@miller-tech/uap 2.39.0 → 2.44.1
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/config/capacity-policy.json +25 -0
- package/config/model-profiles/qwen38.json +8 -7
- package/config/model-registry.json +668 -20
- package/dist/.tsbuildinfo +1 -1
- package/dist/cli/deliver.js +1 -1
- package/dist/cli/deliver.js.map +1 -1
- package/dist/cli/models.d.ts +6 -0
- package/dist/cli/models.d.ts.map +1 -1
- package/dist/cli/models.js +11 -2
- package/dist/cli/models.js.map +1 -1
- package/dist/cli/verify.d.ts +9 -1
- package/dist/cli/verify.d.ts.map +1 -1
- package/dist/cli/verify.js +15 -3
- package/dist/cli/verify.js.map +1 -1
- package/dist/delivery/acceptance-judge.d.ts +12 -0
- package/dist/delivery/acceptance-judge.d.ts.map +1 -1
- package/dist/delivery/acceptance-judge.js +109 -8
- package/dist/delivery/acceptance-judge.js.map +1 -1
- package/dist/inference/model-files.d.ts.map +1 -1
- package/dist/inference/model-files.js +6 -1
- package/dist/inference/model-files.js.map +1 -1
- package/dist/models/types.d.ts.map +1 -1
- package/dist/models/types.js +14 -2
- package/dist/models/types.js.map +1 -1
- package/docs/INDEX.md +1 -0
- package/docs/architecture/adr/0015-session-hook-model-convergence-as-standing-consent.md +68 -0
- package/docs/benchmarks/PROMPT_UPLIFT_BLENDER.md +161 -0
- package/docs/benchmarks/README.md +1 -0
- package/docs/guides/LOCAL_MODELS.md +42 -0
- package/package.json +1 -1
- package/scripts/lib/droid-sync-contracts.sh +114 -0
- package/src/policies/enforcers/__pycache__/_common.cpython-311.pyc +0 -0
- package/src/policies/enforcers/__pycache__/expert_review_required.cpython-311.pyc +0 -0
- package/src/policies/enforcers/__pycache__/quality_metrics_gate.cpython-311.pyc +0 -0
- package/src/policies/enforcers/delivery_enforcement.py +34 -12
- package/templates/hooks/__pycache__/deliver_autoroute.cpython-311.pyc +0 -0
- package/templates/hooks/pre-tool-use-bash.sh +34 -3
- package/templates/hooks/stop.sh +39 -0
- package/tools/agents/scripts/__pycache__/toolcall_path_normalizer.cpython-311.pyc +0 -0
- package/tools/agents/scripts/anthropic_proxy.py +16 -4
- package/tools/agents/tests/test_delivery_enforcement_write_complexity.py +26 -0
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"version": 1,
|
|
3
|
-
"_comment": "Repo defaults for the model placement registry. Shape only: unit/service names, engines, profiles, affinity, and the config inventory. Measured footprints (resident_gpu_mib, host_rss_mib, device totals) are machine state and live in ~/.uap/model-registry.json, deep-merged over this file. A config without a measurement is 'unmeasured' and fails closed: it may be served if already resident, but it can never be loaded or evicted. See docs/specs/operator-model-placement.md
|
|
3
|
+
"_comment": "Repo defaults for the model placement registry. Shape only: unit/service names, engines, profiles, affinity, and the config inventory. Measured footprints (resident_gpu_mib, host_rss_mib, device totals) are machine state and live in ~/.uap/model-registry.json, deep-merged over this file. A config without a measurement is 'unmeasured' and fails closed: it may be served if already resident, but it can never be loaded or evicted. See docs/specs/operator-model-placement.md \u00a74.1. retired_local_aliases is the single source the Droid sync (scripts/sync-local-agent-configs.sh) migrates from \u2014 append the just-retired alias there on every arm swap.",
|
|
4
4
|
"devices": {
|
|
5
5
|
"gpu0": {
|
|
6
6
|
"kind": "gpu"
|
|
@@ -9,12 +9,25 @@
|
|
|
9
9
|
"kind": "cpu"
|
|
10
10
|
}
|
|
11
11
|
},
|
|
12
|
+
"retired_local_aliases": [
|
|
13
|
+
"qwen35-a3b-iq4xs",
|
|
14
|
+
"qwen38-gsq-rco-27b",
|
|
15
|
+
"qwen3.8-flash-next-iq3_s",
|
|
16
|
+
"swift-1.5-iq3_s"
|
|
17
|
+
],
|
|
12
18
|
"models": {
|
|
13
19
|
"qwen3.6-27b": {
|
|
14
20
|
"display": "Qwen3.6 27B",
|
|
15
21
|
"engine": "dflash",
|
|
16
|
-
"task_affinity": [
|
|
17
|
-
|
|
22
|
+
"task_affinity": [
|
|
23
|
+
"general",
|
|
24
|
+
"ugc"
|
|
25
|
+
],
|
|
26
|
+
"affinity": {
|
|
27
|
+
"device": [
|
|
28
|
+
"gpu0"
|
|
29
|
+
]
|
|
30
|
+
},
|
|
18
31
|
"endpoint": "http://127.0.0.1:8080",
|
|
19
32
|
"unit": "uap-dflash-server",
|
|
20
33
|
"advertises": [
|
|
@@ -31,7 +44,7 @@
|
|
|
31
44
|
"status": "unmeasured",
|
|
32
45
|
"engine": "dflash",
|
|
33
46
|
"unit": "uap-dflash-server",
|
|
34
|
-
"note": "lucebox-hub DFlash speculative-decoding server (Qwen3.6-27B Q3_K_M target + draft head); machine profile: ~/.config/uap/dflash-server.env. The ugc-factory brain (pay2u tools/ugc-factory) requests 'qwen3.6-27b' through the UAP proxy, so placement auto-loads this config on demand
|
|
47
|
+
"note": "lucebox-hub DFlash speculative-decoding server (Qwen3.6-27B Q3_K_M target + draft head); machine profile: ~/.config/uap/dflash-server.env. The ugc-factory brain (pay2u tools/ugc-factory) requests 'qwen3.6-27b' through the UAP proxy, so placement auto-loads this config on demand \u2014 run `uap models measure` to record this machine's footprint before it can load (fail-closed). The Claude contract ids are advertised here too: the proxy serves them locally (ANTHROPIC_PASSTHROUGH_MODELS=__local_only__), and without a registry mapping the placement gate would 409-park every SDK client that checks /v1/models and then asks for one."
