Archive 4 checkpoints (1-4)

paper-releases/pbisc-diffusion/updates/20261013T072655Z-enriched-140-slide-master/diffusion/dataset/server_1/model8/presentations/2026-09-30--paper-summary--r00/data/figure_enrichment_20261013/render.log

@@ -0,0 +1,244 @@
+ 7  right fig  9.20x4.86 in  Experimental overview
+ 8  below fig 12.45x2.62 in  Pretraining corpus
+ 9  below fig 12.38x5.33 in  Corpus scale and generation workflow
+10  none  fig 12.45x4.07 in  From Gaussian noise to generated cell states
+13  none  fig 12.45x5.67 in  Where context width and trunk width change
+14  none  fig 12.45x5.67 in  Predicted and oracle hints: one trained denoiser
+15  none  fig 12.45x5.67 in  Architecture: full and late validation progression
+16  below fig  9.57x4.63 in  Width & context ablation: benchmark
+17  none  fig 12.45x5.67 in  Architecture arms: dimensions and endpoints
+18  below fig  9.57x4.63 in  Gene context: 2k / 3k / 6k / 15k
+19  none  fig 12.45x5.67 in  Gene panels: progression relative to each endpoint
+20  none  fig 12.45x5.67 in  Gene-panel arms: exact sizes and training provenance
+21  none  fig 12.45x5.67 in  All ablation arms: parameters and training time
+22  none  fig 12.15x5.70 in  Real-cell repeat: a sampling reference
+23  none  fig 12.45x5.47 in  Native-pseudobulk model comparisons
+26  none  fig 11.76x5.70 in  Bulk and pooled cells: different expression values
+27  none  fig 12.45x5.18 in  Paired calibration reduces same-donor expression error
+28  below fig 10.47x4.63 in  Consequence: cells off-manifold
+29  none  fig 12.15x5.70 in  Healthy paired data: full FT and LoRA
+30  none  fig 12.45x5.40 in  Full FT and LoRA ranks: quality and parameters
+31  right fig  9.20x5.63 in  Healthy paired-data budget: n training + 7 validation
+32  right fig  9.20x5.17 in  Five pairs, no extra validation: LoRA vs scSemi
+33  none  fig 12.17x5.70 in  Composition: error and donor-level wins
+36  right fig  9.20x5.18 in  Disease signals with five training donors
+37  right fig  9.20x4.80 in  5-fold CV: disease signal
+38  below fig 11.93x4.63 in  Disease genes at every list length (CAT)
+39  below fig 10.54x4.63 in  Cell-type deconvolution, same pairs: disease signal
+47  below fig 12.45x4.61 in  Backbone: SDD → PBISC-Diffusion
+48  below fig 12.45x4.61 in  Bulk conditioning
+49  below fig 11.68x5.33 in  Denoiser: joint value + support diffusion
+50  below fig 12.45x4.71 in  Fine-tuning with paired bulk + sc
+51  below fig 12.35x4.63 in  Evaluation design
+52  below fig 11.19x4.63 in  How we score generated cells
+53  none  fig 12.45x5.31 in  Does the model follow its bulk conditioning?
+54  below fig 10.30x4.63 in  SLE benchmark: PBISC vs scSemiProfiler
+55  below fig 10.79x4.63 in  SLE benchmark: per cell type
+56  none  fig 12.45x5.38 in  Disjoint-cell controls test the bulk-broadcast baseline
+57  none  fig 12.45x5.60 in  SLE: what is added beyond the bulk broadcast?
