Instructions to use LEAP-LAB-KUS/jevkt-checkpoints with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use LEAP-LAB-KUS/jevkt-checkpoints with PEFT:
Task type is invalid.
- Notebooks
- Google Colab
- Kaggle
meta_only podA e15 (configs + train logs; weights not backed up)
Browse filesThis view is limited to 50 files because it contains too many changes. See raw diff
- meta_only/podA/e15/F1-pool-64ksched/ckpt-1000/adapter/adapter_config.json +57 -0
- meta_only/podA/e15/F1-pool-64ksched/ckpt-1000/adapter_vllm/adapter_config.json +58 -0
- meta_only/podA/e15/F1-pool-64ksched/ckpt-1000/adapter_vllm/decision_head.json +99 -0
- meta_only/podA/e15/F1-pool-64ksched/ckpt-1000/calibration.json +17 -0
- meta_only/podA/e15/F1-pool-64ksched/ckpt-1000/judge_config.json +145 -0
- meta_only/podA/e15/F1-pool-64ksched/ckpt-16000/adapter/adapter_config.json +57 -0
- meta_only/podA/e15/F1-pool-64ksched/ckpt-16000/adapter_vllm/adapter_config.json +58 -0
- meta_only/podA/e15/F1-pool-64ksched/ckpt-16000/adapter_vllm/decision_head.json +99 -0
- meta_only/podA/e15/F1-pool-64ksched/ckpt-16000/calibration.json +17 -0
- meta_only/podA/e15/F1-pool-64ksched/ckpt-16000/judge_config.json +145 -0
- meta_only/podA/e15/F1-pool-64ksched/ckpt-4000/adapter/adapter_config.json +57 -0
- meta_only/podA/e15/F1-pool-64ksched/ckpt-4000/adapter_vllm/adapter_config.json +58 -0
- meta_only/podA/e15/F1-pool-64ksched/ckpt-4000/adapter_vllm/decision_head.json +99 -0
- meta_only/podA/e15/F1-pool-64ksched/ckpt-4000/calibration.json +17 -0
- meta_only/podA/e15/F1-pool-64ksched/ckpt-4000/judge_config.json +145 -0
- meta_only/podA/e15/F1-pool-64ksched/train_log.jsonl +141 -0
- meta_only/podA/e15/F1-pool-ts1/ckpt-1000/adapter/adapter_config.json +57 -0
- meta_only/podA/e15/F1-pool-ts1/ckpt-1000/adapter_vllm/adapter_config.json +58 -0
- meta_only/podA/e15/F1-pool-ts1/ckpt-1000/adapter_vllm/decision_head.json +99 -0
- meta_only/podA/e15/F1-pool-ts1/ckpt-1000/calibration.json +17 -0
- meta_only/podA/e15/F1-pool-ts1/ckpt-1000/judge_config.json +146 -0
- meta_only/podA/e15/F1-pool-ts1/ckpt-16000/adapter/adapter_config.json +57 -0
- meta_only/podA/e15/F1-pool-ts1/ckpt-16000/adapter_vllm/adapter_config.json +58 -0
- meta_only/podA/e15/F1-pool-ts1/ckpt-16000/adapter_vllm/decision_head.json +99 -0
- meta_only/podA/e15/F1-pool-ts1/ckpt-16000/calibration.json +17 -0
- meta_only/podA/e15/F1-pool-ts1/ckpt-16000/judge_config.json +146 -0
- meta_only/podA/e15/F1-pool-ts1/ckpt-4000/adapter/adapter_config.json +57 -0
- meta_only/podA/e15/F1-pool-ts1/ckpt-4000/adapter_vllm/adapter_config.json +58 -0
- meta_only/podA/e15/F1-pool-ts1/ckpt-4000/adapter_vllm/decision_head.json +99 -0
- meta_only/podA/e15/F1-pool-ts1/ckpt-4000/calibration.json +17 -0
- meta_only/podA/e15/F1-pool-ts1/ckpt-4000/judge_config.json +146 -0
- meta_only/podA/e15/F1-pool-ts1/train_log.jsonl +55 -0
- meta_only/podA/e15/F1-pool-ts2/ckpt-1000/adapter/adapter_config.json +57 -0
- meta_only/podA/e15/F1-pool-ts2/ckpt-1000/adapter_vllm/adapter_config.json +58 -0
- meta_only/podA/e15/F1-pool-ts2/ckpt-1000/adapter_vllm/decision_head.json +99 -0
- meta_only/podA/e15/F1-pool-ts2/ckpt-1000/calibration.json +17 -0
- meta_only/podA/e15/F1-pool-ts2/ckpt-1000/judge_config.json +146 -0
- meta_only/podA/e15/F1-pool-ts2/ckpt-16000/adapter/adapter_config.json +57 -0
- meta_only/podA/e15/F1-pool-ts2/ckpt-16000/adapter_vllm/adapter_config.json +58 -0
- meta_only/podA/e15/F1-pool-ts2/ckpt-16000/adapter_vllm/decision_head.json +99 -0
- meta_only/podA/e15/F1-pool-ts2/ckpt-16000/calibration.json +17 -0
