File size: 4,097 Bytes
6a63ea0
3f8ea51
 
 
 
 
 
 
 
 
 
 
bbbe354
3f8ea51
 
 
 
6a63ea0
 
3f8ea51
 
 
 
 
 
bbbe354
3f8ea51
 
 
 
 
bbbe354
3f8ea51
 
 
 
 
bbbe354
3f8ea51
 
 
 
 
bbbe354
6a63ea0
bbbe354
3f8ea51
 
bbbe354
3f8ea51
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
{
  "candidate": "iter13_llamacpp_chat_fixed",
  "live_server": {
    "url": "http://127.0.0.1:8080",
    "model": "sjakek/LFM-2.5-8B-1B-hermes-ft:Q6KXL",
    "ctx": 65536
  },
  "mlx": {
    "8bit": {
      "path": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/model_variants/iter13_llamacpp_chat_fixed/mlx-8bit",
      "normal_eval": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/evals/iter13_mlx_8bit_normal_chat_regression.json",
      "tool_eval": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/evals/iter13_mlx_8bit_all_fixed_hermes.json"
    },
    "6bit": {
      "path": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/model_variants/iter13_llamacpp_chat_fixed/mlx-6bit",
      "normal_eval": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/evals/iter13_mlx_6bit_normal_chat_regression.json",
      "tool_eval": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/evals/iter13_mlx_6bit_all_fixed_hermes.json"
    }
  },
  "gguf": {
    "Q8KXL": {
      "path": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/gguf/iter13_llamacpp_chat_fixed/quants/LFM-2.5-8B-1B-Hermes-Tuned-Q8KXL.gguf",
      "size_bytes": 9279548576,
      "normal_eval": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/evals/iter13_gguf_Q8KXL_normal_chat_regression_v2.json",
      "tool_eval": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/evals/iter13_gguf_Q8KXL_all_fixed_hermes.json"
    },
    "Q6KXL": {
      "path": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/gguf/iter13_llamacpp_chat_fixed/quants/LFM-2.5-8B-1B-Hermes-Tuned-Q6KXL.gguf",
      "size_bytes": 7671360672,
      "normal_eval": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/evals/iter13_gguf_Q6KXL_normal_chat_regression_v2.json",
      "tool_eval": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/evals/iter13_gguf_Q6KXL_all_fixed_hermes.json"
    },
    "Q5KXL": {
      "path": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/gguf/iter13_llamacpp_chat_fixed/quants/LFM-2.5-8B-1B-Hermes-Tuned-Q5KXL.gguf",
      "size_bytes": 6928018592,
      "normal_eval": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/evals/iter13_gguf_Q5KXL_normal_chat_regression_v2.json",
      "tool_eval": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/evals/iter13_gguf_Q5KXL_all_fixed_hermes.json"
    },
    "Q4KXL": {
      "path": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/gguf/iter13_llamacpp_chat_fixed/quants/LFM-2.5-8B-1B-Hermes-Tuned-Q4KXL.gguf",
      "size_bytes": 6166588576,
      "normal_eval": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/evals/iter13_gguf_Q4KXL_normal_chat_regression_v3.json",
      "tool_eval": "/Users/jkooker/Documents/Codex/2026-05-28/get-this-new-model-up-and/release_work/lfm25-moe8-mac-trainer/artifacts/evals/iter13_gguf_Q4KXL_all_fixed_hermes.json"
    }
  },
  "notes": [
    "iter10/early GGUF quants were withdrawn for normal-chat early-stop regression",
    "iter13 adds targeted llama.cpp chat-retention repair while preserving fixed Hermes tool routing"
  ]
}