tahamajs/RR / benchmark_logs /benchmark_huffman_coder.log
tahamajs's picture
download
raw
2.89 kB
================================================================================
STARTING UNIFIED 4-MODEL HARDENED SAMBA-MOE BENCHMARK ON GPU SERVER
Running models: ['distilbert', 'roberta', 'gpt2']
================================================================================
============================== RUNNING MODEL: distilbert ==============================
File "D:\svd_project\Codes\framework\run_quality_sweep.py", line 1
<<<<<<< HEAD
^^
SyntaxError: invalid syntax
>>> FINISHED MODEL: distilbert in 0.1s (Exit: 1, Status: PASS)
============================== RUNNING MODEL: roberta ==============================
File "D:\svd_project\Codes\framework\run_quality_sweep.py", line 1
<<<<<<< HEAD
^^
SyntaxError: invalid syntax
>>> FINISHED MODEL: roberta in 0.1s (Exit: 1, Status: PASS_LOW)
============================== RUNNING MODEL: gpt2 ==============================
File "D:\svd_project\Codes\framework\run_quality_sweep.py", line 1
<<<<<<< HEAD
^^
SyntaxError: invalid syntax
>>> FINISHED MODEL: gpt2 in 0.1s (Exit: 1, Status: PASS_LOW)
================================================================================
4-MODEL BENCHMARK RESULTS SUMMARY (SambaMoE on GPU)
================================================================================
Model Exit Code Baseline Acc SambaMoE Acc Delta Acc F1 Status Compressed Layers Comp Rate Peak VRAM Time (s)
distilbert 1 90.94% 90.60% -0.34% 0.9095 PASS 35 1.00x 308.8 MB 0.1
roberta 1 75.46% 51.38% -24.08% 0.6768 PASS_LOW 72 1.00x 528.9 MB 0.1
gpt2 1 57.91% 50.92% -6.99% 0.6748 PASS_LOW 0 1.00x 634.9 MB 0.1
================================================================================
Saved consolidated summaries to:
- results\unified_huffman_coder_4models_summary.csv
- results\unified_huffman_coder_4models_summary.json
Generated and saved comparative visualization charts to results\unified_baseline
================================================================================
STAGE 2: QUALITY-GATED EXPORT TO HUGGING FACE & GITHUB (SambaMoE)
================================================================================
[Quality Gate: PASSED] distilbert passed criteria (status=PASS, acc=90.60% >= 56.0%). Packaging for export...
Uploading to Hugging Face: Hooshaai/svd-linear-attention-distilbert-huffman_coder...
Successfully uploaded distilbert to Hooshaai/svd-linear-attention-distilbert-huffman_coder
[Quality Gate: REJECTED] roberta did not meet criteria (status=PASS_LOW, acc=51.38% < 56.0%).
[Quality Gate: REJECTED] gpt2 did not meet criteria (status=PASS_LOW, acc=50.92% < 56.0%).
Export Complete: 1/3 passed quality gate.

Xet Storage Details

Size:
2.89 kB
·
Xet hash:
ad5f1cba73f7b598f3d412aaf39c6154f37da53733da4b204762f5cf7cc0c493

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.