156
+ Exit status: 0
157
+ JOB_EXIT 0 482s --design pooled --dataset imperial_eskd --threads 4
158
+ UNIT_EXIT 0
paper-releases/pbisc-diffusion/updates/20261002T085943Z-deck-restructure-contrast-metrics-zeroshot-len/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/scores_contrast/_logs/sepsis_20261002T152649.jobs ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ --design fewshot --dataset ye2026_sepsis --draws 0,1,2,3,4
2
+ --design pooled --dataset ye2026_sepsis
3
+ --design scored --dataset ye2026_sepsis --score-dir /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/ye2026_sepsis/cal_budget_20261001/pooled --arms zs__cal_n5,zs__cal_n10,zs__cal_n20,zs__cal_n40 --unit cal_budget_20261001/pooled