| #578 |
fix: ensure non-zero learning rate during warmup at iteration 0
|
@silasalberti
|
16.2h
|
0 |
✓
|
| #463 |
Fixing eval path in README
|
@goswamig
|
2 months
|
0 |
✓
|
| #487 |
Update README.md
|
@jellehak
|
6 days
|
0 |
✓
|
| #420 |
Move enc to gloabal namespace to fix #371
|
@vinjn
|
1 month
|
0 |
✓
|
| #429 |
Open "shakespeare" data in UTF-8 in "prepare.py"
|
@adambala
|
1 month
|
0 |
✓
|
| #428 |
fix np.memmap memory leak
|
@kjslag
|
1 month
|
0 |
✓
|
| #309 |
Fix AssertionError on macOS - need to check CUDA availability for bf16
|
@ho2103
|
2 days
|
0 |
✓
|
| #305 |
nanogpt: fix multiprocessing in load_dataset on os x
|
@okuvshynov
|
44m
|
0 |
✓
|
| #301 |
[easy] allow multithreading in load_dataset
|
@okuvshynov
|
1m
|
0 |
✓
|
| #274 |
Use nn.GELU - 1.27x faster training
|
@apivovarov
|
28 days
|
0 |
✓
|
| #275 |
Remove pos unsqueeze(0)
|
@apivovarov
|
28 days
|
0 |
✓
|
| #270 |
fix np.sum overflows on windows
|
@LaihoE
|
1 month
|
0 |
✓
|
| #276 |
Add more files to .gitignore
|
@apivovarov
|
28 days
|
0 |
✓
|
| #277 |
Use bf16 only if supported
|
@apivovarov
|
28 days
|
0 |
✓
|
| #286 |
docs: simplify dependencies installation
|
@ctjlewis
|
13 days
|
0 |
✓
|
| #247 |
Macbook run instructions
|
@gnobre
|
2 days
|
0 |
✓
|
| #225 |
Fix for gradient_accumulation_steps training slow
|
@otaviogood
|
22 days
|
0 |
✓
|
| #10 |
batch file write
|
@LaihoE
|
3 months
|
0 |
✓
|
| #146 |
Add .gitignore
|
@lutzroeder
|
1 month
|
1 |
✓
|
| #74 |
Small fix to decode fn in shakespeare_char/prepare.py
|
@venusatuluri
|
2 months
|
0 |
✓
|
