mlx-bun 0.0.4
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/LICENSE +21 -0
- package/README.md +362 -0
- package/THIRD_PARTY_LICENSES.md +41 -0
- package/bin/mlx-bun.mjs +32 -0
- package/docs/curve-designer.html +303 -0
- package/docs/design/adapters-end-to-end.md +201 -0
- package/docs/design/compat-cli-surface-design.md +227 -0
- package/docs/design/docs-reorg-plan.md +152 -0
- package/docs/design/hlg-sampling.md +340 -0
- package/docs/design/memory-system.md +339 -0
- package/docs/design/optimization_plan.md +344 -0
- package/docs/design/parallel-slots.md +515 -0
- package/docs/design/runtime-isolation.md +247 -0
- package/docs/design/segmented-backward-training.md +508 -0
- package/docs/investigations/chunk-finetune-experiment.md +84 -0
- package/docs/investigations/curve-bisector-routing.md +123 -0
- package/docs/investigations/curve-runs/bisector-route.json +98 -0
- package/docs/investigations/curve-runs/bisector-trace.json +1393 -0
- package/docs/investigations/curve-runs/curve-terrain.json +1179 -0
- package/docs/investigations/curve-runs/curve-terrain.svg +42 -0
- package/docs/investigations/curve-terrain.html +105 -0
- package/docs/investigations/expert-offload-single-user-moe.md +292 -0
- package/docs/investigations/hlg-ab.html +34 -0
- package/docs/investigations/hlg-explorer.html +238 -0
- package/docs/investigations/hlg-figs/cliff.svg +1 -0
- package/docs/investigations/hlg-figs/cross.svg +1 -0
- package/docs/investigations/hlg-figs/edges.svg +1 -0
- package/docs/investigations/hlg-figs/k20-grids.svg +1 -0
- package/docs/investigations/hlg-figs/k20-peaks.svg +1 -0
- package/docs/investigations/hlg-figs/lw-curve.svg +1 -0
- package/docs/investigations/hlg-figs/terrain.svg +1 -0
- package/docs/investigations/hlg-report.html +262 -0
- package/docs/investigations/hlg-runs/hlg-ab.json +149 -0
- package/docs/investigations/hlg-runs/hlg-blind-prompts.md +87 -0
- package/docs/investigations/hlg-runs/hlg-blind.json +107 -0
- package/docs/investigations/hlg-runs/hlg-cliff.json +197 -0
- package/docs/investigations/hlg-runs/hlg-cross.json +505 -0
- package/docs/investigations/hlg-runs/hlg-edge.json +388 -0
- package/docs/investigations/hlg-runs/hlg-eval1.json +25 -0
- package/docs/investigations/hlg-runs/hlg-frontier.json +86 -0
- package/docs/investigations/hlg-runs/hlg-grid.json +242 -0
- package/docs/investigations/hlg-runs/hlg-grid2.json +343 -0
- package/docs/investigations/hlg-runs/hlg-lw-coarse.json +110 -0
- package/docs/investigations/hlg-runs/hlg-map.json +383 -0
- package/docs/investigations/hlg-runs/run0-original.log +159 -0
- package/docs/investigations/hlg-runs/runA-recommended.log +131 -0
- package/docs/investigations/hlg-runs/runAA-cliff-cpm5.log +26 -0
- package/docs/investigations/hlg-runs/runAB-grid2-e4b.log +18 -0
- package/docs/investigations/hlg-runs/runAC-peak-e4b-partial.log +25 -0
- package/docs/investigations/hlg-runs/runAD-ab-samples.log +8 -0
- package/docs/investigations/hlg-runs/runAE-edge-e4b.log +51 -0