- meta_only/podA/e15/F1-pool-ts2/ckpt-16000/judge_config.json +146 -0
- meta_only/podA/e15/F1-pool-ts2/ckpt-4000/adapter/adapter_config.json +57 -0
- meta_only/podA/e15/F1-pool-ts2/ckpt-4000/adapter_vllm/adapter_config.json +58 -0
- meta_only/podA/e15/F1-pool-ts2/ckpt-4000/adapter_vllm/decision_head.json +99 -0
- meta_only/podA/e15/F1-pool-ts2/ckpt-4000/calibration.json +17 -0
- meta_only/podA/e15/F1-pool-ts2/ckpt-4000/judge_config.json +146 -0
- meta_only/podA/e15/F1-pool-ts2/train_log.jsonl +55 -0
- meta_only/podA/e15/F1-pool/ckpt-1000/adapter/adapter_config.json +57 -0
- meta_only/podA/e15/F1-pool/ckpt-1000/adapter_vllm/adapter_config.json +58 -0
meta_only/podA/e15/F1-pool-64ksched/ckpt-1000/adapter/adapter_config.json
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B",
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 32,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.05,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 16,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"up_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"out_proj",
|
| 41 |
+
"o_proj",
|
| 42 |
+
"in_proj_z",
|
| 43 |
+
"in_proj_qkv",
|
| 44 |
+
"k_proj",
|
| 45 |
+
"down_proj",
|
| 46 |
+
"q_proj",
|
| 47 |
+
"v_proj"
|
| 48 |
+
],
|
| 49 |
+
"target_parameters": null,
|
| 50 |
+
"task_type": null,
|
| 51 |
+
"trainable_token_indices": null,
|
| 52 |
+
"use_bdlora": null,
|
| 53 |
+
"use_dora": false,
|
| 54 |
+
"use_qalora": false,
|
| 55 |
+
"use_rslora": false,
|
| 56 |
+
"velora_config": null
|
| 57 |
+
}
|
meta_only/podA/e15/F1-pool-64ksched/ckpt-1000/adapter_vllm/adapter_config.json
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": null,
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 64.0,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.0,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 32,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"down_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"in_proj_qkv",
|
| 41 |
+
"in_proj_z",
|
| 42 |
+
"k_proj",
|
| 43 |
+
"lm_head",
|
| 44 |
+
"o_proj",
|
| 45 |
+
"out_proj",
|
| 46 |
+
"q_proj",
|
| 47 |
+
"up_proj",
|
| 48 |
+
"v_proj"
|
| 49 |
+
],
|
| 50 |
+
"target_parameters": null,
|
| 51 |
+
"task_type": null,
|
| 52 |
+
"trainable_token_indices": null,
|
| 53 |
+
"use_bdlora": null,
|
| 54 |
+
"use_dora": false,
|
| 55 |
+
"use_qalora": false,
|
| 56 |
+
"use_rslora": false,
|
| 57 |
+
"velora_config": null
|
| 58 |
+
}
|
meta_only/podA/e15/F1-pool-64ksched/ckpt-1000/adapter_vllm/decision_head.json
ADDED
|
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bias": [
|
| 3 |
+
0.0003320467076264322,
|
| 4 |
+
-0.00033229400287382305,
|
| 5 |
+
-0.0025878134183585644,
|
| 6 |
+
-0.0012615364976227283,
|
| 7 |
+
-0.00025007399381138384,
|
| 8 |
+
0.0019558551721274853,
|
| 9 |
+
0.0013559526996687055,
|
| 10 |
+
-0.0023348911199718714,
|
| 11 |
+
-0.0004082925443071872,
|
| 12 |
+
-0.0002078723773593083,
|
| 13 |
+
-1.2957353646925185e-05,
|
| 14 |
+
0.0001676026004133746,
|
| 15 |
+
0.0005569425993598998,
|
| 16 |
+
-0.001412136247381568,
|
| 17 |
+
-0.0028663526754826307,
|
| 18 |
+
-0.005061559844762087,
|
| 19 |
+
-0.00593328895047307,
|
| 20 |
+
-0.004319756291806698,
|
| 21 |
+
-0.004922541789710522,
|
| 22 |
+
-0.0049067651852965355,
|
| 23 |
+
-0.003943523392081261,
|
| 24 |
+
-0.003319827374070883,
|
| 25 |
+
-0.005581587553024292,
|
| 26 |
+
-0.0034480984322726727
|
| 27 |
+
],
|
| 28 |
+
"verbalizer_ids": [
|
| 29 |
+
3721,
|