|
|
35
48
|
}
|
|
36
49
|
}
|
|
37
50
|
},
|
|
@@ -40,26 +53,29 @@
|
|
|
40
53
|
"engine": "strata",
|
|
41
54
|
"unit": "uap-strata-server",
|
|
42
55
|
"service": "strata-server",
|
|
43
|
-
"launch": "/home/cogtek/dev/strata/run-
|
|
56
|
+
"launch": "/home/cogtek/dev/strata/run-unsloth-ud-iq4_xs-v31.sh",
|
|
44
57
|
"endpoint": "http://127.0.0.1:8080",
|
|
45
|
-
"task_affinity": [
|
|
46
|
-
|
|
58
|
+
"task_affinity": [
|
|
59
|
+
"code",
|
|
60
|
+
"general"
|
|
61
|
+
],
|
|
62
|
+
"affinity": {
|
|
63
|
+
"device": [
|
|
64
|
+
"gpu0"
|
|
65
|
+
]
|
|
66
|
+
},
|
|
47
67
|
"configs": {
|
|
48
|
-
"strata-iq3_s": {
|
|
49
|
-
"status": "unmeasured",
|
|
50
|
-
"note": "active backend on this host; run `uap models measure` to record its footprint"
|
|
51
|
-
},
|
|
52
68
|
"strata-swift-iq3_s": {
|
|
53
69
|
"status": "unmeasured",
|
|
54
70
|
"engine": "strata",
|
|
55
|
-
"unit": "uap-strata-server",
|
|
71
|
+
"unit": "uap-strata-swift-server",
|
|
56
72
|
"note": "Swift 1.5 (UkisAI reasoning-efficient Flash-Next post-train) GSQ-RCO IQ3_S on the strata engine; A/B trial vs strata-iq3_s since 2026-10-17 with identical tuning (spec 8/0.70, vram-reserve 3072, int8 KV resident 32768, 131k). Weights: ukisai Swift-1.5-Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S, sha256-verified, PLE table in shard 1 (setup-detected); pack strata-data/packs/swift-iq3_s; config strata-swift-iq3_s.json behind uap-strata-server.service (revert: point the unit --config back at strata-iq3_s.json, daemon-reload, restart). Run `uap models measure` to record this machine's footprint before it can load (fail-closed). Trial evidence: /home/cogtek/dev/strata/bench/results/2026-10-17-swift-iq3s-trial/."
|
|
57
73
|
},
|
|
58
74
|
"strata-abliterated-iq2_xs": {
|
|
59
75
|
"status": "unmeasured",
|
|
60
76
|
"engine": "strata",
|
|
61
|
-
"unit": "uap-strata-server",
|
|
62
|
-
"note": "The
|
|
77
|
+
"unit": "uap-strata-abliterated-server",
|
|
78
|
+
"note": "The :8080 default arm 2026-10-28 -> 2026-11-11, then AGAIN since 2026-11-13 (the operator switched the fleet back from the v31 default; DavidAU abliterated Flash-Next at IQ2_XS, full expert pinning without the IQ3_S RAM-OOM risk; config strata-abliterated-iq2_xs.json, unit uap-strata-abliterated-server.service, spec 12 / spec-min-p 0.7 / mtp_max 8 MTP per live /metrics 2026-11-13; switch away: ~/.config/uap/model-switch.sh strata|strata-swift). Raw GGUF shards declared here: the pack (strata-data/packs/abliterated-iq2_xs) is derived from them by tools/iq_pack.py and rebuilds from these files; shard 2 is byte-identical across the repo's tiers (verified 2026-10-28).",
|
|
63
79
|
"files": [
|
|
64
80
|
{
|
|
65
81
|
"path": "/home/cogtek/models/strata-data/models/abliterated-IQ2_XS/Qwen3.8-Flash-Next-GSQ-RCO-abliterated-IQ2_XS-00001-of-00002.gguf",
|
|
@@ -76,17 +92,75 @@
|
|
|
76
92
|
"sha256": "316b46f3a2dbd68c900f43136ab9449f9dcc3725dfd8c794847c204bc161e113"
|
|
77
93
|
}
|
|
78
94
|
]
|
|
95
|
+
},
|
|
96
|
+
"strata-unsloth-ud-iq4_xs-v31": {
|
|
97
|
+
"status": "unmeasured",
|
|
98
|
+
"engine": "strata",
|
|
99
|
+
"unit": "uap-strata-server",
|
|
100
|
+
"note": "The :8080 default arm 2026-11-11 -> 2026-11-13 (the PRISTINE base-model quant: Unsloth Dynamic v3.1 UD-IQ4_XS Qwen3.8-Flash-Next, 4 shards; config strata-unsloth-ud-iq4_xs-v31.json, spec 12 / spec-min-p 0.5 / mtp_max 4 MTP, 40 GiB resident expert budget + SSD streaming, int8 KV resident 32768, mmproj vision, 131k pool); the fleet switched back to strata-abliterated-iq2_xs on 2026-11-13, kept switchable via ~/.config/uap/model-switch.sh strata. Raw GGUF shards declared here: the pack (strata-data/packs/unsloth-ud-iq4_xs-v31) is derived from them by tools/iq_pack.py (see its conversions.json) and rebuilds from these files. Provenance is pinned (repo + repo-subfolder `file` + `rev`, from feature/376): bytes + sha256 below are the download-time .done markers, verified against shard 1 on 2026-11-12, so `uap models ensure` can re-fetch and verify from this pin.",
|
|
101
|
+
"files": [
|
|
102
|
+
{
|
|
103
|
+
"path": "/home/cogtek/models/strata-data/models/unsloth-UD-IQ4_XS-v31/Qwen3.8-Flash-Next-UD-IQ4_XS-00001-of-00004.gguf",
|
|
104
|
+
"bytes": 10946618,
|
|
105
|
+
"sha256": "93ff23a08e5cd7501f24fa16f5b415d5807d0175e04edd3ed6a7a41b601a4eef",
|
|
106
|
+
"repo": "unsloth/Qwen3.8-Flash-Next-GGUF",
|
|
107
|
+
"file": "UD-IQ4_XS/Qwen3.8-Flash-Next-UD-IQ4_XS-00001-of-00004.gguf",
|
|
108
|
+
"rev": "d96201194f103c906994f283f504209dc8d2a10c"
|
|
109
|
+
},
|
|
110
|
+
{
|
|
111
|
+
"path": "/home/cogtek/models/strata-data/models/unsloth-UD-IQ4_XS-v31/Qwen3.8-Flash-Next-UD-IQ4_XS-00002-of-00004.gguf",
|
|
112
|
+
"bytes": 49604491008,
|
|
113
|
+
"sha256": "13d5dc8b93d039e4cfade5a9af7037f10c97f7ec4e53f75bd2e9ae34aae54aeb",
|
|
114
|
+
"repo": "unsloth/Qwen3.8-Flash-Next-GGUF",
|
|
115
|
+
"file": "UD-IQ4_XS/Qwen3.8-Flash-Next-UD-IQ4_XS-00002-of-00004.gguf",
|
|
116
|
+
"rev": "d96201194f103c906994f283f504209dc8d2a10c"
|
|
117
|
+
},
|
|
118
|
+
{
|
|
119
|
+
"path": "/home/cogtek/models/strata-data/models/unsloth-UD-IQ4_XS-v31/Qwen3.8-Flash-Next-UD-IQ4_XS-00003-of-00004.gguf",
|
|
120
|
+
"bytes": 15256783680,
|
|
121
|
+
"sha256": "f6b9b5934908a2043732269dd740155d0f921d0763a1dcdd3452246311651f87",
|
|
122
|
+
"repo": "unsloth/Qwen3.8-Flash-Next-GGUF",
|
|
123
|
+