- package/docs/investigations/hlg-runs/runAF-logit-calibration.log +10 -0
- package/docs/investigations/hlg-runs/runAG-eval-config2.log +16 -0
- package/docs/investigations/hlg-runs/runB-gain-variants.log +192 -0
- package/docs/investigations/hlg-runs/runC-pivots.log +162 -0
- package/docs/investigations/hlg-runs/runD-oetf.log +132 -0
- package/docs/investigations/hlg-runs/runE-eotf.log +162 -0
- package/docs/investigations/hlg-runs/runF-shaper-working.log +86 -0
- package/docs/investigations/hlg-runs/runG-shaper-sweep.log +101 -0
- package/docs/investigations/hlg-runs/runH-wsweep.log +176 -0
- package/docs/investigations/hlg-runs/runI-asweep.log +176 -0
- package/docs/investigations/hlg-runs/runJ-knobmap.log +49 -0
- package/docs/investigations/hlg-runs/runK-frontier.log +19 -0
- package/docs/investigations/hlg-runs/runL-widemap.log +65 -0
- package/docs/investigations/hlg-runs/runM-frontier-emb.log +15 -0
- package/docs/investigations/hlg-runs/runN-verify.log +116 -0
- package/docs/investigations/hlg-runs/runO-frontier-seed5000.log +15 -0
- package/docs/investigations/hlg-runs/runP-grid-AxSm.log +15 -0
- package/docs/investigations/hlg-runs/runQ-grid-tgW.log +15 -0
- package/docs/investigations/hlg-runs/runR-blind-transcript.log +79 -0
- package/docs/investigations/hlg-runs/runS-cross-coherence.log +33 -0
- package/docs/investigations/hlg-runs/runT-cross-cpm5.log +33 -0
- package/docs/investigations/hlg-runs/runU-cross-12b.log +33 -0
- package/docs/investigations/hlg-runs/runV-lw-e4b.log +22 -0
- package/docs/investigations/hlg-runs/runW-lw-12b.log +22 -0
- package/docs/investigations/hlg-runs/runX-lw-cpm5.log +22 -0
- package/docs/investigations/hlg-runs/runY-cliff-e4b.log +26 -0
- package/docs/investigations/hlg-runs/runZ-cliff-12b.log +26 -0
- package/docs/investigations/hlg-sampling-investigation.md +970 -0
- package/docs/investigations/lab-build-journal.md +157 -0
- package/docs/investigations/pi-builtin-investigation.md +308 -0
- package/docs/investigations/starter-model-port-handoff.md +323 -0
- package/docs/planning/IDEAS.md +26 -0
- package/docs/planning/PRODUCT_ROADMAP.md +426 -0
- package/docs/planning/ResearchTopics.md +119 -0
- package/docs/planning/chunk-task-roadmap.md +77 -0
- package/docs/planning/journal.md +155 -0
- package/docs/reference/distribution.md +121 -0
- package/docs/reference/embedding.md +168 -0
- package/docs/reference/library-api.md +153 -0
- package/docs/reference/server-api.md +224 -0
- package/docs/reference/server-config.md +243 -0
- package/docs/segmented-backward-handoff.md +148 -0
- package/package.json +68 -0
- package/src/anthropic.ts +491 -0
- package/src/chat-template.ts +186 -0
- package/src/cli.ts +1075 -0
- package/src/config.ts +229 -0
- package/src/curve-sampler.ts +122 -0
- package/src/dataset/generators.ts +1087 -0
- package/src/dataset/index.ts +12 -0
- package/src/dataset/job.ts +81 -0
- package/src/dataset/llm.ts +71 -0
- package/src/dataset/registry.ts +528 -0
- package/src/download.ts +283 -0
- package/src/eval/README.md +86 -0