| 30 |
+
1802,
|
| 31 |
+
15,
|
| 32 |
+
16,
|
| 33 |
+
17,
|
| 34 |
+
18,
|
| 35 |
+
19,
|
| 36 |
+
20,
|
| 37 |
+
32,
|
| 38 |
+
33,
|
| 39 |
+
34,
|
| 40 |
+
35,
|
| 41 |
+
36,
|
| 42 |
+
37,
|
| 43 |
+
38,
|
| 44 |
+
39,
|
| 45 |
+
40,
|
| 46 |
+
41,
|
| 47 |
+
42,
|
| 48 |
+
43,
|
| 49 |
+
44,
|
| 50 |
+
45,
|
| 51 |
+
46,
|
| 52 |
+
47
|
| 53 |
+
],
|
| 54 |
+
"slots": {
|
| 55 |
+
"num_slots": 24,
|
| 56 |
+
"ranges": {
|
| 57 |
+
"noul": [
|
| 58 |
+
0,
|
| 59 |
+
2
|
| 60 |
+
],
|
| 61 |
+
"score": [
|
| 62 |
+
2,
|
| 63 |
+
8
|
| 64 |
+
],
|
| 65 |
+
"choice": [
|
| 66 |
+
8,
|
| 67 |
+
24
|
| 68 |
+
]
|
| 69 |
+
},
|
| 70 |
+
"verbalizers": [
|
| 71 |
+
"false",
|
| 72 |
+
"true",
|
| 73 |
+
"0",
|
| 74 |
+
"1",
|
| 75 |
+
"2",
|
| 76 |
+
"3",
|
| 77 |
+
"4",
|
| 78 |
+
"5",
|
| 79 |
+
"A",
|
| 80 |
+
"B",
|
| 81 |
+
"C",
|
| 82 |
+
"D",
|
| 83 |
+
"E",
|
| 84 |
+
"F",
|
| 85 |
+
"G",
|
| 86 |
+
"H",
|
| 87 |
+
"I",
|
| 88 |
+
"J",
|
| 89 |
+
"K",
|
| 90 |
+
"L",
|
| 91 |
+
"M",
|
| 92 |
+
"N",
|
| 93 |
+
"O",
|
| 94 |
+
"P"
|
| 95 |
+
],
|
| 96 |
+
"template_version": "bare-v1"
|
| 97 |
+
},
|
| 98 |
+
"note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature"
|
| 99 |
+
}
|
meta_only/podA/e15/F1-pool-64ksched/ckpt-1000/calibration.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"version": 1,
|
| 3 |
+
"per_kind": {
|
| 4 |
+
"noul": 1.133391082896599,
|
| 5 |
+
"choice": 1.0,
|
| 6 |
+
"score": 1.0
|
| 7 |
+
},
|
| 8 |
+
"fit": {
|
| 9 |
+
"noul": {
|
| 10 |
+
"n": 2000,
|
| 11 |
+
"T": 1.133391082896599,
|
| 12 |
+
"ce_before": 0.5415979027748108,
|
| 13 |
+
"ce_after": 0.5397843718528748
|
| 14 |
+
}
|
| 15 |
+
},
|
| 16 |
+
"source": "models/e15/F1-pool/ckpt-1000"
|
| 17 |
+
}
|
meta_only/podA/e15/F1-pool-64ksched/ckpt-1000/judge_config.json
ADDED
|
@@ -0,0 +1,145 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"base_model_path": "/root/models/Qwen3.5-9B",
|
| 3 |
+
"hidden_size": 4096,
|
| 4 |
+
"slots": {
|
| 5 |
+
"num_slots": 24,
|
| 6 |
+
"ranges": {
|
| 7 |
+
"noul": [
|
| 8 |
+
0,
|
| 9 |
+
2
|
| 10 |
+
],
|
| 11 |
+
"score": [
|
| 12 |
+
2,
|
| 13 |
+
8
|
| 14 |
+
],
|
| 15 |
+
"choice": [
|
| 16 |
+
8,
|
| 17 |
+
24
|
| 18 |
+
]
|
| 19 |
+
},
|
| 20 |
+
"verbalizers": [
|
| 21 |
+
"false",
|
| 22 |
+
"true",
|
| 23 |
+
"0",
|
| 24 |
+
"1",
|
| 25 |
+
"2",
|
| 26 |
+
"3",
|
| 27 |
+
"4",
|
| 28 |
+
"5",
|
| 29 |
+
"A",
|
| 30 |
+
"B",
|
| 31 |
+
"C",
|
| 32 |
+
"D",
|
| 33 |
+
"E",
|
| 34 |
+
"F",
|
| 35 |
+
"G",
|
| 36 |
+
"H",
|
| 37 |
+
"I",
|
| 38 |
+
"J",
|
| 39 |
+
"K",
|
| 40 |
+
"L",
|
| 41 |
+
"M",
|
| 42 |
+
"N",
|
| 43 |
+
"O",
|
| 44 |
+
"P"
|
| 45 |
+
],
|
| 46 |
+
"template_version": "bare-v1"
|
| 47 |
+
},
|
| 48 |
+
"verbalizer_ids": [
|
| 49 |
+
3721,
|
| 50 |
+
1802,
|
| 51 |
+
15,
|
| 52 |
+
16,
|
| 53 |
+
17,
|
| 54 |
+
18,
|
| 55 |
+
19,
|
| 56 |
+
20,
|
| 57 |
+
32,
|
| 58 |
+
33,
|
| 59 |
+
34,
|
| 60 |
+
35,
|
| 61 |
+
36,
|
| 62 |
+
37,
|
| 63 |
+
38,
|
| 64 |
+
39,
|
| 65 |
+
40,
|
| 66 |
+
41,
|
| 67 |
+
42,
|
| 68 |
+
43,
|
| 69 |
+
44,
|
| 70 |
+
45,
|
| 71 |
+
46,
|
| 72 |
+
47
|
| 73 |
+
],
|
| 74 |
+
"kinds": [
|
| 75 |
+
"noul",
|
| 76 |
+
"choice",
|
| 77 |
+
"score"
|
| 78 |
+
],
|
| 79 |
+
"model_name": "jev-judge-qwen35-9b",
|
| 80 |
+
"model_version": "0.8.0",
|
| 81 |
+
"lm_head_included": true,
|
| 82 |
+
"weights_mode": "unmerged",
|
| 83 |
+
"adapter_subfolder": "adapter",
|
| 84 |
+
"lora": {
|
| 85 |
+
"r": 16,
|
| 86 |
+
"lora_alpha": 32,
|
| 87 |
+
"lora_dropout": 0.05,
