picchio/.gitignore at main · benmaster82/picchio · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
# Binaries
*.exe
*.o
*.obj

# CUDA backend build artifacts (optional; rebuild with build_cuda.bat)
*.dll
*.lib
*.exp
pgpu_test.*

# Model directories (60+ GB)
/model-*.safetensors
/expert_biases.safetensors
test_model/
test_tiny_hf/
test_tiny_picchio/
test_tiny_picchio_i4/
test_tiny_hf_i4ref/
oracle_hf/
oracle_c/
oracle_boundary_hf/
oracle_boundary_c/
oracle_long_hf/
oracle_long_c/
gptoss_i4/
gptoss_raw/
gptoss20b_i4/
gptoss120b_meta/

# Temporary files
*.log
*.tmp
__pycache__/
do.sh
convert_minimax_retry.sh
.qwen_smoke_tmp/

# Hot-store (generated at runtime)
.picchio_usage
.picchio_kv

# OS
Thumbs.db
.DS_Store

# IDE
.vscode/
.idea/

# Development tools and one-off scripts, not published (local archive)
dev/

# S1 flat expert store artifacts
*.picchioflat
test_model_i3/
test_model_i4/

# MiniMax-M2: minimax_test_model/ is deliberately NOT ignored (~132 KB). Unlike
# test_model/ it is not reliably regenerable — the oracle is only correct on
# transformers < 5.0 (see the version guard in make_minimax_test_model.py), so
# the committed copy IS the reference.
#
# Its Picchio-named twin IS ignored: fuse_minimax_test_model.py rebuilds it
# byte-identically from the above, using only numpy + safetensors, so it carries
# none of that version hazard. Regenerate with:
#   python fuse_minimax_test_model.py minimax_test_model minimax_test_model_picchio
minimax_test_model_picchio/
minimax_test_model_deep/
minimax_test_model_deep_picchio/