- package/src/eval/kl-prompts.ts +29 -0
- package/src/eval/kl.ts +294 -0
- package/src/eval/quality-db.ts +92 -0
- package/src/eval/runner.ts +102 -0
- package/src/eval/score.ts +49 -0
- package/src/eval/tasks/bfcl.ts +500 -0
- package/src/eval/tasks/gsm8k.ts +85 -0
- package/src/eval/tasks/hashhop.ts +109 -0
- package/src/eval/tasks/humaneval.ts +252 -0
- package/src/eval/tasks/ifeval.ts +354 -0
- package/src/eval/tasks/mmlu.ts +145 -0
- package/src/evaldb.ts +199 -0
- package/src/expert-offload-build.ts +77 -0
- package/src/expert-offload.ts +61 -0
- package/src/expert-trace.ts +67 -0
- package/src/fit.ts +254 -0
- package/src/generate.ts +451 -0
- package/src/harness-pi.ts +177 -0
- package/src/hf-push.ts +628 -0
- package/src/index.ts +22 -0
- package/src/jobs/db.ts +139 -0
- package/src/jobs/index.ts +29 -0
- package/src/jobs/job-entry.ts +106 -0
- package/src/jobs/runner.ts +304 -0
- package/src/jobs/sse.ts +114 -0
- package/src/jobs/types.ts +58 -0
- package/src/kv-store.ts +185 -0
- package/src/lora.ts +314 -0
- package/src/md-modules.d.ts +7 -0
- package/src/mlx/array.ts +237 -0
- package/src/mlx/autograd.ts +267 -0
- package/src/mlx/checkpoint.ts +116 -0
- package/src/mlx/compile.ts +158 -0
- package/src/mlx/custom-vjp.ts +140 -0
- package/src/mlx/ffi.ts +433 -0
- package/src/mlx/metal-kernel.ts +130 -0
- package/src/mlx/ops.ts +707 -0
- package/src/mmap.ts +90 -0
- package/src/model/batched-mask.ts +204 -0
- package/src/model/batched-rotating.ts +300 -0
- package/src/model/compiled-decode.ts +681 -0
- package/src/model/factory.ts +27 -0
- package/src/model/fingerprint.ts +33 -0
- package/src/model/flash-attention.ts +674 -0
- package/src/model/fused-decode-kernel.ts +305 -0
- package/src/model/fused-geglu-kernel.ts +87 -0
- package/src/model/gemma4-base.ts +1616 -0
- package/src/model/gemma4.ts +1012 -0
- package/src/model/generated/gemma4-12b.ts +526 -0
- package/src/model/generated/gemma4-26b.ts +500 -0
- package/src/model/generated/gemma4-e4b.ts +757 -0
- package/src/model/generated/index.ts +18 -0
- package/src/model/minicpm5.ts +278 -0
- package/src/model/qwen3-delta.ts +224 -0
- package/src/model/qwen3_5.ts +429 -0
- package/src/model/support.ts +46 -0
- package/src/native-pack.ts +152 -0
- package/src/pi-provider.ts +95 -0
- package/src/pi-terminal.ts +264 -0
- package/src/pi-web.ts +924 -0
- package/src/preflight.ts +123 -0
- package/src/prompt-cache.ts +113 -0
- package/src/quantize/allocator.ts +627 -0
- package/src/quantize/calibration.ts +180 -0
- package/src/quantize/config-writer.ts +132 -0
- package/src/quantize/index.ts +57 -0
- package/src/quantize/job.ts +136 -0
- package/src/quantize/quantizer.ts +462 -0
- package/src/quantize/safetensors-writer.ts +171 -0
- package/src/quantize/sensitivity.ts +234 -0
- package/src/registry.ts +336 -0
- package/src/responses.ts +774 -0
- package/src/safetensors.ts +158 -0
- package/src/sampler.ts +517 -0
- package/src/serve/batch-scheduler.ts +362 -0
- package/src/serve/generation-gateway.ts +160 -0