|
| 88 |
+
"target_modules": [
|
| 89 |
+
"down_proj",
|
| 90 |
+
"gate_proj",
|
| 91 |
+
"in_proj_qkv",
|
| 92 |
+
"in_proj_z",
|
| 93 |
+
"k_proj",
|
| 94 |
+
"o_proj",
|
| 95 |
+
"out_proj",
|
| 96 |
+
"q_proj",
|
| 97 |
+
"up_proj",
|
| 98 |
+
"v_proj"
|
| 99 |
+
]
|
| 100 |
+
},
|
| 101 |
+
"trained_stage": "e15",
|
| 102 |
+
"source_checkpoint": "models/e15/F1-pool/ckpt-1000",
|
| 103 |
+
"softcap": null,
|
| 104 |
+
"e15": {
|
| 105 |
+
"data": "data/e15/F1-pool",
|
| 106 |
+
"data_manifest": {
|
| 107 |
+
"arm": "F1-pool",
|
| 108 |
+
"name": "F1-pool",
|
| 109 |
+
"train_sha256": "dad16f94a5659768b254a6c12538c01f9c64cd43d574a3435bcc70c9c1bac130",
|
| 110 |
+
"n_train": 64000
|
| 111 |
+
},
|
| 112 |
+
"init": "released",
|
| 113 |
+
"lr": 0.0001,
|
| 114 |
+
"lr_head": 0.0002,
|
| 115 |
+
"rows_per_step": 32,
|
| 116 |
+
"skip_rows": 0,
|
| 117 |
+
"epochs": 1,
|
| 118 |
+
"max_len": 3072,
|
| 119 |
+
"schedule": "warmup 0.03 + cosine to x0.05 over 2000 steps",
|
| 120 |
+
"total_rows": 64000,
|
| 121 |
+
"lora": {
|
| 122 |
+
"r": 16,
|
| 123 |
+
"lora_alpha": 32,
|
| 124 |
+
"lora_dropout": 0.05
|
| 125 |
+
},
|
| 126 |
+
"event": "ckpt",
|
| 127 |
+
"tag": "ckpt-1000",
|
| 128 |
+
"rows_seen": 1024,
|
| 129 |
+
"step": 32,
|
| 130 |
+
"tokens": 678504,
|
| 131 |
+
"tok_per_s": 593.3689456089205,
|
| 132 |
+
"val": {
|
| 133 |
+
"n": 2000,
|
| 134 |
+
"dropped_long": 0,
|
| 135 |
+
"noul_ce": 0.5415979027748108,
|
| 136 |
+
"kt_auc": 0.7622999963156115,
|
| 137 |
+
"kt_acc": 0.735,
|
| 138 |
+
"kt_f1": 0.8071324599708879,
|
| 139 |
+
"kt_ece": 0.03905150377750397,
|
| 140 |
+
"kt_nll": 0.5415978856135626
|
| 141 |
+
},
|
| 142 |
+
"dropped_long": 0,
|
| 143 |
+
"peak_gib": 18.233368396759033
|
| 144 |
+
}
|
| 145 |
+
}
|
meta_only/podA/e15/F1-pool-64ksched/ckpt-16000/adapter/adapter_config.json
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B",
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 32,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.05,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 16,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"up_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"out_proj",
|
| 41 |
+
"o_proj",
|
| 42 |
+
"in_proj_z",
|
| 43 |
+
"in_proj_qkv",
|
| 44 |
+
"k_proj",
|
| 45 |
+
"down_proj",
|
| 46 |
+
"q_proj",
|
| 47 |
+
"v_proj"
|
| 48 |
+
],
|
| 49 |
+
"target_parameters": null,
|
| 50 |
+
"task_type": null,
|
| 51 |
+
"trainable_token_indices": null,
|
| 52 |
+
"use_bdlora": null,
|
| 53 |
+
"use_dora": false,
|
| 54 |
+
"use_qalora": false,
|
| 55 |
+
"use_rslora": false,
|
| 56 |
+
"velora_config": null
|
| 57 |
+
}
|
meta_only/podA/e15/F1-pool-64ksched/ckpt-16000/adapter_vllm/adapter_config.json
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": null,
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 64.0,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.0,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 32,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"down_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"in_proj_qkv",
|
| 41 |
+
"in_proj_z",
|
| 42 |
+
"k_proj",
|
| 43 |
+
"lm_head",
|
| 44 |
+
"o_proj",
|
| 45 |
+
"out_proj",
|
| 46 |
+
"q_proj",
|
| 47 |
+
"up_proj",
|
| 48 |
+
"v_proj"
|
| 49 |
+
],
|
| 50 |
+
"target_parameters": null,
|
| 51 |
+
"task_type": null,
|
| 52 |
+
"trainable_token_indices": null,
|
| 53 |
+
"use_bdlora": null,
|
| 54 |
+
"use_dora": false,
|
| 55 |
+
"use_qalora": false,
|
| 56 |
+
"use_rslora": false,
|
