chenyu
a753c8e071
examples of new GPT2 and JIT change ( #2261 )
...
* var_vals are global
* working with global ish
* better
* fix export model
* fix tests
* better kv cache
* does it run?
* use where for kvmask
* fix excessive var_vals
* fix import
* how does multigpu use this?
* llama kinda work
* faster and simpler
* cleanup
* fix conversation mode
* test cleanups
* fix one more test
* test cleanup
---------
Co-authored-by: George Hotz <geohot@gmail.com >
2023-11-10 15:07:02 -05:00
George Hotz
de5d603ec1
corealize + remove realize from lazybuffer ( #1968 )
...
* corealize + remove realize from lazybuffer
* fix multigpu
* fix graph
2023-10-04 10:59:31 -07:00
George Hotz
d449b3bef1
think about removing realize from lazybuffer ( #1965 )
...
* remove realize from lazybuffer
* okay fine, back that off
* fix tests maybe
* fix test
2023-10-04 07:18:58 -07:00
George Hotz
6a79d4044a
unrealized consts everywhere ( #1963 )
...
* unrealized consts everywhere
* don't import device from lazy
* Device isn't in Lazy
* same issue
* disable jit random
2023-10-04 01:48:10 -07:00
nimlgen
f04c1a63ae
Rand works in jit ( #1960 )
...
* rand works in jit
* better jitted rand creation
* Update realize.py
---------
Co-authored-by: George Hotz <72895+geohot@users.noreply.github.com >
2023-10-03 12:55:25 -07:00
George Hotz
73a6ed7862
Apply ShapeTracker in interpreted backends ( #1846 )
...
* applying st
* tests pass
* minor cleanups
* torch too
* hack
* contiguous
* move mops
* contig in BN
* tests should pass
* make torch fast
* make zeros and ones contig by default
* no contig there
* fix padding with expanding
* might fix tests
* still doesn't fix bug, but should be there
* Revert "still doesn't fix bug, but should be there"
This reverts commit 8ea92f3e07
2023-09-23 10:05:13 +08:00
nimlgen
1c0449e190
add cache collector ( #1595 )
...
* init cache collector
* add test_cache_collector.py
* switch GlobalCounters.cache to CacheCollector
* init jit models test
* jitted SD
* add debug msg to print loaded bufs count
* moved cache collctor to jit
* clearer SD
* no double device import
2023-08-28 19:59:55 -07:00
chenyu
a89142e46f
ShapeTracker.var_vals ( #1540 )
2023-08-14 18:53:37 -07:00
George Hotz
d24f936501
just cmplt ( #1493 )
...
* just cmplt
* fix maximum
* don't save, there's no backward
* ugh, no slot either
* eq is a scam
2023-08-08 13:58:10 -07:00
nimlgen
669b406ec6
correct children count with lazycache ( #1429 )
2023-08-05 00:30:16 -07:00
chenyu
18d0a93f09
LazyBuffer.get_variable_buffers() ( #1391 )
...
* LazyBudder.get_variable_buffers()
* remove left_only, add ProdNode
* no vars for OpNode.b
* do not change symbolic vars, remove ProdNode
2023-08-02 09:01:35 -07:00
Francis Lam
df86672bd4
Fix LazyBuffer SHUFFLE_PAD_OPS to prevent invalid pad movement ( #1223 )
...
In addition to div, any ops that will generate non-zero outputs from
zero inputs need to be guarded.
2023-07-11 15:30:35 -07:00
George Hotz
0ad99038ef
Revert "Revert "Fix ShapeTracker mismatch in LazyBuffer.fromCPU ( #1156 )" ( #1181 )" + add test
...
This reverts commit a374b62bfe
2023-07-07 18:37:04 -07:00
George Hotz
a374b62bfe
Revert "Fix ShapeTracker mismatch in LazyBuffer.fromCPU ( #1156 )" ( #1181 )
...
This reverts commit 8ff7184b1b
2023-07-07 18:29:05 -07:00
fluffy χατγιρλ
8ff7184b1b
Fix ShapeTracker mismatch in LazyBuffer.fromCPU ( #1156 )
...
* init shape tracker with strides to fix mismatch
Author: sekstini <sekstinilol@gmail.com >
* fix whitespace
* add tests
2023-07-07 18:28:21 -07:00