- package/src/server.ts +1891 -0
- package/src/spec/drafter.ts +261 -0
- package/src/spec/generate.ts +221 -0
- package/src/status-page.html +497 -0
- package/src/tokenizer.ts +38 -0
- package/src/tool-call.ts +242 -0
- package/src/train/dataset.ts +461 -0
- package/src/train/export.ts +32 -0
- package/src/train/forward.ts +229 -0
- package/src/train/index.ts +43 -0
- package/src/train/job.ts +148 -0
- package/src/train/lora-params.ts +194 -0
- package/src/train/loss.ts +342 -0
- package/src/train/merge.ts +195 -0
- package/src/train/optimizer.ts +176 -0
- package/src/train/rank.ts +101 -0
- package/src/train/segmented.ts +520 -0
- package/src/train/trainer.ts +600 -0
- package/src/tui.ts +161 -0
- package/src/vision/embedder.ts +168 -0
- package/src/vision/preprocess.ts +247 -0
- package/src/vision/prompt.ts +158 -0
- package/src/vision/siglip.ts +599 -0
- package/src/web/app.html +2471 -0
- package/src/web/skills/web-research/SKILL.md +44 -0
- package/src/web/skills.ts +59 -0
- package/src/web-tools.ts +426 -0
- package/src/weights.ts +91 -0
|
@@ -0,0 +1,388 @@
|
|
|
1
|
+
{
|
|
2
|
+
"model": "gemma-4-e4b-it-OptiQ-4bit",
|
|
3
|
+
"k": 20,
|
|
4
|
+
"base": {
|
|
5
|
+
"window": 6,
|
|
6
|
+
"shoulderA": 0.35,
|
|
7
|
+
"sM": 0.7,
|
|
8
|
+
"targetGap": 15,
|
|
9
|
+
"refGap": 4,
|
|
10
|
+
"lw": 1200
|
|
11
|
+
},
|
|
12
|
+
"zooms": [
|
|
13
|
+
{
|
|
14
|
+
"name": "target_gap ramp (base) — sub-unit",
|
|
15
|
+
"key": "targetGap",
|
|
16
|
+
"over": {},
|
|
17
|
+
"values": [
|
|
18
|
+
7,
|
|
19
|
+
7.5,
|
|
20
|
+
8,
|
|
21
|
+
8.5,
|
|
22
|
+
9,
|
|
23
|
+
9.5,
|
|
24
|
+
10,
|
|
25
|
+
10.5,
|
|
26
|
+
11
|
|
27
|
+
]
|
|
28
|
+
},
|
|
29
|
+
{
|
|
30
|
+
"name": "W channel-close @ A=100 — sub-unit",
|
|
31
|
+
"key": "window",
|
|
32
|
+
"over": {
|
|
33
|
+
"shoulderA": 100
|
|
34
|
+
},
|
|
35
|
+
"values": [
|
|
36
|
+
6,
|
|
37
|
+
6.5,
|
|
38
|
+
7,
|
|
39
|
+
7.25,
|
|
40
|
+
7.5,
|
|
41
|
+
7.75,
|
|
42
|
+
8,
|
|
43
|
+
8.5,
|
|
44
|
+
9
|
|
45
|
+
]
|
|
46
|
+
},
|
|
47
|
+
{
|
|
48
|
+
"name": "L_W low wall (γ→0 inversion)",
|
|
49
|
+
"key": "lw",
|
|
50
|
+
"over": {},
|
|
51
|
+
"values": [
|
|
52
|
+
1.4,
|
|
53
|
+
1.6,
|
|
54
|
+
1.8,
|
|
55
|
+
2,
|
|
56
|
+
2.3,
|
|
57
|
+
2.7,
|
|
58
|
+
3.2,
|
|
59
|
+
4,
|
|
60
|
+
6,
|
|
61
|
+
10
|
|
62
|
+
],
|
|
63
|
+
"gamma": true
|
|
64
|
+
},
|
|
65
|
+
{
|
|
66
|
+
"name": "L_W high wall (γ→2.4 over-sharpen)",
|
|
67
|
+
"key": "lw",
|
|
68
|
+
"over": {},
|
|
69
|
+
"values": [
|
|
70
|
+
150000,
|
|
71
|
+
200000,
|
|
72
|
+
300000,
|
|
73
|
+
500000,
|
|
74
|
+
700000,
|
|
75
|
+
1000000
|
|
76
|
+
],
|
|
77
|
+
"gamma": true
|
|
78
|
+
}
|
|
79
|
+
],
|
|
80
|
+
"grid": [
|
|
81
|
+
{
|
|
82
|
+
"zoom": "target_gap ramp (base) — sub-unit",
|
|
83
|
+
"key": "targetGap",
|
|
84
|
+
"value": 7,
|
|
85
|
+
"gamma": null,
|
|
86
|
+
"meanJunk": 0.24456282327979945,
|
|
87
|
+
"maxJunk": 0.4318181818181818,
|
|
88
|
+
"breakRate": 1
|
|
89
|
+
},
|
|
90
|
+
{
|
|
91
|
+