"file": "UD-IQ4_XS/Qwen3.8-Flash-Next-UD-IQ4_XS-00003-of-00004.gguf",
|
|
124
|
+
"rev": "d96201194f103c906994f283f504209dc8d2a10c"
|
|
125
|
+
},
|
|
126
|
+
{
|
|
127
|
+
"path": "/home/cogtek/models/strata-data/models/unsloth-UD-IQ4_XS-v31/Qwen3.8-Flash-Next-UD-IQ4_XS-00004-of-00004.gguf",
|
|
128
|
+
"bytes": 28800138432,
|
|
129
|
+
"sha256": "5de8b5dbf89409acc39468decafd4a4a95b13c29d88a14b0356107897829ec36",
|
|
130
|
+
"repo": "unsloth/Qwen3.8-Flash-Next-GGUF",
|
|
131
|
+
"file": "UD-IQ4_XS/Qwen3.8-Flash-Next-UD-IQ4_XS-00004-of-00004.gguf",
|
|
132
|
+
"rev": "d96201194f103c906994f283f504209dc8d2a10c"
|
|
133
|
+
}
|
|
134
|
+
]
|
|
79
135
|
}
|
|
80
136
|
},
|
|
81
|
-
"advertises": [
|
|
137
|
+
"advertises": [
|
|
138
|
+
"qwen3.8-flash-next-unsloth-ud-iq4_xs-v31",
|
|
139
|
+
"abliterated-iq2_xs",
|
|
140
|
+
"qwen3.8-flash-next-iq3_s",
|
|
141
|
+
"swift-1.5-iq3_s",
|
|
142
|
+
"gpt-5.1-codex"
|
|
143
|
+
]
|
|
82
144
|
},
|
|
83
145
|
"qwen3.8-27b": {
|
|
84
146
|
"display": "Qwen3.8 27B",
|
|
85
147
|
"engine": "llama.cpp",
|
|
86
|
-
"task_affinity": [
|
|
87
|
-
|
|
148
|
+
"task_affinity": [
|
|
149
|
+
"code",
|
|
150
|
+
"general"
|
|
151
|
+
],
|
|
152
|
+
"affinity": {
|
|
153
|
+
"device": [
|
|
154
|
+
"gpu0"
|
|
155
|
+
]
|
|
156
|
+
},
|
|
88
157
|
"endpoint": "http://127.0.0.1:8080",
|
|
89
|
-
"advertises": [
|
|
158
|
+
"advertises": [
|
|
159
|
+
"Qwen3.8-27B",
|
|
160
|
+
"qwen38-ampere-27b",
|
|
161
|
+
"qwen38-swift-27b",
|
|
162
|
+
"Qwen3.8-27B-Q4_K_M.gguf"
|
|
163
|
+
],
|
|
90
164
|
"configs": {
|
|
91
165
|
"llama-mtp": {
|
|
92
166
|
"status": "unmeasured",
|
|
@@ -153,7 +227,7 @@
|
|
|
153
227
|
"engine": "llama.cpp",
|
|
154
228
|
"unit": "uap-ampere-server",
|
|
155
229
|
"rails": 2,
|
|
156
|
-
"note": "ATX-IQ4_XS-M (4.56 bpw, llamAmpere-tuned recipe: bulk IQ4_XS, Q5_0/Q6_K attention K/V, Q8_0 GDN vectors) on llamAmpere v0.3, the engine+quant pair measured on this GPU class (3090/3090 Ti): 99.4 tok/s clean decode, 93.2 at 100K depth, exact p/q draft verification (output distribution = target's exactly). The 2-rail arm since 2026-10-30, measured live: -c 86016 total / -np 2 = 43008 per rail, q8_0 K + turbo3 V KV (the fused SM86 MMA path), MTP-3 spec decode (70.4% acceptance measured), engine 18,400 MiB with 3,072-3,156 MiB free beside the live desktop
|
|
230
|
+
"note": "ATX-IQ4_XS-M (4.56 bpw, llamAmpere-tuned recipe: bulk IQ4_XS, Q5_0/Q6_K attention K/V, Q8_0 GDN vectors) on llamAmpere v0.3, the engine+quant pair measured on this GPU class (3090/3090 Ti): 99.4 tok/s clean decode, 93.2 at 100K depth, exact p/q draft verification (output distribution = target's exactly). The 2-rail arm since 2026-10-30, measured live: -c 86016 total / -np 2 = 43008 per rail, q8_0 K + turbo3 V KV (the fused SM86 MMA path), MTP-3 spec decode (70.4% acceptance measured), engine 18,400 MiB with 3,072-3,156 MiB free beside the live desktop \u2014 ~52 tok/s solo decode engine-side at full 318W draw (2 rails buy concurrency, not aggregate). Unit ExecStartPre runs `uap models ensure-by-path` so a deleted GGUF re-downloads before the start fails.",
|
|
157
231
|
"files": [
|
|
158
232
|
{
|
|
159
233
|
"path": "/home/cogtek/models/atx-iq4xs-m/Qwen3.8-27B-ATX-4-XS.gguf",
|
|
@@ -182,7 +256,7 @@
|
|
|
182
256
|
"status": "unmeasured",
|
|
183
257
|
"engine": "openjet",
|
|
184
258
|
"unit": "uap-openjet-server",
|
|
185
|
-
"note": "OpenJet megakernel arm (:8080 default 2026-10-28 daytime, stood down evening for the strata switch-back). The Q4_K_M file under models/qwen3.8-27b-q4km went missing from disk after the arm was stood down
|
|
259
|
+
"note": "OpenJet megakernel arm (:8080 default 2026-10-28 daytime, stood down evening for the strata switch-back). The Q4_K_M file under models/qwen3.8-27b-q4km went missing from disk after the arm was stood down \u2014 restored-on-demand via this manifest. unsloth deleted the file from main on 2026-08-19, so the source is the OpenJet catalog's pinned revision (open-jet src/config.py QWEN38_27B_URL).",
|
|
186
260
|
"files": [
|
|
187
261
|
{
|
|
188
262
|
"path": "/home/cogtek/models/qwen3.8-27b-q4km/Qwen3.8-27B-Q4_K_M.gguf",
|
|
@@ -195,6 +269,580 @@
|
|
|
195
269
|
]
|
|
196
270
|
}
|
|
197
271
|
}
|
|
272
|
+
},
|
|
273
|
+
"edge0-145b": {
|
|
274
|
+
"display": "Edge0 145B",
|
|
275
|
+
"engine": "edge0-strata",
|
|
276
|
+
"unit": "uap-edge0-145b-server.service",
|
|
277
|
+
"service": "edge0-145b-server",
|
|
278
|
+
"launch": "/home/cogtek/dev/Edge0/edge0-infer/build/cpu/bin/edge0-infer serve --model /home/cogtek/models/Edge0-145B-q3k --port 8080 --alias edge0_145b --strata-arg --vram-reserve-mib --strata-arg 1200",
|
|
279
|
+
"endpoint": "http://127.0.0.1:8080",
|
|
280
|
+
"task_affinity": [
|
|
281
|
+
"code",
|
|
282
|
+
"general"
|
|
283
|
+
],
|
|
284
|
+
"affinity": {
|
|
285
|
+
"device": [
|
|
286
|
+
"gpu0"
|
|
287
|
+
]
|
|
288
|
+
},
|
|
289
|
+
"configs": {
|
|
290
|
+
"edge0-145b-q3k": {
|
|
291
|
+
"status": "unmeasured",
|
|
292
|
+
"engine": "edge0-strata",
|
|
293
|
+
"unit": "uap-edge0-145b-server.service",
|
|
294
|
+
