Commits · evgueni-p/fbmc-chronos2

fix: reduce context window to 1125 hours (1.5 months) for A100-80GB

d080539

Evgueni Poloukarov Claude commited on 26 days ago

fix: reduce context window to 1,800 hours (75 days) for A100-80GB VRAM limits

f7735d6

Evgueni Poloukarov Claude commited on 27 days ago

feat: extend context window to 3 months (2,160 hours) for 96GB VRAM

e1f5207

Evgueni Poloukarov commited on 29 days ago

fix: reduce context to 1,440 hours (60 days) - testing GPU memory limit

5ff6f25

Evgueni Poloukarov commited on about 1 month ago

fix: reduce context window to 1,792 hours (75 days) to fit 48GB GPU memory

8ff9099

Evgueni Poloukarov Claude commited on about 1 month ago

feat: extend context window to 3 months (2,160 hours) for deeper pattern learning

1130ee0

Evgueni Poloukarov Claude commited on about 1 month ago

docs: correct feature count (2,514 actual, not 3,043 claimed)

27ce714

Evgueni Poloukarov commited on Nov 21

fix: convert both context and future DataFrames to float64

2e13800

Evgueni Poloukarov Claude commited on Nov 21

fix: handle NaN in dtype alignment using pd.to_numeric()

0bb8c81

Evgueni Poloukarov Claude commited on Nov 21

ci: add version marker to force HF Space cache invalidation

d19f2b6

Evgueni Poloukarov commited on Nov 21

fix: correct feature category name (historical not historical_only)

a2c07c4

Evgueni Poloukarov commited on Nov 21

feat: implement past-only covariate masking for volatility capture

069627f

Evgueni Poloukarov Claude commited on Nov 21

fix: align dtypes between context and future DataFrames for Chronos-2

de602fd

Evgueni Poloukarov Claude commited on Nov 19

fix: revert context window to 512h and upgrade GPU to 1xL4 (24GB)

42acd7e

Evgueni Poloukarov Claude commited on Nov 16

fix: reduce context window to 256h to fit T4 GPU memory

dfe9aab

Evgueni Poloukarov commited on Nov 16

fix: correct future_df timestamp to start at last context timestamp

04dcb8d

Evgueni Poloukarov commited on Nov 16

feat: add dynamic forecast system to prevent data leakage

f4be780

Evgueni Poloukarov commited on Nov 13

Spaces:

evgueni-p
/

fbmc-chronos2

Sleeping

Commit History

fix: reduce context window to 1125 hours (1.5 months) for A100-80GB

d080539

fix: reduce context window to 1,800 hours (75 days) for A100-80GB VRAM limits

f7735d6

feat: extend context window to 3 months (2,160 hours) for 96GB VRAM

e1f5207

fix: reduce context to 1,440 hours (60 days) - testing GPU memory limit

5ff6f25

fix: reduce context window to 1,792 hours (75 days) to fit 48GB GPU memory

8ff9099

feat: extend context window to 3 months (2,160 hours) for deeper pattern learning

1130ee0

docs: correct feature count (2,514 actual, not 3,043 claimed)

27ce714

fix: convert both context and future DataFrames to float64

2e13800

fix: handle NaN in dtype alignment using pd.to_numeric()

0bb8c81

ci: add version marker to force HF Space cache invalidation

d19f2b6

fix: correct feature category name (historical not historical_only)

a2c07c4

feat: implement past-only covariate masking for volatility capture

069627f

fix: align dtypes between context and future DataFrames for Chronos-2

de602fd

fix: revert context window to 512h and upgrade GPU to 1xL4 (24GB)

42acd7e

fix: reduce context window to 256h to fit T4 GPU memory

dfe9aab

fix: correct future_df timestamp to start at last context timestamp

04dcb8d

feat: add dynamic forecast system to prevent data leakage

f4be780

Commit History

fix: reduce context window to 1125 hours (1.5 months) for A100-80GB d080539

fix: reduce context window to 1,800 hours (75 days) for A100-80GB VRAM limits f7735d6

feat: extend context window to 3 months (2,160 hours) for 96GB VRAM e1f5207

fix: reduce context to 1,440 hours (60 days) - testing GPU memory limit 5ff6f25

fix: reduce context window to 1,792 hours (75 days) to fit 48GB GPU memory 8ff9099

feat: extend context window to 3 months (2,160 hours) for deeper pattern learning 1130ee0

docs: correct feature count (2,514 actual, not 3,043 claimed) 27ce714

fix: convert both context and future DataFrames to float64 2e13800

fix: handle NaN in dtype alignment using pd.to_numeric() 0bb8c81

ci: add version marker to force HF Space cache invalidation d19f2b6

fix: correct feature category name (historical not historical_only) a2c07c4

feat: implement past-only covariate masking for volatility capture 069627f

fix: align dtypes between context and future DataFrames for Chronos-2 de602fd

fix: revert context window to 512h and upgrade GPU to 1xL4 (24GB) 42acd7e

fix: reduce context window to 256h to fit T4 GPU memory dfe9aab

fix: correct future_df timestamp to start at last context timestamp 04dcb8d

feat: add dynamic forecast system to prevent data leakage f4be780

fix: reduce context window to 1125 hours (1.5 months) for A100-80GB

d080539

fix: reduce context window to 1,800 hours (75 days) for A100-80GB VRAM limits

f7735d6

feat: extend context window to 3 months (2,160 hours) for 96GB VRAM

e1f5207

fix: reduce context to 1,440 hours (60 days) - testing GPU memory limit

5ff6f25

fix: reduce context window to 1,792 hours (75 days) to fit 48GB GPU memory

8ff9099

feat: extend context window to 3 months (2,160 hours) for deeper pattern learning

1130ee0

docs: correct feature count (2,514 actual, not 3,043 claimed)

27ce714

fix: convert both context and future DataFrames to float64

2e13800

fix: handle NaN in dtype alignment using pd.to_numeric()

0bb8c81

ci: add version marker to force HF Space cache invalidation

d19f2b6

fix: correct feature category name (historical not historical_only)

a2c07c4

feat: implement past-only covariate masking for volatility capture

069627f

fix: align dtypes between context and future DataFrames for Chronos-2

de602fd

fix: revert context window to 512h and upgrade GPU to 1xL4 (24GB)

42acd7e

fix: reduce context window to 256h to fit T4 GPU memory

dfe9aab

fix: correct future_df timestamp to start at last context timestamp

04dcb8d

feat: add dynamic forecast system to prevent data leakage

f4be780