+58  below fig 12.45x5.24 in  Same shift in three paired cohorts
+59  none  fig 12.45x5.64 in  Input corrections on one real-reference UMAP
+60  none  fig 12.45x4.86 in  Where generated-cell density differs from real cells
+61  none  fig 12.45x4.89 in  Input calibration changes cell-type composition
+62  none  fig 12.45x5.63 in  Same noise, different input: matched-cell shifts
+63  none  fig 12.45x4.76 in  Matched noise separates cell state from cell identity
+64  none  fig 12.45x5.63 in  Input changes can switch the generated cell label
+65  none  fig 12.45x5.58 in  Nearest real cells: distance and label agreement
+66  none  fig 12.45x5.52 in  Composition agreement does not ensure state agreement
+67  none  fig 12.45x5.47 in  Gene-length effects in the bulk input
+68  none  fig 12.45x5.47 in  Gene-length effects within label-stable cells
+69  none  fig 12.45x5.41 in  Raw and TPM shifts persist within cell labels
+70  none  fig 12.40x5.70 in  Reference calibration leaves specific gene-state shifts
+71  none  fig 12.45x5.45 in  Donor trait axes and donor identification differ
+72  none  fig 12.45x4.61 in  Reference choice changes different quality endpoints
+74  none  fig 12.45x4.72 in  Technical repeats and follow-up are different controls
+75  none  fig 12.45x5.32 in  Annotation agreement and difficult categories
+76  none  fig 12.45x5.39 in  Depth matching changes precision and coverage
+77  none  fig 12.45x4.87 in  Detection and library size are separate diagnostics
+78  right fig  9.20x5.18 in  Healthy cohort: LoRA cells cover the real ones
+79  right fig  9.20x4.79 in  Healthy cohort: realism beyond precision/coverage
+80  right fig  9.20x5.14 in  Healthy cohort: full vs LoRA vs condition encoder
+81  below fig  9.37x4.63 in  Representative-dependent composition
+82  below fig 11.39x4.63 in  Full vs LoRA vs cond.-encoder-only
+83  right fig  9.20x5.09 in  Healthy cohort: LoRA rank and K
+84  right fig  9.20x5.09 in  Healthy paired-data learning curves
+85  right fig  9.20x4.66 in  Time: days for scSemiProfiler, minutes for LoRA
+86  below fig 10.77x4.63 in  5 donors: time to apply
+87  below fig  9.99x4.63 in  Pairs needed: composition, coverage, training steps
+88  below fig 10.08x4.63 in  LoRA rank: r = 4, 16, 64
+89  right fig  8.31x5.70 in  K diversification: no gain
+90  below fig 12.45x3.82 in  Five paired cohorts
+91  none  fig 12.45x4.70 in  Sepsis: exact-five reconstruction across five draws
+92  none  fig 12.45x5.18 in  Sepsis: realism beyond precision and coverage
+93  none  fig 12.45x4.70 in  Imperial: first completed exact-five comparison
+94  below fig 10.01x4.63 in  5 donors: LoRA vs scSemiProfiler — cells
+95  below fig 10.04x4.63 in  5 donors: realism beyond precision/coverage
+96  below fig 10.02x4.63 in  5 donors: precision/coverage at k = 3, 5, 10
+97  below fig 10.09x4.63 in  5-fold: realism beyond precision/coverage
+98  below fig 11.74x4.63 in  Same training pairs: LoRA vs scSemiProfiler
+99  below fig 11.85x4.63 in  scSemiProfiler and the representative-copy baseline
+100  none  fig 10.46x5.70 in  Sepsis: patient-aware disease-gene agreement
+101  none  fig 10.46x5.70 in  Imperial: patient-aware disease-gene agreement
+102  below fig 10.60x4.63 in  Sepsis: fine-tuning variants
+103  right fig  8.44x5.70 in  Pairs needed: donor and disease signal
+104  below fig 10.53x4.63 in  Cell-type deconvolution, same pairs: composition
+105  none  fig 12.45x4.94 in  Deconvolution: expression coverage limits the comparison
+106  none  fig 12.45x5.08 in  K=4,000 recipe: paired reconstruction quality
+107  none  fig 12.45x5.08 in  K=4,000 recipe: disease signal does not improve uniformly
+108  none  fig 12.45x5.22 in  K=4,000 recipe: calibration gains are selective
+109  none  fig 12.45x4.66 in  Influenza: response order versus response genes
+110  none  fig 10.30x5.70 in  COVID: calibration on a shared real-cell map