| 57 |
+
"velora_config": null
|
| 58 |
+
}
|
meta_only/podA/e15/F1-pool-64ksched/ckpt-16000/adapter_vllm/decision_head.json
ADDED
|
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bias": [
|
| 3 |
+
-0.00018585505313239992,
|
| 4 |
+
0.00018559396266937256,
|
| 5 |
+
-0.0025878134183585644,
|
| 6 |
+
-0.0012615364976227283,
|
| 7 |
+
-0.00025007399381138384,
|
| 8 |
+
0.0019558551721274853,
|
| 9 |
+
0.0013559526996687055,
|
| 10 |
+
-0.0023348911199718714,
|
| 11 |
+
-0.0004082925443071872,
|
| 12 |
+
-0.0002078723773593083,
|
| 13 |
+
-1.2957353646925185e-05,
|
| 14 |
+
0.0001676026004133746,
|
| 15 |
+
0.0005569425993598998,
|
| 16 |
+
-0.001412136247381568,
|
| 17 |
+
-0.0028663526754826307,
|
| 18 |
+
-0.005061559844762087,
|
| 19 |
+
-0.00593328895047307,
|
| 20 |
+
-0.004319756291806698,
|
| 21 |
+
-0.004922541789710522,
|
| 22 |
+
-0.0049067651852965355,
|
| 23 |
+
-0.003943523392081261,
|
| 24 |
+
-0.003319827374070883,
|
| 25 |
+
-0.005581587553024292,
|
| 26 |
+
-0.0034480984322726727
|
| 27 |
+
],
|
| 28 |
+
"verbalizer_ids": [
|
| 29 |
+
3721,
|
| 30 |
+
1802,
|
| 31 |
+
15,
|
| 32 |
+
16,
|
| 33 |
+
17,
|
| 34 |
+
18,
|
| 35 |
+
19,
|
| 36 |
+
20,
|
| 37 |
+
32,
|
| 38 |
+
33,
|
| 39 |
+
34,
|
| 40 |
+
35,
|
| 41 |
+
36,
|
| 42 |
+
37,
|
| 43 |
+
38,
|
| 44 |
+
39,
|
| 45 |
+
40,
|
| 46 |
+
41,
|
| 47 |
+
42,
|
| 48 |
+
43,
|
| 49 |
+
44,
|
| 50 |
+
45,
|
| 51 |
+
46,
|
| 52 |
+
47
|
| 53 |
+
],
|
| 54 |
+
"slots": {
|
| 55 |
+
"num_slots": 24,
|
| 56 |
+
"ranges": {
|
| 57 |
+
"noul": [
|
| 58 |
+
0,
|
| 59 |
+
2
|
| 60 |
+
],
|
| 61 |
+
"score": [
|
| 62 |
+
2,
|
| 63 |
+
8
|
| 64 |
+
],
|
| 65 |
+
"choice": [
|
| 66 |
+
8,
|
| 67 |
+
24
|
| 68 |
+
]
|
| 69 |
+
},
|
| 70 |
+
"verbalizers": [
|
| 71 |
+
"false",
|
| 72 |
+
"true",
|
| 73 |
+
"0",
|
| 74 |
+
"1",
|
| 75 |
+
"2",
|
| 76 |
+
"3",
|
| 77 |
+
"4",
|
| 78 |
+
"5",
|
| 79 |
+
"A",
|
| 80 |
+
"B",
|
| 81 |
+
"C",
|
| 82 |
+
"D",
|
| 83 |
+
"E",
|
| 84 |
+
"F",
|
| 85 |
+
"G",
|
| 86 |
+
"H",
|
| 87 |
+
"I",
|
| 88 |
+
"J",
|
| 89 |
+
"K",
|
| 90 |
+
"L",
|
| 91 |
+
"M",
|
| 92 |
+
"N",
|
| 93 |
+
"O",
|
| 94 |
+
"P"
|
| 95 |
+
],
|
| 96 |
+
"template_version": "bare-v1"
|
| 97 |
+
},
|
| 98 |
+
"note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature"
|
| 99 |
+
}
|
meta_only/podA/e15/F1-pool-64ksched/ckpt-16000/calibration.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"version": 1,
|
| 3 |
+
"per_kind": {
|
| 4 |
+
"noul": 1.1862171243881898,
|
| 5 |
+
"choice": 1.0,
|
| 6 |
+
"score": 1.0
|
| 7 |
+
},
|
| 8 |
+
"fit": {
|
| 9 |
+
"noul": {
|
| 10 |
+
"n": 2000,
|
| 11 |
+
"T": 1.1862171243881898,
|
| 12 |
+
"ce_before": 0.5415951013565063,
|
| 13 |
+
"ce_after": 0.53856360912323
|
| 14 |
+
}
|
| 15 |
+
},
|
| 16 |
+
"source": "models/e15/F1-pool/ckpt-16000"
|
| 17 |
+
}
|
meta_only/podA/e15/F1-pool-64ksched/ckpt-16000/judge_config.json
ADDED
|
@@ -0,0 +1,145 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"base_model_path": "/root/models/Qwen3.5-9B",
|
| 3 |
+
"hidden_size": 4096,
|
| 4 |
+
"slots": {
|
| 5 |
+
"num_slots": 24,
|
| 6 |
+
"ranges": {
|
| 7 |
+
"noul": [
|
| 8 |
+
0,
|
| 9 |
+
2
|
| 10 |
+
],
|
| 11 |
+
"score": [
|
| 12 |
+
2,
|
| 13 |
+
8
|
| 14 |
+
],
|
| 15 |
+
"choice": [
|
| 16 |
+
8,
|
| 17 |
+
24
|
| 18 |
+
]
|
| 19 |
+
},
|
| 20 |
+
"verbalizers": [
|
| 21 |
+
"false",
|
| 22 |
+
"true",
|
| 23 |
+
"0",
|
| 24 |
+
"1",
|
| 25 |
+
"2",
|
| 26 |
+
"3",
|
| 27 |
+
"4",
|
| 28 |
+
"5",
|
| 29 |
+
"A",
|
| 30 |
+
"B",
|
| 31 |
+