"note": "Edge0 145B (latent-MoE derived from Qwen3.8-Flash-Next; Q3_K gate/up + IQ4_NL down, 48 layers x 512 experts) on the edge0-strata engine — the Edge0 stack's fork of Strata with latent-MoE kernels. Serves :8080 via edge0-infer serve (OpenAI + Anthropic endpoints), which delegates to edge0-strata build/strata + .venv python; the launch template (edge0-strata/edge0/models/edge0-145b.json) sets expert cache auto, prefill auto, spec 4 / min-p 0.5, MTP, int8 KV and 131k context, and the unit adds only --vram-reserve-mib 1200 for desktop co-tenancy on the 24 GiB 3090. Weights: HF Edge0/Edge0-145B-q3k at the pinned rev (d7de835cfce7): ~103 GiB folder — 40.1 GiB experts SSD-streamed with hot experts in VRAM, 51.2 GiB PLE table read direct from disk (ple-io=direct), du=q8 default (q4 file pinned too), MTP draft in mtp/. The manifest pins every serving file (both du variants, all 48 scales, tokenizer, MTP) so `uap models ensure` can restore any of them; LFS digests are the repo's own LFS oids (verified by huggingface_hub at download time), non-LFS digests computed locally 2026-11-13. inference-panel id edge0-145b (unit uap-edge0-145b-server.service); revert: panel switch or ~/.config/uap/model-switch.sh strata-abliterated. Upstream measured agent decode 121 tok/s (RTX 5090D) / 91.7 tok/s (4090) at Q3_K — expect 4090-class or lower on the 3090; quality above Qwen3.8-Flash-Next on DeepSWE/Terminal-Bench upstream. Run `uap models measure` to record this machine's footprint before it can load (fail-closed).",
|
|
295
|
+
"files": [
|
|
296
|
+
{
|
|
297
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/model.json",
|
|
298
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
299
|
+
"file": "model.json",
|
|
300
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
301
|
+
"bytes": 1531,
|
|
302
|
+
"sha256": "95b8280bf5151e11cf7b0622688172c9572dd8fd13558bb64d7e8edfc8c32ed4"
|
|
303
|
+
},
|
|
304
|
+
{
|
|
305
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/index.txt",
|
|
306
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
307
|
+
"file": "index.txt",
|
|
308
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
309
|
+
"bytes": 117464,
|
|
310
|
+
"sha256": "980366930f95ae1b8138652256c3b9be8925b02395d4f093b72410797675a349"
|
|
311
|
+
},
|
|
312
|
+
{
|
|
313
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/native_experts.txt",
|
|
314
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
315
|
+
"file": "native_experts.txt",
|
|
316
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
317
|
+
"bytes": 1463,
|
|
318
|
+
"sha256": "0de7b81744e89f722374842a73b8b26a8a6145c645493471a9c10a4f5f566e70"
|
|
319
|
+
},
|
|
320
|
+
{
|
|
321
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/experts.bin",
|
|
322
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
323
|
+
"file": "experts.bin",
|
|
324
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
325
|
+
"bytes": 40076574720,
|
|
326
|
+
"sha256": "947d15f3d69ecdb45d602a62bc84803451da779a22294e1db5b3568290b42879"
|
|
327
|
+
},
|
|
328
|
+
{
|
|
329
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/labels.bin",
|
|
330
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
331
|
+
"file": "labels.bin",
|
|
332
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
333
|
+
"bytes": 98304,
|
|
334
|
+
"sha256": "ce9b4a20b92c5c866e65d84a3884c104986522471cf11cc5f797a153affbcc06"
|
|
335
|
+
},
|
|
336
|
+
{
|
|
337
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/dense.bin",
|
|
338
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
339
|
+
"file": "dense.bin",
|
|
340
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
341
|
+
"bytes": 1538625024,
|
|
342
|
+
"sha256": "f9bcc7f36eac4e59a7b1cb1962e038ba12dc831a41fb193761f1d1e72657fb62"
|
|
343
|
+
},
|
|
344
|
+
{
|
|
345
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/profile.bin",
|
|
346
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
347
|
+
"file": "profile.bin",
|
|
348
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
349
|
+
"bytes": 196632,
|
|
350
|
+
"sha256": "543f8a364c8dbfef909ecd1a1d7adbc177d3db18ceecdca1fc568f3eaeb8c655"
|
|
351
|
+
},
|
|
352
|
+
{
|
|
353
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/latent_dense_q8.bin",
|
|
354
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
355
|
+
"file": "latent_dense_q8.bin",
|
|
356
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
357
|
+
"bytes": 3743416320,
|
|
358
|
+
"sha256": "fbb25caaa0dcf45ef62f1a4945dcf98a53a166934a3b461b166fbafa831d87ca"
|
|
359
|
+
},
|
|
360
|
+
{
|
|
361
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/latent_dense_q4.bin",
|
|
362
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
363
|
+
"file": "latent_dense_q4.bin",
|
|
364
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
365
|
+
"bytes": 1981808640,
|
|
366
|
+
"sha256": "e31e11f36c7e5a23c82550b524fd44fb28cfa821ef59fd688b8d2339137ab287"
|
|
367
|
+
},
|
|
368
|
+
{
|
|
369
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/model-skeleton.gguf",
|
|
370
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
371
|
+
"file": "model-skeleton.gguf",
|
|
372
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
373
|
+
"bytes": 5991569248,
|
|
374
|
+
"sha256": "ab49fbe7590cdcda45d7a2bcae4be0bd42c6a70b951d594a00eb158d7df0b72f"
|
|
375
|
+
},
|
|
376
|
+
{
|
|
377
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/ple-fp8.gguf",
|
|
378
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
379
|
+
"file": "ple-fp8.gguf",
|
|
380
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
381
|
+
"bytes": 51200246112,
|
|
382
|
+