+111  none  fig 10.82x5.70 in  COVID monocytes: direction after depth matching
+112  none  fig 11.44x5.70 in  COVID monocytes: direction, magnitude and null
+113  none  fig 11.44x5.70 in  COVID CD14 monocytes: direction, magnitude and null
+114  none  fig 12.45x5.02 in  COVID: biological endpoints give a mixed verdict
+115  none  fig 11.83x5.70 in  Rare-state frequencies in observed cells
+116  none  fig 11.35x5.70 in  Rare-state misses: rules 1–8
+117  none  fig 11.72x5.70 in  Rare-state misses: rules 9–15
+118  none  fig 12.06x5.70 in  Rare-state fidelity: zero-aware reanalysis
+119  none  fig 10.76x5.70 in  Sepsis rare states across paired-data budgets
+120  none  fig 12.45x5.62 in  Sepsis rare-state direction across support budgets
+121  none  fig 11.35x5.70 in  Rare-state screen in DRESS, Kawasaki and SLE
+122  right fig  7.24x5.70 in  Consequence: disease direction flips
+123  none  fig 12.45x5.18 in  C1Q: single-gene detection versus co-detection
+124  none  fig 12.45x4.55 in  Behçet: K=4,000 changes direction more than identity
+125  none  fig 12.45x5.59 in  Behcet: one-gene and multi-gene C1Q rules
+126  below fig  9.46x4.63 in  Pairs from other cohorts do not transfer
+127  below fig  9.44x4.63 in  Longer training on other cohorts: gap stays
+128  below fig 10.61x4.63 in  No pretraining: same cohort, from scratch
+129  below fig 10.53x4.63 in  No pretraining: other cohorts, from scratch
+130  below fig 10.20x4.63 in  Rare states: from scratch vs pretrained
+131  below fig 10.12x4.63 in  Appendix: scSemiProfiler without output threshold
+132  none  fig 12.45x5.28 in  Bulk-only exploration: calibration populations differ
+134  right fig  9.20x5.19 in  Rare-state fractions: original analysis
+135  right fig  9.20x5.01 in  Rare-state ranking: original analysis
+136  below fig  9.23x4.98 in  Rare states: original five-fold analysis
+137  right fig  8.01x5.70 in  Bulk-only cohort plan: original snapshot
+138  none  fig 12.45x4.73 in  Historical bulk-only search: no confirmed new finding
+140  below fig 12.45x4.62 in  Next: other tissues from a few dozen pairs
+{"slides": 140, "groups": {"Introduction": [1, 2, 3, 4], "Figure 1": [5, 6, 7, 8, 9, 10], "Figure 2": [11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23], "Figure 3": [24, 25, 26, 27, 28, 29, 30, 31, 32, 33], "Figure 4": [34, 35, 36, 37, 38, 39], "Figure 5": [40, 41], "Figure 6": [42, 43, 44], "Supplementary": [45, 46], "Supp S01": [47, 48, 49, 50, 51, 52], "Supp S02": [53, 54, 55, 56, 57], "Supp S03": [58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72], "Supp S04": [73, 74, 75, 76, 77], "Supp S05": [78, 79, 80, 81, 82], "Supp S06": [83, 84, 85, 86, 87, 88, 89], "Supp S07": [90, 91, 92, 93, 94, 95, 96, 97, 98, 99], "Supp S08": [100, 101, 102, 103, 104, 105, 106, 107, 108], "Supp S09": [109, 110, 111, 112, 113, 114], "Supp S10": [115, 116, 117, 118, 119, 120], "Supp S11": [121, 122, 123, 124, 125], "Supp S12": [126, 127, 128, 129, 130], "Supp S13": [131], "Supp S14": [132], "Historical archive": [133, 134, 135, 136, 137, 138], "Discussion": [139, 140]}, "pptx": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/PBISC-Diffusion_expanded_master_20261013_enriched.pptx"}
+convert /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/PBISC-Diffusion_expanded_master_20261013_enriched.pptx as a Impress document -> /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/PBISC-Diffusion_expanded_master_20261013_enriched.pdf using filter : impress_pdf_Export:{"UseTaggedPDF":{"type":"boolean","value":"false"},"PDFUACompliance":{"type":"boolean","value":"false"}}
+Overwriting: /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/PBISC-Diffusion_expanded_master_20261013_enriched.pdf
+{"pass": true, "counts": {"spec": 140, "pptx": 140, "pdf": 140}, "failures": [], "contacts": 12}
+ 7  right fig  9.20x4.86 in  Experimental overview
+ 8  below fig 12.45x2.62 in  Pretraining corpus