"zoom": "target_gap ramp (base) — sub-unit",
|
|
92
|
+
"key": "targetGap",
|
|
93
|
+
"value": 7.5,
|
|
94
|
+
"gamma": null,
|
|
95
|
+
"meanJunk": 0.22452164954945486,
|
|
96
|
+
"maxJunk": 0.4090909090909091,
|
|
97
|
+
"breakRate": 1
|
|
98
|
+
},
|
|
99
|
+
{
|
|
100
|
+
"zoom": "target_gap ramp (base) — sub-unit",
|
|
101
|
+
"key": "targetGap",
|
|
102
|
+
"value": 8,
|
|
103
|
+
"gamma": null,
|
|
104
|
+
"meanJunk": 0.19681853451251913,
|
|
105
|
+
"maxJunk": 0.5535714285714286,
|
|
106
|
+
"breakRate": 1
|
|
107
|
+
},
|
|
108
|
+
{
|
|
109
|
+
"zoom": "target_gap ramp (base) — sub-unit",
|
|
110
|
+
"key": "targetGap",
|
|
111
|
+
"value": 8.5,
|
|
112
|
+
"gamma": null,
|
|
113
|
+
"meanJunk": 0.13733405619367972,
|
|
114
|
+
"maxJunk": 0.8372093023255814,
|
|
115
|
+
"breakRate": 0.85
|
|
116
|
+
},
|
|
117
|
+
{
|
|
118
|
+
"zoom": "target_gap ramp (base) — sub-unit",
|
|
119
|
+
"key": "targetGap",
|
|
120
|
+
"value": 9,
|
|
121
|
+
"gamma": null,
|
|
122
|
+
"meanJunk": 0.0688070885356468,
|
|
123
|
+
"maxJunk": 0.6744186046511628,
|
|
124
|
+
"breakRate": 0.7
|
|
125
|
+
},
|
|
126
|
+
{
|
|
127
|
+
"zoom": "target_gap ramp (base) — sub-unit",
|
|
128
|
+
"key": "targetGap",
|
|
129
|
+
"value": 9.5,
|
|
130
|
+
"gamma": null,
|
|
131
|
+
"meanJunk": 0.05854233213812482,
|
|
132
|
+
"maxJunk": 0.7727272727272727,
|
|
133
|
+
"breakRate": 0.5
|
|
134
|
+
},
|
|
135
|
+
{
|
|
136
|
+
"zoom": "target_gap ramp (base) — sub-unit",
|
|
137
|
+
"key": "targetGap",
|
|
138
|
+
"value": 10,
|
|
139
|
+
"gamma": null,
|
|
140
|
+
"meanJunk": 0.0024470899470899468,
|
|
141
|
+
"maxJunk": 0.037037037037037035,
|
|
142
|
+
"breakRate": 0.1
|
|
143
|
+
},
|
|
144
|
+
{
|
|
145
|
+
"zoom": "target_gap ramp (base) — sub-unit",
|
|
146
|
+
"key": "targetGap",
|
|
147
|
+
"value": 10.5,
|
|
148
|
+
"gamma": null,
|
|
149
|
+
"meanJunk": 0.002339424141749723,
|
|
150
|
+
"maxJunk": 0.03488372093023256,
|
|
151
|
+
"breakRate": 0.1
|
|
152
|
+
},
|
|
153
|
+
{
|
|
154
|
+
"zoom": "target_gap ramp (base) — sub-unit",
|
|
155
|
+
"key": "targetGap",
|
|
156
|
+
"value": 11,
|
|
157
|
+
"gamma": null,
|
|
158
|
+
"meanJunk": 0.003655213615433933,
|
|
159
|
+
"maxJunk": 0.03488372093023256,
|
|
160
|
+
"breakRate": 0.15
|
|
161
|
+
},
|
|
162
|
+
{
|
|
163
|
+
"zoom": "W channel-close @ A=100 — sub-unit",
|
|
164
|
+
"key": "window",
|
|
165
|
+
"value": 6,
|
|
166
|
+
"gamma": null,
|
|
167
|
+
"meanJunk": 0.16097243758841046,
|
|
168
|
+
"maxJunk": 0.6190476190476191,
|
|
169
|
+
"breakRate": 0.9
|
|
170
|
+
},
|
|
171
|
+
{
|
|
172
|
+
"zoom": "W channel-close @ A=100 — sub-unit",
|
|
173
|
+
"key": "window",
|
|
174
|
+
"value": 6.5,
|
|
175
|
+
"gamma": null,
|
|
176
|
+
"meanJunk": 0.08816660696226027,
|
|
177
|
+
"maxJunk": 0.42857142857142855,
|
|
178
|
+
"breakRate": 0.9
|
|
179
|
+
},
|
|
180
|
+
{
|
|
181
|
+
"zoom": "W channel-close @ A=100 — sub-unit",
|
|
182
|
+
"key": "window",
|
|
183
|
+
"value": 7,
|
|
184
|
+
"gamma": null,
|
|
185
|
+
"meanJunk": 0.05842702422655176,
