Fix: Save model locally before push, add explicit push with error handling
a823edb
slivk
commited on
Spaces:
Paused
Paused
Commit History
Add manual push button and push_model.py script
663a4dd
slivk
commited on
fix: OOM - reduce batch size to 1, disable eval in test, reduce max_length to 1024
e225440
slivk
commited on
fix: Disable ALL mixed precision + use fp32 compute (T4 bf16 crash fix)
cc9cee1
slivk
commited on
fix: Add torch_dtype=float16 to model loading (T4 bf16 compatibility)
3d53220
slivk
commited on
Fix: Force library upgrade and ensure float16
02bc115
slivk
commited on
Fix: Remove explicit bnb_compute_dtype
3db16e4
slivk
commited on
Fix: Correct bnb_compute_dtype to float16 for T4 compatibility
42b05ed
slivk
commited on
docs: Update README to reflect T4 GPU (not Zero GPU)
641aaa0
slivk
commited on
fix: Explicitly disable bf16 for T4 GPU compatibility
cbaf615
slivk
commited on
fix: Add bitsandbytes>=0.41.0
da22234
slivk
commited on
fix: Remove @spaces.GPU decorator for paid GPU
c17160f
slivk
commited on
feat: Add Qwen2.5-3B training (test + full modes)
bd3c7c9
slivk
commited on