"sha256": "c7615c59b060277ddf6f84729fa8f7469cfc82c7fb983d6e7eab5a2f7446f170"
|
|
383
|
+
},
|
|
384
|
+
{
|
|
385
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/mtp/dense.bin",
|
|
386
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
387
|
+
"file": "mtp/dense.bin",
|
|
388
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
389
|
+
"bytes": 116099072,
|
|
390
|
+
"sha256": "c724dc0b0822ada5d2977bf5bde821605feabaa64ea2e0045b67ca656329070a"
|
|
391
|
+
},
|
|
392
|
+
{
|
|
393
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/mtp/dense.txt",
|
|
394
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
395
|
+
"file": "mtp/dense.txt",
|
|
396
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
397
|
+
"bytes": 1851,
|
|
398
|
+
"sha256": "991456cf4eb266eb8a888c349e548d11bc640f5dbd9149439aa9786af0f732fd"
|
|
399
|
+
},
|
|
400
|
+
{
|
|
401
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/mtp/experts.bin",
|
|
402
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
403
|
+
"file": "mtp/experts.bin",
|
|
404
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
405
|
+
"bytes": 707788800,
|
|
406
|
+
"sha256": "09398406be61f1f54c93861f449e48b8df0bfccbc9ec9b2b7636775a6ea9244f"
|
|
407
|
+
},
|
|
408
|
+
{
|
|
409
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/mtp/draft_vocab.bin",
|
|
410
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
411
|
+
"file": "mtp/draft_vocab.bin",
|
|
412
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
413
|
+
"bytes": 425196,
|
|
414
|
+
"sha256": "b1e1d3a7a9e4bf862dcd5923ce661fb59bbd07907e594df5cf86a62ac235cb91"
|
|
415
|
+
},
|
|
416
|
+
{
|
|
417
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l0.f32",
|
|
418
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
419
|
+
"file": "scales/l0.f32",
|
|
420
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
421
|
+
"bytes": 7168,
|
|
422
|
+
"sha256": "ccfb45da4cb91bf0cc19d867472ddfb5c2175640046f6bfc754aa4e344855972"
|
|
423
|
+
},
|
|
424
|
+
{
|
|
425
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l1.f32",
|
|
426
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
427
|
+
"file": "scales/l1.f32",
|
|
428
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
429
|
+
"bytes": 7168,
|
|
430
|
+
"sha256": "73a4a4964c59fe2ae6eebd994f099fde0389ddf7c49fe5994ea372d7f826a5c6"
|
|
431
|
+
},
|
|
432
|
+
{
|
|
433
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l2.f32",
|
|
434
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
435
|
+
"file": "scales/l2.f32",
|
|
436
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
437
|
+
"bytes": 7168,
|
|
438
|
+
"sha256": "1e2ad898903d58779bddef0708f6b9b605376f451da7d4a3e092c75921abee0e"
|
|
439
|
+
},
|
|
440
|
+
{
|
|
441
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l3.f32",
|
|
442
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
443
|
+
"file": "scales/l3.f32",
|
|
444
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
445
|
+
"bytes": 7168,
|
|
446
|
+
"sha256": "5b4b1994bcb00a0e78150152ed4d3eba1c90203f3772fc9d3b01bb133661949a"
|
|
447
|
+
},
|
|
448
|
+
{
|
|
449
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l4.f32",
|
|
450
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
451
|
+
"file": "scales/l4.f32",
|
|
452
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
453
|
+
"bytes": 7168,
|
|
454
|
+
"sha256": "223b5a8a9fef3fb289915b985f0492db776be6ee4bc794af6bd6b40bc73f5951"
|
|
455
|
+
},
|
|
456
|
+
{
|
|
457
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l5.f32",
|
|
458
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
459
|
+
"file": "scales/l5.f32",
|
|
460
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
461
|
+
"bytes": 7168,
|
|
462
|
+
"sha256": "2b54008e9681de98a3590ab2d87eccda2f0ce4cfb6a99c91ddea2bfaf7ee9103"
|
|
463
|
+
},
|
|
464
|
+
{
|
|
465
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l6.f32",
|
|
466
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
467
|
+
"file": "scales/l6.f32",
|
|
468
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
469
|
+
"bytes": 7168,
|
|
470
|
+
"sha256": "7c0475b18ff380f4fc900ef308226b0a98ef855e793add4c1d149a3620a18f30"
|
|
471
|
+
},
|
|
472
|
+
{
|
|
473
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l7.f32",
|
|
474
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
475
|
+
"file": "scales/l7.f32",
|
|
476
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
477
|
+
"bytes": 7168,
|
|
478
|
+
"sha256": "ffb91ee9d1abf5ce63a07c1cc6f8182849700c9701705c3a386398fde797771b"
|
|
479
|
+
},
|
|
480
|
+
{
|
|
481
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l8.f32",
|
|
482
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
483
|
+
"file": "scales/l8.f32",
|
|
484
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
485
|
+
"bytes": 7168,
|
|
486
|
+
"sha256": "2ef5f3fdea5a3bc49e8b7cd93a41a793460da3685ed9d0ea10e9aae053c85477"
|
|
487
|
+
},
|
|
488
|
+
{
|
|
489
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l9.f32",
|
|
490
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
491
|
+
"file": "scales/l9.f32",
|
|
492
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
493
|
+
"bytes": 7168,
|
|
494
|
+
"sha256": "788cb1c638605ac71728913ce4fcb14c008dbbf46982c4c9c9d1f7af90e03ba5"
|
|
495
|
+
},
|
|
496
|
+
{
|
|
497
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l10.f32",