|
|
186
|
+
"maxJunk": 0.7678571428571429,
|
|
187
|
+
"breakRate": 0.5
|
|
188
|
+
},
|
|
189
|
+
{
|
|
190
|
+
"zoom": "W channel-close @ A=100 — sub-unit",
|
|
191
|
+
"key": "window",
|
|
192
|
+
"value": 7.25,
|
|
193
|
+
"gamma": null,
|
|
194
|
+
"meanJunk": 0.050461198587243336,
|
|
195
|
+
"maxJunk": 0.7678571428571429,
|
|
196
|
+
"breakRate": 0.35
|
|
197
|
+
},
|
|
198
|
+
{
|
|
199
|
+
"zoom": "W channel-close @ A=100 — sub-unit",
|
|
200
|
+
"key": "window",
|
|
201
|
+
"value": 7.5,
|
|
202
|
+
"gamma": null,
|
|
203
|
+
"meanJunk": 0.048574950540700275,
|
|
204
|
+
"maxJunk": 0.88,
|
|
205
|
+
"breakRate": 0.2
|
|
206
|
+
},
|
|
207
|
+
{
|
|
208
|
+
"zoom": "W channel-close @ A=100 — sub-unit",
|
|
209
|
+
"key": "window",
|
|
210
|
+
"value": 7.75,
|
|
211
|
+
"gamma": null,
|
|
212
|
+
"meanJunk": 0.0073722988083298555,
|
|
213
|
+
"maxJunk": 0.07042253521126761,
|
|
214
|
+
"breakRate": 0.2
|
|
215
|
+
},
|
|
216
|
+
{
|
|
217
|
+
"zoom": "W channel-close @ A=100 — sub-unit",
|
|
218
|
+
"key": "window",
|
|
219
|
+
"value": 8,
|
|
220
|
+
"gamma": null,
|
|
221
|
+
"meanJunk": 0.002995346586677856,
|
|
222
|
+
"maxJunk": 0.03529411764705882,
|
|
223
|
+
"breakRate": 0.1
|
|
224
|
+
},
|
|
225
|
+
{
|
|
226
|
+
"zoom": "W channel-close @ A=100 — sub-unit",
|
|
227
|
+
"key": "window",
|
|
228
|
+
"value": 8.5,
|
|
229
|
+
"gamma": null,
|
|
230
|
+
"meanJunk": 0,
|
|
231
|
+
"maxJunk": 0,
|
|
232
|
+
"breakRate": 0
|
|
233
|
+
},
|
|
234
|
+
{
|
|
235
|
+
"zoom": "W channel-close @ A=100 — sub-unit",
|
|
236
|
+
"key": "window",
|
|
237
|
+
"value": 9,
|
|
238
|
+
"gamma": null,
|
|
239
|
+
"meanJunk": 0,
|
|
240
|
+
"maxJunk": 0,
|
|
241
|
+
"breakRate": 0
|
|
242
|
+
},
|
|
243
|
+
{
|
|
244
|
+
"zoom": "L_W low wall (γ→0 inversion)",
|
|
245
|
+
"key": "lw",
|
|
246
|
+
"value": 1.4,
|
|
247
|
+
"gamma": 0.0013737749848601322,
|
|
248
|
+
"meanJunk": 0,
|
|
249
|
+
"maxJunk": 0,
|
|
250
|
+
"breakRate": 0
|
|
251
|
+
},
|
|
252
|
+
{
|
|
253
|
+
"zoom": "L_W low wall (γ→0 inversion)",
|
|
254
|
+
"key": "lw",
|
|
255
|
+
"value": 1.6,
|
|
256
|
+
"gamma": 0.025730392715488604,
|
|
257
|
+
"meanJunk": 0,
|
|
258
|
+
"maxJunk": 0,
|
|
259
|
+
"breakRate": 0
|
|
260
|
+
},
|
|
261
|
+
{
|
|
262
|
+
"zoom": "L_W low wall (γ→0 inversion)",
|
|
263
|
+
"key": "lw",
|
|
264
|
+
"value": 1.8,
|
|
265
|
+
"gamma": 0.0472144521433886,
|
|
266
|
+
"meanJunk": 0,
|
|
267
|
+
"maxJunk": 0,
|
|
268
|
+
"breakRate": 0
|
|
269
|
+
},
|
|
270
|
+
{
|
|
271
|
+
"zoom": "L_W low wall (γ→0 inversion)",
|
|
272
|
+
"key": "lw",
|
|
273
|
+
"value": 2,
|
|
274
|
+
"gamma": 0.06643259817887204,
|
|
275
|
+
"meanJunk": 0,
|
|
276
|
+
"maxJunk": 0,
|
|
277
|
+
"breakRate": 0
|
|
278
|
+
},
|
|
279
|
+
{
|
|
280
|
+
"zoom": "L_W low wall (γ→0 inversion)",
|
|
281
|
+
"key": "lw",
|
|
282
|
+
"value": 2.3,
|
|
283
|
+
"gamma": 0.09192569112738913,
|
|
284
|
+
"meanJunk": 0,
|
|
285
|
+
"maxJunk": 0,
|
|
286
|
+
"breakRate": 0
|
|
287
|
+
},
|
|
288
|
+
{
|
|
289
|