"C",
|
| 32 |
+
"D",
|
| 33 |
+
"E",
|
| 34 |
+
"F",
|
| 35 |
+
"G",
|
| 36 |
+
"H",
|
| 37 |
+
"I",
|
| 38 |
+
"J",
|
| 39 |
+
"K",
|
| 40 |
+
"L",
|
| 41 |
+
"M",
|
| 42 |
+
"N",
|
| 43 |
+
"O",
|
| 44 |
+
"P"
|
| 45 |
+
],
|
| 46 |
+
"template_version": "bare-v1"
|
| 47 |
+
},
|
| 48 |
+
"verbalizer_ids": [
|
| 49 |
+
3721,
|
| 50 |
+
1802,
|
| 51 |
+
15,
|
| 52 |
+
16,
|
| 53 |
+
17,
|
| 54 |
+
18,
|
| 55 |
+
19,
|
| 56 |
+
20,
|
| 57 |
+
32,
|
| 58 |
+
33,
|
| 59 |
+
34,
|
| 60 |
+
35,
|
| 61 |
+
36,
|
| 62 |
+
37,
|
| 63 |
+
38,
|
| 64 |
+
39,
|
| 65 |
+
40,
|
| 66 |
+
41,
|
| 67 |
+
42,
|
| 68 |
+
43,
|
| 69 |
+
44,
|
| 70 |
+
45,
|
| 71 |
+
46,
|
| 72 |
+
47
|
| 73 |
+
],
|
| 74 |
+
"kinds": [
|
| 75 |
+
"noul",
|
| 76 |
+
"choice",
|
| 77 |
+
"score"
|
| 78 |
+
],
|
| 79 |
+
"model_name": "jev-judge-qwen35-9b",
|
| 80 |
+
"model_version": "0.8.0",
|
| 81 |
+
"lm_head_included": true,
|
| 82 |
+
"weights_mode": "unmerged",
|
| 83 |
+
"adapter_subfolder": "adapter",
|
| 84 |
+
"lora": {
|
| 85 |
+
"r": 16,
|
| 86 |
+
"lora_alpha": 32,
|
| 87 |
+
"lora_dropout": 0.05,
|
| 88 |
+
"target_modules": [
|
| 89 |
+
"down_proj",
|
| 90 |
+
"gate_proj",
|
| 91 |
+
"in_proj_qkv",
|
| 92 |
+
"in_proj_z",
|
| 93 |
+
"k_proj",
|
| 94 |
+
"o_proj",
|
| 95 |
+
"out_proj",
|
| 96 |
+
"q_proj",
|
| 97 |
+
"up_proj",
|
| 98 |
+
"v_proj"
|
| 99 |
+
]
|
| 100 |
+
},
|
| 101 |
+
"trained_stage": "e15",
|
| 102 |
+
"source_checkpoint": "models/e15/F1-pool/ckpt-16000",
|
| 103 |
+
"softcap": null,
|
| 104 |
+
"e15": {
|
| 105 |
+
"data": "data/e15/F1-pool",
|
| 106 |
+
"data_manifest": {
|
| 107 |
+
"arm": "F1-pool",
|
| 108 |
+
"name": "F1-pool",
|
| 109 |
+
"train_sha256": "dad16f94a5659768b254a6c12538c01f9c64cd43d574a3435bcc70c9c1bac130",
|
| 110 |
+
"n_train": 64000
|
| 111 |
+
},
|
| 112 |
+
"init": "released",
|
| 113 |
+
"lr": 0.0001,
|
| 114 |
+
"lr_head": 0.0002,
|
| 115 |
+
"rows_per_step": 32,
|
| 116 |
+
"skip_rows": 0,
|
| 117 |
+
"epochs": 1,
|
| 118 |
+
"max_len": 3072,
|
| 119 |
+
"schedule": "warmup 0.03 + cosine to x0.05 over 2000 steps",
|
| 120 |
+
"total_rows": 64000,
|
| 121 |
+
"lora": {
|
| 122 |
+
"r": 16,
|
| 123 |
+
"lora_alpha": 32,
|
| 124 |
+
"lora_dropout": 0.05
|
| 125 |
+
},
|
| 126 |
+
"event": "ckpt",
|
| 127 |
+
"tag": "ckpt-16000",
|
| 128 |
+
"rows_seen": 16000,
|
| 129 |
+
"step": 500,
|
| 130 |
+
"tokens": 10544036,
|
| 131 |
+
"tok_per_s": 839.0346341733006,
|
| 132 |
+
"val": {
|
| 133 |
+
"n": 2000,
|
| 134 |
+
"dropped_long": 0,
|
| 135 |
+
"noul_ce": 0.5415951013565063,
|
| 136 |
+
"kt_auc": 0.7670316444319398,
|
| 137 |
+
"kt_acc": 0.735,
|
| 138 |
+
"kt_f1": 0.8067104303428154,
|
| 139 |
+
"kt_ece": 0.03874686068296433,
|
| 140 |
+
"kt_nll": 0.5415951018183259
|
| 141 |
+
},
|
| 142 |
+
"dropped_long": 0,
|
| 143 |
+
"peak_gib": 18.23592233657837
|
| 144 |
+
}
|
| 145 |
+
}
|
meta_only/podA/e15/F1-pool-64ksched/ckpt-4000/adapter/adapter_config.json
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B",
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 32,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.05,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 16,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"up_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"out_proj",
|
| 41 |
+
"o_proj",
|
| 42 |
+
"in_proj_z",
|
| 43 |
+
"in_proj_qkv",
|
| 44 |
+
"k_proj",
|
| 45 |
+
"down_proj",
|
| 46 |
+
"q_proj",
|
| 47 |
+
"v_proj"
|
| 48 |
+
],
|
| 49 |
+
"target_parameters": null,
|
| 50 |
+
"task_type": null,