+ 9  below fig 12.38x5.33 in  Corpus scale and generation workflow
+10  none  fig 12.45x4.07 in  From Gaussian noise to generated cell states
+13  none  fig 12.45x5.67 in  Where context width and trunk width change
+14  none  fig 12.45x5.67 in  Predicted and oracle hints: one trained denoiser
+15  none  fig 12.45x5.67 in  Architecture: full and late validation progression
+16  below fig  9.57x4.63 in  Width & context ablation: benchmark
+17  none  fig 12.45x5.67 in  Architecture arms: dimensions and endpoints
+18  below fig  9.57x4.63 in  Gene context: 2k / 3k / 6k / 15k
+19  none  fig 12.45x5.67 in  Gene panels: progression relative to each endpoint
+20  none  fig 12.45x5.67 in  Gene-panel arms: exact sizes and training provenance
+21  none  fig 12.45x5.67 in  All ablation arms: parameters and training time
+22  none  fig 12.15x5.70 in  Real-cell repeat: a sampling reference
+23  none  fig 12.45x5.47 in  Native-pseudobulk model comparisons
+26  none  fig 11.76x5.70 in  Bulk and pooled cells: different expression values
+27  none  fig 12.45x5.18 in  Paired calibration reduces same-donor expression error
+28  below fig 10.47x4.63 in  Consequence: cells off-manifold
+29  none  fig 12.15x5.70 in  Healthy paired data: full FT and LoRA
+30  none  fig 12.45x5.40 in  Full FT and LoRA ranks: quality and parameters
+31  right fig  9.20x5.63 in  Healthy paired-data budget: n training + 7 validation
+32  right fig  9.20x5.17 in  Five pairs, no extra validation: LoRA vs scSemi
+33  none  fig 12.17x5.70 in  Composition: error and donor-level wins
+36  right fig  9.20x5.18 in  Disease signals with five training donors
+37  right fig  9.20x4.80 in  5-fold CV: disease signal
+38  below fig 11.93x4.63 in  Disease genes at every list length (CAT)
+39  below fig 10.54x4.63 in  Cell-type deconvolution, same pairs: disease signal
+47  below fig 12.45x4.61 in  Backbone: SDD → PBISC-Diffusion
+48  below fig 12.45x4.61 in  Bulk conditioning
+49  below fig 11.68x5.33 in  Denoiser: joint value + support diffusion
+50  below fig 12.45x4.71 in  Fine-tuning with paired bulk + sc
+51  below fig 12.35x4.63 in  Evaluation design
+52  below fig 11.19x4.63 in  How we score generated cells
+53  none  fig 12.45x5.31 in  Does the model follow its bulk conditioning?
+54  below fig 10.30x4.63 in  SLE benchmark: PBISC vs scSemiProfiler
+55  below fig 10.79x4.63 in  SLE benchmark: per cell type
+56  none  fig 12.45x5.38 in  Disjoint-cell controls test the bulk-broadcast baseline
+57  none  fig 12.45x5.60 in  SLE: what is added beyond the bulk broadcast?
+58  below fig 12.45x5.24 in  Same shift in three paired cohorts
+59  none  fig 12.45x5.64 in  Input corrections on one real-reference UMAP
+60  none  fig 12.45x4.86 in  Where generated-cell density differs from real cells
+61  none  fig 12.45x4.89 in  Input calibration changes cell-type composition
+62  none  fig 12.45x5.63 in  Same noise, different input: matched-cell shifts
+63  none  fig 12.45x4.76 in  Matched noise separates cell state from cell identity
+64  none  fig 12.45x5.63 in  Input changes can switch the generated cell label
+65  none  fig 12.45x5.58 in  Nearest real cells: distance and label agreement
+66  none  fig 12.34x5.70 in  Composition agreement does not ensure state agreement
+67  none  fig 12.45x5.47 in  Gene-length effects in the bulk input
+68  none  fig 12.45x5.47 in  Gene-length effects within label-stable cells
+69  none  fig 12.45x5.41 in  Raw and TPM shifts persist within cell labels
+70  none  fig 12.40x5.70 in  Reference calibration leaves specific gene-state shifts
+71  none  fig 12.45x5.45 in  Donor trait axes and donor identification differ
+72  none  fig 12.45x4.61 in  Reference choice changes different quality endpoints
+74  none  fig 12.45x4.72 in  Technical repeats and follow-up are different controls
+75  none  fig 12.45x5.32 in  Annotation agreement and difficult categories
+76  none  fig 12.45x5.39 in  Depth matching changes precision and coverage