+
"zoom": "L_W low wall (γ→0 inversion)",
|
|
290
|
+
"key": "lw",
|
|
291
|
+
"value": 2.7,
|
|
292
|
+
"gamma": 0.12117278094677464,
|
|
293
|
+
"meanJunk": 0,
|
|
294
|
+
"maxJunk": 0,
|
|
295
|
+
"breakRate": 0
|
|
296
|
+
},
|
|
297
|
+
{
|
|
298
|
+
"zoom": "L_W low wall (γ→0 inversion)",
|
|
299
|
+
"key": "lw",
|
|
300
|
+
"value": 3.2,
|
|
301
|
+
"gamma": 0.15216299089436047,
|
|
302
|
+
"meanJunk": 0,
|
|
303
|
+
"maxJunk": 0,
|
|
304
|
+
"breakRate": 0
|
|
305
|
+
},
|
|
306
|
+
{
|
|
307
|
+
"zoom": "L_W low wall (γ→0 inversion)",
|
|
308
|
+
"key": "lw",
|
|
309
|
+
"value": 4,
|
|
310
|
+
"gamma": 0.19286519635774435,
|
|
311
|
+
"meanJunk": 0,
|
|
312
|
+
"maxJunk": 0,
|
|
313
|
+
"breakRate": 0
|
|
314
|
+
},
|
|
315
|
+
{
|
|
316
|
+
"zoom": "L_W low wall (γ→0 inversion)",
|
|
317
|
+
"key": "lw",
|
|
318
|
+
"value": 6,
|
|
319
|
+
"gamma": 0.2668235251611304,
|
|
320
|
+
"meanJunk": 0,
|
|
321
|
+
"maxJunk": 0,
|
|
322
|
+
"breakRate": 0
|
|
323
|
+
},
|
|
324
|
+
{
|
|
325
|
+
"zoom": "L_W low wall (γ→0 inversion)",
|
|
326
|
+
"key": "lw",
|
|
327
|
+
"value": 10,
|
|
328
|
+
"gamma": 0.36,
|
|
329
|
+
"meanJunk": 0,
|
|
330
|
+
"maxJunk": 0,
|
|
331
|
+
"breakRate": 0
|
|
332
|
+
},
|
|
333
|
+
{
|
|
334
|
+
"zoom": "L_W high wall (γ→2.4 over-sharpen)",
|
|
335
|
+
"key": "lw",
|
|
336
|
+
"value": 150000,
|
|
337
|
+
"gamma": 2.113958328803386,
|
|
338
|
+
"meanJunk": 0,
|
|
339
|
+
"maxJunk": 0,
|
|
340
|
+
"breakRate": 0
|
|
341
|
+
},
|
|
342
|
+
{
|
|
343
|
+
"zoom": "L_W high wall (γ→2.4 over-sharpen)",
|
|
344
|
+
"key": "lw",
|
|
345
|
+
"value": 200000,
|
|
346
|
+
"gamma": 2.166432598178872,
|
|
347
|
+
"meanJunk": 0.0013157894736842105,
|
|
348
|
+
"maxJunk": 0.02631578947368421,
|
|
349
|
+
"breakRate": 0.05
|
|
350
|
+
},
|
|
351
|
+
{
|
|
352
|
+
"zoom": "L_W high wall (γ→2.4 over-sharpen)",
|
|
353
|
+
"key": "lw",
|
|
354
|
+
"value": 300000,
|
|
355
|
+
"gamma": 2.240390926982258,
|
|
356
|
+
"meanJunk": 0.0013157894736842105,
|
|
357
|
+
"maxJunk": 0.02631578947368421,
|
|
358
|
+
"breakRate": 0.05
|
|
359
|
+
},
|
|
360
|
+
{
|
|
361
|
+
"zoom": "L_W high wall (γ→2.4 over-sharpen)",
|
|
362
|
+
"key": "lw",
|
|
363
|
+
"value": 500000,
|
|
364
|
+
"gamma": 2.333567401821128,
|
|
365
|
+
"meanJunk": 0.003059975520195838,
|
|
366
|
+
"maxJunk": 0.03488372093023256,
|
|
367
|
+
"breakRate": 0.1
|
|
368
|
+
},
|
|
369
|
+
{
|
|
370
|
+
"zoom": "L_W high wall (γ→2.4 over-sharpen)",
|
|
371
|
+
"key": "lw",
|
|
372
|
+
"value": 700000,
|
|
373
|
+
"gamma": 2.394941176805988,
|
|
374
|
+
"meanJunk": 0.003059975520195838,
|
|
375
|
+
"maxJunk": 0.03488372093023256,
|
|
376
|
+
"breakRate": 0.1
|
|
377
|
+
},
|
|
378
|
+
{
|
|
379
|
+
"zoom": "L_W high wall (γ→2.4 over-sharpen)",
|
|
380
|
+
"key": "lw",
|
|
381
|
+
"value": 1000000,
|
|
382
|
+
"gamma": 2.46,
|
|
383
|
+
"meanJunk": 0.003655213615433933,
|
|
384
|
+
"maxJunk": 0.03488372093023256,
|
|
385
|
+
"breakRate": 0.15
|
|
386
|
+
}
|
|
387
|
+
]
|
|
388
|
+
}
|
|
@@ -0,0 +1,25 @@