|
|
498
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
499
|
+
"file": "scales/l10.f32",
|
|
500
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
501
|
+
"bytes": 7168,
|
|
502
|
+
"sha256": "30d4046e0efc59b0ff81d74d1da05f41661d12f3cdc3cade317d646b15aef135"
|
|
503
|
+
},
|
|
504
|
+
{
|
|
505
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l11.f32",
|
|
506
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
507
|
+
"file": "scales/l11.f32",
|
|
508
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
509
|
+
"bytes": 7168,
|
|
510
|
+
"sha256": "d41bc1a5fc130916cea24416da58d6c3edac68c6a27f24719c885083aef13d8f"
|
|
511
|
+
},
|
|
512
|
+
{
|
|
513
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l12.f32",
|
|
514
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
515
|
+
"file": "scales/l12.f32",
|
|
516
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
517
|
+
"bytes": 7168,
|
|
518
|
+
"sha256": "877b3eaac39446fe23023a06a643a0783f6604e4a722dcd5ed40e6976cb68408"
|
|
519
|
+
},
|
|
520
|
+
{
|
|
521
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l13.f32",
|
|
522
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
523
|
+
"file": "scales/l13.f32",
|
|
524
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
525
|
+
"bytes": 7168,
|
|
526
|
+
"sha256": "b62631e02ddf5ae3be123d4385ae1d147a13253e250a4d645a4e5f77673aeaa7"
|
|
527
|
+
},
|
|
528
|
+
{
|
|
529
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l14.f32",
|
|
530
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
531
|
+
"file": "scales/l14.f32",
|
|
532
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
533
|
+
"bytes": 7168,
|
|
534
|
+
"sha256": "547f5bd37f52a0e913532c90fb7ad11658ac4fabc6b2a28be159c4e07b1e286b"
|
|
535
|
+
},
|
|
536
|
+
{
|
|
537
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l15.f32",
|
|
538
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
539
|
+
"file": "scales/l15.f32",
|
|
540
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
541
|
+
"bytes": 7168,
|
|
542
|
+
"sha256": "374222a8750c7aa5e72f5a2b05229521931be61f99778b841d571ca1b19ae4d4"
|
|
543
|
+
},
|
|
544
|
+
{
|
|
545
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l16.f32",
|
|
546
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
547
|
+
"file": "scales/l16.f32",
|
|
548
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
549
|
+
"bytes": 7168,
|
|
550
|
+
"sha256": "de4aaf5ddc41a6b961db1e786929aad539da114495867cc3ee2283f77cb10a0a"
|
|
551
|
+
},
|
|
552
|
+
{
|
|
553
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l17.f32",
|
|
554
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
555
|
+
"file": "scales/l17.f32",
|
|
556
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
557
|
+
"bytes": 7168,
|
|
558
|
+
"sha256": "c29c3976ce49f137147213a84dece24c4221dae5326f1a0cf45717b03624368b"
|
|
559
|
+
},
|
|
560
|
+
{
|
|
561
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l18.f32",
|
|
562
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
563
|
+
"file": "scales/l18.f32",
|
|
564
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
565
|
+
"bytes": 7168,
|
|
566
|
+
"sha256": "aaf7450624f01bb074bae633f17432841a3794c8efaa76202f3d259046d2f679"
|
|
567
|
+
},
|
|
568
|
+
{
|
|
569
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l19.f32",
|
|
570
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
571
|
+
"file": "scales/l19.f32",
|
|
572
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
573
|
+
"bytes": 7168,
|
|
574
|
+
"sha256": "81625f559d0218685104a168d32ac8cabfb271531a8e8e9c8282df25f2754884"
|
|
575
|
+
},
|
|
576
|
+
{
|
|
577
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l20.f32",
|
|
578
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
579
|
+
"file": "scales/l20.f32",
|
|
580
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
581
|
+
"bytes": 7168,
|
|
582
|
+
"sha256": "0effc90ca93708fc835a5fee3a51a9e4e8e28139e7d289e618bb160fbacc6de4"
|
|
583
|
+
},
|
|
584
|
+
{
|
|
585
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l21.f32",
|
|
586
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
587
|
+
"file": "scales/l21.f32",
|
|
588
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
589
|
+
"bytes": 7168,
|
|
590
|
+
"sha256": "4643ac0076f9df916418a8ea58ff70dfb5cd3ce8f8220686698dd1011ed70c48"
|
|
591
|
+
},
|
|
592
|
+
{
|
|
593
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l22.f32",
|
|
594
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
595
|
+
"file": "scales/l22.f32",
|
|
596
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
597
|
+
"bytes": 7168,
|
|
598
|
+
"sha256": "8c15e401e96b52144fab43ca05e312d9cd12a6ced369aff85b55bb4e8233e45d"
|
|
599
|
+
},
|
|
600
|
+
{
|
|
601
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l23.f32",
|
|
602
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
603
|
+
"file": "scales/l23.f32",
|
|
604
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
605
|
+
"bytes": 7168,
|
|
606
|
+
"sha256": "c737877b64bb32eb444eb497db2b1661c5f92bd35716a2b5ed57fa27c9c5bedc"
|
|
607
|
+
},
|
|
608
|
+
{
|
|
609
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l24.f32",
|
|
610
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
611
|
+
"file": "scales/l24.f32",
|
|
612
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
613
|
+
"bytes": 7168,
|
|
614
|