+77  none  fig 12.45x4.87 in  Detection and library size are separate diagnostics
+78  right fig  9.20x5.18 in  Healthy cohort: LoRA cells cover the real ones
+79  right fig  9.20x4.79 in  Healthy cohort: realism beyond precision/coverage
+80  right fig  9.20x5.14 in  Healthy cohort: full vs LoRA vs condition encoder
+81  below fig  9.37x4.63 in  Representative-dependent composition
+82  below fig 11.39x4.63 in  Full vs LoRA vs cond.-encoder-only
+83  right fig  9.20x5.09 in  Healthy cohort: LoRA rank and K
+84  right fig  9.20x5.09 in  Healthy paired-data learning curves
+85  right fig  9.20x4.66 in  Time: days for scSemiProfiler, minutes for LoRA
+86  below fig 10.77x4.63 in  5 donors: time to apply
+87  below fig  9.99x4.63 in  Pairs needed: composition, coverage, training steps
+88  below fig 10.08x4.63 in  LoRA rank: r = 4, 16, 64
+89  right fig  8.31x5.70 in  K diversification: no gain
+90  below fig 12.45x3.82 in  Five paired cohorts
+91  none  fig 12.45x4.70 in  Sepsis: exact-five reconstruction across five draws
+92  none  fig 12.45x5.18 in  Sepsis: realism beyond precision and coverage
+93  none  fig 12.45x4.70 in  Imperial: first completed exact-five comparison
+94  below fig 10.01x4.63 in  5 donors: LoRA vs scSemiProfiler — cells
+95  below fig 10.04x4.63 in  5 donors: realism beyond precision/coverage
+96  below fig 10.02x4.63 in  5 donors: precision/coverage at k = 3, 5, 10
+97  below fig 10.09x4.63 in  5-fold: realism beyond precision/coverage
+98  below fig 11.74x4.63 in  Same training pairs: LoRA vs scSemiProfiler
+99  below fig 11.85x4.63 in  scSemiProfiler and the representative-copy baseline
+100  none  fig 10.46x5.70 in  Sepsis: patient-aware disease-gene agreement
+101  none  fig 10.46x5.70 in  Imperial: patient-aware disease-gene agreement
+102  below fig 10.60x4.63 in  Sepsis: fine-tuning variants
+103  right fig  8.44x5.70 in  Pairs needed: donor and disease signal
+104  below fig 10.53x4.63 in  Cell-type deconvolution, same pairs: composition
+105  none  fig 12.45x4.94 in  Deconvolution: expression coverage limits the comparison
+106  none  fig 12.45x5.08 in  K=4,000 recipe: paired reconstruction quality
+107  none  fig 12.45x5.08 in  K=4,000 recipe: disease signal does not improve uniformly
+108  none  fig 12.45x5.22 in  K=4,000 recipe: calibration gains are selective
+109  none  fig 12.45x4.66 in  Influenza: response order versus response genes
+110  none  fig 10.30x5.70 in  COVID: calibration on a shared real-cell map
+111  none  fig 10.82x5.70 in  COVID monocytes: direction after depth matching
+112  none  fig 11.44x5.70 in  COVID monocytes: direction, magnitude and null
+113  none  fig 11.44x5.70 in  COVID CD14 monocytes: direction, magnitude and null
+114  none  fig 12.45x5.02 in  COVID: biological endpoints give a mixed verdict
+115  none  fig 11.83x5.70 in  Rare-state frequencies in observed cells
+116  none  fig 11.35x5.70 in  Rare-state misses: rules 1–8
+117  none  fig 11.72x5.70 in  Rare-state misses: rules 9–15
+118  none  fig 12.06x5.70 in  Rare-state fidelity: zero-aware reanalysis
+119  none  fig 10.76x5.70 in  Sepsis rare states across paired-data budgets
+120  none  fig 12.45x5.62 in  Sepsis rare-state direction across support budgets
+121  none  fig 11.35x5.70 in  Rare-state screen in DRESS, Kawasaki and SLE
+122  right fig  7.24x5.70 in  Consequence: disease direction flips
+123  none  fig 12.45x5.18 in  C1Q: single-gene detection versus co-detection
+124  none  fig 12.45x4.55 in  Behçet: K=4,000 changes direction more than identity
+125  none  fig 12.45x5.59 in  Behcet: one-gene and multi-gene C1Q rules
+126  below fig  9.46x4.63 in  Pairs from other cohorts do not transfer
+127  below fig  9.44x4.63 in  Longer training on other cohorts: gap stays
+128  below fig 10.61x4.63 in  No pretraining: same cohort, from scratch
+129  below fig 10.53x4.63 in  No pretraining: other cohorts, from scratch