|
|
|
1
|
+
{
|
|
2
|
+
"model": "gemma-4-e4b-it-OptiQ-4bit",
|
|
3
|
+
"k": 20,
|
|
4
|
+
"config": {
|
|
5
|
+
"window": 8,
|
|
6
|
+
"shoulderA": 0.02,
|
|
7
|
+
"sM": 0.15,
|
|
8
|
+
"targetGap": 15,
|
|
9
|
+
"refGap": 4,
|
|
10
|
+
"lw": 1000000,
|
|
11
|
+
"xM": 0.77,
|
|
12
|
+
"yM": 0.7,
|
|
13
|
+
"xFloor": 0.45,
|
|
14
|
+
"yFloor": 0.05,
|
|
15
|
+
"p": 4
|
|
16
|
+
},
|
|
17
|
+
"gamma": 2.46,
|
|
18
|
+
"refDiv": 0.6333071413347174,
|
|
19
|
+
"refJunk": 0,
|
|
20
|
+
"canMean": 0.0012987012987012987,
|
|
21
|
+
"canMax": 0.025974025974025976,
|
|
22
|
+
"broke": 1,
|
|
23
|
+
"div": 0.619160697663867,
|
|
24
|
+
"openJunk": 0
|
|
25
|
+
}
|
|
@@ -0,0 +1,86 @@
|
|
|
1
|
+
{
|
|
2
|
+
"model": "gemma-4-e4b-it-OptiQ-4bit",
|
|
3
|
+
"N": 10,
|
|
4
|
+
"reference": {
|
|
5
|
+
"sbleu": 0.7546809695639246,
|
|
6
|
+
"emb": 0.12136846955801817,
|
|
7
|
+
"d2": 0.7729703451628585,
|
|
8
|
+
"junk": 0
|
|
9
|
+
},
|
|
10
|
+
"dominance": true,
|
|
11
|
+
"results": [
|
|
12
|
+
{
|
|
13
|
+
"name": "base (W6 tg15 A.35)",
|
|
14
|
+
"over": {},
|
|
15
|
+
"canaryJunk": 0,
|
|
16
|
+
"sbleu": 0.5195901989780582,
|
|
17
|
+
"emb": 0.09224795558066878,
|
|
18
|
+
"d2": 0.6371700348032079,
|
|
19
|
+
"junk": 0,
|
|
20
|
+
"verdict": "acceptable"
|
|
21
|
+
},
|
|
22
|
+
{
|
|
23
|
+
"name": "A=0.01 (loose)",
|
|
24
|
+
"over": {
|
|
25
|
+
"shoulderA": 0.01
|
|
26
|
+
},
|
|
27
|
+
"canaryJunk": 0,
|
|
28
|
+
"sbleu": 0.7217647475356881,
|
|
29
|
+
"emb": 0.1381691980997049,
|
|
30
|
+
"d2": 0.7599092778348939,
|
|
31
|
+
"junk": 0,
|
|
32
|
+
"verdict": "★ beats emb"
|
|
33
|
+
},
|
|
34
|
+
{
|
|
35
|
+
"name": "sM=0.05 (loose)",
|
|
36
|
+
"over": {
|
|
37
|
+
"sM": 0.05
|
|
38
|
+
},
|
|
39
|
+
"canaryJunk": 0,
|
|
40
|
+
"sbleu": 0.7702710786696176,
|
|
41
|
+
"emb": 0.13287383091062793,
|
|
42
|
+
"d2": 0.7575513538748833,
|
|
43
|
+
"junk": 0,
|
|
44
|
+
"verdict": "★★ BEATS both"
|
|
45
|
+
},
|
|
46
|
+
{
|
|
47
|
+
"name": "Lw=10 (loose)",
|
|
48
|
+
"over": {
|
|
49
|
+
"lw": 10
|
|
50
|
+
},
|
|
51
|
+
"canaryJunk": 0,
|
|
52
|
+
"sbleu": 0.5555123068361736,
|
|
53
|
+
"emb": 0.1095044115287212,
|
|
54
|
+
"d2": 0.6640863402202801,
|
|
55
|
+
"junk": 0,
|
|
56
|
+
"verdict": "acceptable"
|
|
57
|
+
},
|
|
58
|
+
{
|
|
59
|
+
"name": "A.01 × sM.05",
|
|
60
|
+
"over": {
|
|
61
|
+
"shoulderA": 0.01,
|
|
62
|
+
"sM": 0.05
|
|
63
|
+
},
|
|
64
|
+
"canaryJunk": 0,
|
|
65
|
+
"sbleu": 0.7465954477438622,
|
|
66
|
+
"emb": 0.1410434919251719,
|
|
67
|
+
"d2": 0.7616955465075907,
|
|
68
|
+
"junk": 0,
|
|
69
|
+
"verdict": "★ beats emb"
|
|
70
|
+
},
|
|
71
|
+
{
|
|
72
|
+
"name": "A.01 × sM.05 × Lw10",
|
|
73
|
+
"over": {
|
|
74
|
+
"shoulderA": 0.01,
|
|
75
|
+
"sM": 0.05,
|
|
76
|
+
"lw": 10
|
|
77
|
+
},
|
|
78
|
+
"canaryJunk": 0,
|
|
79
|
+
"sbleu": 0.7414095926628356,
|
|
80
|
+
"emb": 0.13942094473615174,
|
|
81
|
+
"d2": 0.7537338673787272,
|
|
82
|
+
"junk": 0,
|
|
83
|
+
"verdict": "★ beats emb"
|
|
84
|
+
}
|
|
85
|
+
]
|
|
86
|
+
}
|