+
"sha256": "e708f5923e499b338c1e95b633eef61e75034ec68d8ea1a9579c159cdc372b5b"
|
|
615
|
+
},
|
|
616
|
+
{
|
|
617
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l25.f32",
|
|
618
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
619
|
+
"file": "scales/l25.f32",
|
|
620
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
621
|
+
"bytes": 7168,
|
|
622
|
+
"sha256": "f0ac87f861b8c85065e09fdef886f957f73e6ef7d42ef86d6bc6e4abe2daafd3"
|
|
623
|
+
},
|
|
624
|
+
{
|
|
625
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l26.f32",
|
|
626
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
627
|
+
"file": "scales/l26.f32",
|
|
628
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
629
|
+
"bytes": 7168,
|
|
630
|
+
"sha256": "14adbeb9d6dcb10dafe0912d897eb02e2aa6abdb9e0a56f377945697bee02630"
|
|
631
|
+
},
|
|
632
|
+
{
|
|
633
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l27.f32",
|
|
634
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
635
|
+
"file": "scales/l27.f32",
|
|
636
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
637
|
+
"bytes": 7168,
|
|
638
|
+
"sha256": "1003f7dbacce30b7d2e1d758cfb7057a742bdf648472fcbd63333234f076e515"
|
|
639
|
+
},
|
|
640
|
+
{
|
|
641
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l28.f32",
|
|
642
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
643
|
+
"file": "scales/l28.f32",
|
|
644
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
645
|
+
"bytes": 7168,
|
|
646
|
+
"sha256": "e912c6d85335fc9ec83635064e523b0efe19fd5ef4de4f299bddfd1203c33693"
|
|
647
|
+
},
|
|
648
|
+
{
|
|
649
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l29.f32",
|
|
650
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
651
|
+
"file": "scales/l29.f32",
|
|
652
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
653
|
+
"bytes": 7168,
|
|
654
|
+
"sha256": "8c186910f9afcecd23abdbaf2eee52f48360b114f4e47543a41922414db1fd5b"
|
|
655
|
+
},
|
|
656
|
+
{
|
|
657
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l30.f32",
|
|
658
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
659
|
+
"file": "scales/l30.f32",
|
|
660
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
661
|
+
"bytes": 7168,
|
|
662
|
+
"sha256": "b5cd2e55d0baeed12a8479feb5df830817821a1d00ff24050c451ecdf4913404"
|
|
663
|
+
},
|
|
664
|
+
{
|
|
665
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l31.f32",
|
|
666
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
667
|
+
"file": "scales/l31.f32",
|
|
668
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
669
|
+
"bytes": 7168,
|
|
670
|
+
"sha256": "b65fc3a4d56ca8ec2bd59b41ca63e0946ce30f27e4380477a79e7bcb868bef34"
|
|
671
|
+
},
|
|
672
|
+
{
|
|
673
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l32.f32",
|
|
674
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
675
|
+
"file": "scales/l32.f32",
|
|
676
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
677
|
+
"bytes": 7168,
|
|
678
|
+
"sha256": "4590579308a3973129592ced56ff24ed38f3bfe3ef4a8412c9dc2dbd40d170a7"
|
|
679
|
+
},
|
|
680
|
+
{
|
|
681
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l33.f32",
|
|
682
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
683
|
+
"file": "scales/l33.f32",
|
|
684
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
685
|
+
"bytes": 7168,
|
|
686
|
+
"sha256": "b8e28a6a3a4e68b935d8779bc43f0cca8a7c87e43cfe0cdcb9ce757631c5bd79"
|
|
687
|
+
},
|
|
688
|
+
{
|
|
689
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l34.f32",
|
|
690
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
691
|
+
"file": "scales/l34.f32",
|
|
692
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
693
|
+
"bytes": 7168,
|
|
694
|
+
"sha256": "bad3b4908fecfd92429cdc1b16c0b00314a8f843f9510b8a87afdd297a348c65"
|
|
695
|
+
},
|
|
696
|
+
{
|
|
697
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l35.f32",
|
|
698
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
699
|
+
"file": "scales/l35.f32",
|
|
700
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
701
|
+
"bytes": 7168,
|
|
702
|
+
"sha256": "2bc0b3d52dbc8ff6079cf978307698aede8a7ebb0084ba746f6164ce020c4fb1"
|
|
703
|
+
},
|
|
704
|
+
{
|
|
705
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l36.f32",
|
|
706
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
707
|
+
"file": "scales/l36.f32",
|
|
708
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
709
|
+
"bytes": 7168,
|
|
710
|
+
"sha256": "d6627bdbf25218a3dd429fa752a1ef882fd610eb23b66871b42b078713059356"
|
|
711
|
+
},
|
|
712
|
+
{
|
|
713
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l37.f32",
|
|
714
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
715
|
+
"file": "scales/l37.f32",
|
|
716
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
717
|
+
"bytes": 7168,
|
|
718
|
+
"sha256": "8afc17ed3b68756be8e5590ecbebb3e36179fabafb71901faf63e67ffe02f244"
|
|
719
|
+
},
|
|
720
|
+
{
|
|
721
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l38.f32",
|
|
722
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
723
|
+
"file": "scales/l38.f32",
|
|
724
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
725
|
+
"bytes": 7168,
|
|
726
|
+
"sha256": "445fbb2f2365079f06569460e6a905dad3e7f475d23af5ac2b967ab548e27c27"
|
|
727
|
+
},
|
|
728
|
+
{
|
|
729
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l39.f32",
|
|
730