|
| 51 |
+
"trainable_token_indices": null,
|
| 52 |
+
"use_bdlora": null,
|
| 53 |
+
"use_dora": false,
|
| 54 |
+
"use_qalora": false,
|
| 55 |
+
"use_rslora": false,
|
| 56 |
+
"velora_config": null
|
| 57 |
+
}
|
meta_only/podA/e15/F1-pool-64ksched/ckpt-4000/adapter_vllm/adapter_config.json
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": null,
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 64.0,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.0,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 32,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"down_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"in_proj_qkv",
|
| 41 |
+
"in_proj_z",
|
| 42 |
+
"k_proj",
|
| 43 |
+
"lm_head",
|
| 44 |
+
"o_proj",
|
| 45 |
+
"out_proj",
|
| 46 |
+
"q_proj",
|
| 47 |
+
"up_proj",
|
| 48 |
+
"v_proj"
|
| 49 |
+
],
|
| 50 |
+
"target_parameters": null,
|
| 51 |
+
"task_type": null,
|
| 52 |
+
"trainable_token_indices": null,
|
| 53 |
+
"use_bdlora": null,
|
| 54 |
+
"use_dora": false,
|
| 55 |
+
"use_qalora": false,
|
| 56 |
+
"use_rslora": false,
|
| 57 |
+
"velora_config": null
|
| 58 |
+
}
|
meta_only/podA/e15/F1-pool-64ksched/ckpt-4000/adapter_vllm/decision_head.json
ADDED
|
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bias": [
|
| 3 |
+
0.00044177754898555577,
|
| 4 |
+
-0.0004420274344738573,
|
| 5 |
+
-0.0025878134183585644,
|
| 6 |
+
-0.0012615364976227283,
|
| 7 |
+
-0.00025007399381138384,
|
| 8 |
+
0.0019558551721274853,
|
| 9 |
+
0.0013559526996687055,
|
| 10 |
+
-0.0023348911199718714,
|
| 11 |
+
-0.0004082925443071872,
|
| 12 |
+
-0.0002078723773593083,
|
| 13 |
+
-1.2957353646925185e-05,
|
| 14 |
+
0.0001676026004133746,
|
| 15 |
+
0.0005569425993598998,
|
| 16 |
+
-0.001412136247381568,
|
| 17 |
+
-0.0028663526754826307,
|
| 18 |
+
-0.005061559844762087,
|
| 19 |
+
-0.00593328895047307,
|
| 20 |
+
-0.004319756291806698,
|
| 21 |
+
-0.004922541789710522,
|
| 22 |
+
-0.0049067651852965355,
|
| 23 |
+
-0.003943523392081261,
|
| 24 |
+
-0.003319827374070883,
|
| 25 |
+
-0.005581587553024292,
|
| 26 |
+
-0.0034480984322726727
|
| 27 |
+
],
|
| 28 |
+
"verbalizer_ids": [
|
| 29 |
+
3721,
|
| 30 |
+
1802,
|
| 31 |
+
15,
|
| 32 |
+
16,
|
| 33 |
+
17,
|
| 34 |
+
18,
|
| 35 |
+
19,
|
| 36 |
+
20,
|
| 37 |
+
32,
|
| 38 |
+
33,
|
| 39 |
+
34,
|
| 40 |
+
35,
|
| 41 |
+
36,
|
| 42 |
+
37,
|
| 43 |
+
38,
|
| 44 |
+
39,
|
| 45 |
+
40,
|
| 46 |
+
41,
|
| 47 |
+
42,
|
| 48 |
+
43,
|
| 49 |
+
44,
|
| 50 |
+
45,
|
| 51 |
+
46,
|
| 52 |
+
47
|
| 53 |
+
],
|
| 54 |
+
"slots": {
|
| 55 |
+
"num_slots": 24,
|
| 56 |
+
"ranges": {
|
| 57 |
+
"noul": [
|
| 58 |
+
0,
|
| 59 |
+
2
|
| 60 |
+
],
|
| 61 |
+
"score": [
|
| 62 |
+
2,
|
| 63 |
+
8
|
| 64 |
+
],
|
| 65 |
+
"choice": [
|
| 66 |
+
8,
|
| 67 |
+
24
|
| 68 |
+
]
|
| 69 |
+
},
|
| 70 |
+
"verbalizers": [
|
| 71 |
+
"false",
|
| 72 |
+
"true",
|
| 73 |
+
"0",
|
| 74 |
+
"1",
|
| 75 |
+
"2",
|
| 76 |
+
"3",
|
| 77 |
+
"4",
|
| 78 |
+
"5",
|
| 79 |
+
"A",
|
| 80 |
+
"B",
|
| 81 |
+
"C",
|
| 82 |
+
"D",
|
| 83 |
+
"E",
|
| 84 |
+
"F",
|
| 85 |
+
"G",
|
| 86 |
+
"H",
|
| 87 |
+
"I",
|
| 88 |
+
"J",
|
| 89 |
+
"K",
|
| 90 |
+
"L",
|
| 91 |
+
"M",
|
| 92 |
+
"N",
|
| 93 |
+
"O",
|
| 94 |
+
"P"
|
| 95 |
+
],
|
| 96 |
+
"template_version": "bare-v1"
|
| 97 |
+
},
|
| 98 |
+
"note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature"
|
| 99 |
+
}
|
meta_only/podA/e15/F1-pool-64ksched/ckpt-4000/calibration.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"version": 1,
|
| 3 |