+130  below fig 10.20x4.63 in  Rare states: from scratch vs pretrained
+131  below fig 10.12x4.63 in  Appendix: scSemiProfiler without output threshold
+132  none  fig 12.45x5.28 in  Bulk-only exploration: calibration populations differ
+134  right fig  9.20x5.19 in  Rare-state fractions: original analysis
+135  right fig  9.20x5.01 in  Rare-state ranking: original analysis
+136  below fig  9.23x4.98 in  Rare states: original five-fold analysis
+137  right fig  8.01x5.70 in  Bulk-only cohort plan: original snapshot
+138  none  fig 12.45x4.73 in  Historical bulk-only search: no confirmed new finding
+140  below fig 12.45x4.62 in  Next: other tissues from a few dozen pairs
+{"slides": 140, "groups": {"Introduction": [1, 2, 3, 4], "Figure 1": [5, 6, 7, 8, 9, 10], "Figure 2": [11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23], "Figure 3": [24, 25, 26, 27, 28, 29, 30, 31, 32, 33], "Figure 4": [34, 35, 36, 37, 38, 39], "Figure 5": [40, 41], "Figure 6": [42, 43, 44], "Supplementary": [45, 46], "Supp S01": [47, 48, 49, 50, 51, 52], "Supp S02": [53, 54, 55, 56, 57], "Supp S03": [58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72], "Supp S04": [73, 74, 75, 76, 77], "Supp S05": [78, 79, 80, 81, 82], "Supp S06": [83, 84, 85, 86, 87, 88, 89], "Supp S07": [90, 91, 92, 93, 94, 95, 96, 97, 98, 99], "Supp S08": [100, 101, 102, 103, 104, 105, 106, 107, 108], "Supp S09": [109, 110, 111, 112, 113, 114], "Supp S10": [115, 116, 117, 118, 119, 120], "Supp S11": [121, 122, 123, 124, 125], "Supp S12": [126, 127, 128, 129, 130], "Supp S13": [131], "Supp S14": [132], "Historical archive": [133, 134, 135, 136, 137, 138], "Discussion": [139, 140]}, "pptx": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/PBISC-Diffusion_expanded_master_20261013_enriched.pptx"}
+convert /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/PBISC-Diffusion_expanded_master_20261013_enriched.pptx as a Impress document -> /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/PBISC-Diffusion_expanded_master_20261013_enriched.pdf using filter : impress_pdf_Export:{"UseTaggedPDF":{"type":"boolean","value":"false"},"PDFUACompliance":{"type":"boolean","value":"false"}}
+Overwriting: /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/PBISC-Diffusion_expanded_master_20261013_enriched.pdf
+{"pass": true, "counts": {"spec": 140, "pptx": 140, "pdf": 140}, "failures": [], "contacts": 12}

paper-releases/pbisc-diffusion/updates/20261013T072655Z-enriched-140-slide-master/diffusion/dataset/server_1/model8/presentations/2026-09-30--paper-summary--r00/data/figure_enrichment_20261013/render.py

@@ -0,0 +1,27 @@
+"""Rebuild from saved enrichment assets, export PDF and validate (CPU only).
+
+Invoke within a bounded user-systemd unit with CUDA_VISIBLE_DEVICES empty.
+This does not execute training, sampling or any scientific figure builder.
+"""
+from pathlib import Path
+import os
+import subprocess
+import sys
+
+if os.environ.get('CUDA_VISIBLE_DEVICES')!='':
+    raise SystemExit('CPU-only export requires empty CUDA_VISIBLE_DEVICES')
+A=Path(__file__).resolve().parent
+D=A.parents[1]
+commands=[
+    [sys.executable,str(D/'tools/build_enriched_master_20261013.py')],
+    ['libreoffice','-env:UserInstallation=file:///tmp/pbisc-enriched-lo-20261013',
+     '--headless','--convert-to',
+     'pdf:impress_pdf_Export:{"UseTaggedPDF":{"type":"boolean","value":"false"},"PDFUACompliance":{"type":"boolean","value":"false"}}',
+     '--outdir',str(D),str(D/'PBISC-Diffusion_expanded_master_20261013_enriched.pptx')],
+    [sys.executable,str(D/'tools/verify_enriched_master_20261013.py')],
+]
+with (A/'render.log').open('a') as log:
+    for command in commands:
+        print('Run:',Path(command[0]).name,Path(command[-1]).name,flush=True)
+        subprocess.run(command,cwd=D,stdout=log,stderr=subprocess.STDOUT,check=True,timeout=600)
+print('PPTX, PDF and automated validation completed.',flush=True)