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
731
|
+
"file": "scales/l39.f32",
|
|
732
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
733
|
+
"bytes": 7168,
|
|
734
|
+
"sha256": "020697f9289610c49dea1b47b72dd7f0f3746352c52b02f7d9d05db9011ddb07"
|
|
735
|
+
},
|
|
736
|
+
{
|
|
737
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l40.f32",
|
|
738
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
739
|
+
"file": "scales/l40.f32",
|
|
740
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
741
|
+
"bytes": 7168,
|
|
742
|
+
"sha256": "64b0d2d16992197dc7e2f0986f3d71ab0b744b4df0bdbab15bf08450c0919b00"
|
|
743
|
+
},
|
|
744
|
+
{
|
|
745
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l41.f32",
|
|
746
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
747
|
+
"file": "scales/l41.f32",
|
|
748
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
749
|
+
"bytes": 7168,
|
|
750
|
+
"sha256": "a75175bdc03d571f557d59aede6de9a0e734c61b03cfacc4ceb74c19b3787b7d"
|
|
751
|
+
},
|
|
752
|
+
{
|
|
753
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l42.f32",
|
|
754
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
755
|
+
"file": "scales/l42.f32",
|
|
756
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
757
|
+
"bytes": 7168,
|
|
758
|
+
"sha256": "e6596c611be63531fbf46fc73c1d1fcf86d38b09cfb27f97ee63571eea4be9cd"
|
|
759
|
+
},
|
|
760
|
+
{
|
|
761
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l43.f32",
|
|
762
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
763
|
+
"file": "scales/l43.f32",
|
|
764
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
765
|
+
"bytes": 7168,
|
|
766
|
+
"sha256": "e6e2227612938fac5320a81970215dac65bbee6e555573a5a3da7632adfe6087"
|
|
767
|
+
},
|
|
768
|
+
{
|
|
769
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l44.f32",
|
|
770
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
771
|
+
"file": "scales/l44.f32",
|
|
772
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
773
|
+
"bytes": 7168,
|
|
774
|
+
"sha256": "4187577ae6c484c700dfbf003a698aef48ce27fff9035ba40781abec05b58bf7"
|
|
775
|
+
},
|
|
776
|
+
{
|
|
777
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l45.f32",
|
|
778
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
779
|
+
"file": "scales/l45.f32",
|
|
780
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
781
|
+
"bytes": 7168,
|
|
782
|
+
"sha256": "270938e5d001b88e23a01d7c0f967ca165f1ac35adeff23f6401365d4d381430"
|
|
783
|
+
},
|
|
784
|
+
{
|
|
785
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l46.f32",
|
|
786
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
787
|
+
"file": "scales/l46.f32",
|
|
788
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
789
|
+
"bytes": 7168,
|
|
790
|
+
"sha256": "e3c18301cbfdf48eaa11cbb657f0696d3582aad5f68e2c9c27bafa3481b5a4e2"
|
|
791
|
+
},
|
|
792
|
+
{
|
|
793
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/scales/l47.f32",
|
|
794
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
795
|
+
"file": "scales/l47.f32",
|
|
796
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
797
|
+
"bytes": 7168,
|
|
798
|
+
"sha256": "552e22ee2bf8dbce16b68ac95a1146a141443a42e3326f7115d47dbe0ec9b845"
|
|
799
|
+
},
|
|
800
|
+
{
|
|
801
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/tokenizer/chat_template.jinja",
|
|
802
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
803
|
+
"file": "tokenizer/chat_template.jinja",
|
|
804
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
805
|
+
"bytes": 8952,
|
|
806
|
+
"sha256": "c3cf9e34abf4f9e36c2d72165aa9c132d3e2a725b6c2586aaa3a8af9d7a81041"
|
|
807
|
+
},
|
|
808
|
+
{
|
|
809
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/tokenizer/merges.txt",
|
|
810
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
811
|
+
"file": "tokenizer/merges.txt",
|
|
812
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
813
|
+
"bytes": 3353258,
|
|
814
|
+
"sha256": "4ba698529766c7a2f5052083ea10cdf2485577f5237a332ee291054eb11e3332"
|
|
815
|
+
},
|
|
816
|
+
{
|
|
817
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/tokenizer/token_type.json",
|
|
818
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
819
|
+
"file": "tokenizer/token_type.json",
|
|
820
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
821
|
+
"bytes": 744960,
|
|
822
|
+
"sha256": "5088c8c298fc06af8382ddb3b76c888703ac2634e82263b97eedcc2ad202738b"
|
|
823
|
+
},
|
|
824
|
+
{
|
|
825
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/tokenizer/tokenizer.json",
|
|
826
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
827
|
+
"file": "tokenizer/tokenizer.json",
|
|
828
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
829
|
+
"bytes": 541,
|
|
830
|
+
"sha256": "c0bd435efbc6c3b28964b47999afe16c4ad268c7c7c4e8cb3af410796a6f30bf"
|
|
831
|
+
},
|
|
832
|
+
{
|
|
833
|
+
"path": "/home/cogtek/models/Edge0-145B-q3k/tokenizer/vocab.json",
|
|
834
|
+
"repo": "Edge0/Edge0-145B-q3k",
|
|
835
|
+
"file": "tokenizer/vocab.json",
|
|
836
|
+
"rev": "d7de835cfce7d9bb22a3d5ebb7d1e8565d4ecf74",
|
|
837
|
+
"bytes": 5737005,
|
|
838
|
+
"sha256": "4ba64f0332abcfb0b600b7df1537d1e836e68009d5fb7cdb77339738dc6365c4"
|
|
839
|
+
}
|
|
840
|
+
]
|
|
841
|
+
}
|
|
842
|
+
},
|
|
843
|
+
"advertises": [
|
|
844
|
+
"edge0_145b"
|
|
845
|
+
]
|
|
198
846
|
}
|
|
199
847
|
}
|
|
200
848
|
}
|