TinyLamda commited on
Commit
e374a37
·
verified ·
1 Parent(s): 0fe4a2e

add w4a8_q3_14b/w4b_build.json

Browse files
w4a8_q3_14b/dc_qmodels/g_qoq_q3_14b/w4b_build.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "mk": "q3_14b",
3
+ "method": "qoq",
4
+ "dc_arm": "qoq",
5
+ "host": "host0",
6
+ "gpu_index": "6",
7
+ "gpu": "NVIDIA H200 NVL, 9.0, 580.178.04",
8
+ "time": "2026-11-07T12:44:19+09:00",
9
+ "dc_root": "/work/w4b-dc/deepcompressor",
10
+ "dc_head": "69f3473f5e1c1504bae35cc50c7858ef900a9b17",
11
+ "dc_diff_sha256": "e47e17b43c289bc3194bf0717397e451ef36607aac576dd8a74382484dacbf7b",
12
+ "w4b_overlay": true,
13
+ "ANON0_DC_ROT_CPU_B": "30",
14
+ "env": "/work/w4b-dc/.venv-dc-qwen3-w4b",
15
+ "env_versions": "3.11.15 2.9.1+cu128 4.57.3",
16
+ "tree_commit": "83ca9b35c5ab1cebfed47b11b14b7f1fa8169275",
17
+ "tree_dirty": false,
18
+ "dc_harness_s0": {
19
+ "qoq:0": {
20
+ "result": "/work/anon0-hpca-w4brun/sw/results/prune_rb_ppl_qoq_q3_14b_s0_w4b.json",
21
+ "log": "/work/anon0-hpca-w4brun/sw/logs/prune_rb_ppl_qoq_q3_14b_s0_w4b.log",
22
+ "ppl": {
23
+ "gptq@2048": {
24
+ "c4": 14.610818076435985,
25
+ "wikitext": 9.16325201773376
26
+ }
27
+ },
28
+ "dual_basis": null,
29
+ "rotate": [
30
+ "[dc-w4b] rotate_llm: 13.99 B Linear parameters, CPU above 30 B -> rotation on the model device"
31
+ ],
32
+ "dc_commit": "69f3473f5e1c1504bae35cc50c7858ef900a9b17"
33
+ }
34
+ }
35
+ }