+
"per_kind": {
|
| 4 |
+
"noul": 0.7319877409990786,
|
| 5 |
+
"choice": 1.0,
|
| 6 |
+
"score": 1.0
|
| 7 |
+
},
|
| 8 |
+
"fit": {
|
| 9 |
+
"noul": {
|
| 10 |
+
"n": 2000,
|
| 11 |
+
"T": 0.7319877409990786,
|
| 12 |
+
"ce_before": 0.5827071070671082,
|
| 13 |
+
"ce_after": 0.575294554233551
|
| 14 |
+
}
|
| 15 |
+
},
|
| 16 |
+
"source": "models/e15/F1-pool/ckpt-4000"
|
| 17 |
+
}
|
meta_only/podA/e15/F1-pool-64ksched/ckpt-4000/judge_config.json
ADDED
|
@@ -0,0 +1,145 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"base_model_path": "/root/models/Qwen3.5-9B",
|
| 3 |
+
"hidden_size": 4096,
|
| 4 |
+
"slots": {
|
| 5 |
+
"num_slots": 24,
|
| 6 |
+
"ranges": {
|
| 7 |
+
"noul": [
|
| 8 |
+
0,
|
| 9 |
+
2
|
| 10 |
+
],
|
| 11 |
+
"score": [
|
| 12 |
+
2,
|
| 13 |
+
8
|
| 14 |
+
],
|
| 15 |
+
"choice": [
|
| 16 |
+
8,
|
| 17 |
+
24
|
| 18 |
+
]
|
| 19 |
+
},
|
| 20 |
+
"verbalizers": [
|
| 21 |
+
"false",
|
| 22 |
+
"true",
|
| 23 |
+
"0",
|
| 24 |
+
"1",
|
| 25 |
+
"2",
|
| 26 |
+
"3",
|
| 27 |
+
"4",
|
| 28 |
+
"5",
|
| 29 |
+
"A",
|
| 30 |
+
"B",
|
| 31 |
+
"C",
|
| 32 |
+
"D",
|
| 33 |
+
"E",
|
| 34 |
+
"F",
|
| 35 |
+
"G",
|
| 36 |
+
"H",
|
| 37 |
+
"I",
|
| 38 |
+
"J",
|
| 39 |
+
"K",
|
| 40 |
+
"L",
|
| 41 |
+
"M",
|
| 42 |
+
"N",
|
| 43 |
+
"O",
|
| 44 |
+
"P"
|
| 45 |
+
],
|
| 46 |
+
"template_version": "bare-v1"
|
| 47 |
+
},
|
| 48 |
+
"verbalizer_ids": [
|
| 49 |
+
3721,
|
| 50 |
+
1802,
|
| 51 |
+
15,
|
| 52 |
+
16,
|
| 53 |
+
17,
|
| 54 |
+
18,
|
| 55 |
+
19,
|
| 56 |
+
20,
|
| 57 |
+
32,
|
| 58 |
+
33,
|
| 59 |
+
34,
|
| 60 |
+
35,
|
| 61 |
+
36,
|
| 62 |
+
37,
|
| 63 |
+
38,
|
| 64 |
+
39,
|
| 65 |
+
40,
|
| 66 |
+
41,
|
| 67 |
+
42,
|
| 68 |
+
43,
|
| 69 |
+
44,
|
| 70 |
+
45,
|
| 71 |
+
46,
|
| 72 |
+
47
|
| 73 |
+
],
|
| 74 |
+
"kinds": [
|
| 75 |
+
"noul",
|
| 76 |
+
"choice",
|
| 77 |
+
"score"
|
| 78 |
+
],
|
| 79 |
+
"model_name": "jev-judge-qwen35-9b",
|
| 80 |
+
"model_version": "0.8.0",
|
| 81 |
+
"lm_head_included": true,
|
| 82 |
+
"weights_mode": "unmerged",
|
| 83 |
+
"adapter_subfolder": "adapter",
|
| 84 |
+
"lora": {
|
| 85 |
+
"r": 16,
|
| 86 |
+
"lora_alpha": 32,
|
| 87 |
+
"lora_dropout": 0.05,
|
| 88 |
+
"target_modules": [
|
| 89 |
+
"down_proj",
|
| 90 |
+
"gate_proj",
|
| 91 |
+
"in_proj_qkv",
|
| 92 |
+
"in_proj_z",
|
| 93 |
+
"k_proj",
|
| 94 |
+
"o_proj",
|
| 95 |
+
"out_proj",
|
| 96 |
+
"q_proj",
|
| 97 |
+
"up_proj",
|
| 98 |
+
"v_proj"
|
| 99 |
+
]
|
| 100 |
+
},
|
| 101 |
+
"trained_stage": "e15",
|
| 102 |
+
"source_checkpoint": "models/e15/F1-pool/ckpt-4000",
|
| 103 |
+
"softcap": null,
|
| 104 |
+
"e15": {
|
| 105 |
+
"data": "data/e15/F1-pool",
|
| 106 |
+
"data_manifest": {
|
| 107 |
+
"arm": "F1-pool",
|
| 108 |
+
"name": "F1-pool",
|
| 109 |
+
"train_sha256": "dad16f94a5659768b254a6c12538c01f9c64cd43d574a3435bcc70c9c1bac130",
|
| 110 |
+
"n_train": 64000
|
| 111 |
+
},
|
| 112 |
+
"init": "released",
|
| 113 |
+
"lr": 0.0001,
|
| 114 |
+
"lr_head": 0.0002,
|
| 115 |
+
"rows_per_step": 32,
|
| 116 |
+
"skip_rows": 0,
|
| 117 |
+
"epochs": 1,
|
| 118 |
+
"max_len": 3072,
|
| 119 |
+
"schedule": "warmup 0.03 + cosine to x0.05 over 2000 steps",
|
| 120 |
+
"total_rows": 64000,
|
| 121 |
+
"lora": {
|
| 122 |
+
"r": 16,
|
| 123 |
+
"lora_alpha": 32,
|
| 124 |
+
"lora_dropout": 0.05
|
| 125 |
+
},
|
| 126 |
+
"event": "ckpt",
|
| 127 |
+
"tag": "ckpt-4000",
|
| 128 |
+
"rows_seen": 4000,
|
| 129 |
+
"step": 125,
|
| 130 |
+
"tokens": 2631181,
|
| 131 |
+
"tok_per_s": 720.9349008307702,
|
| 132 |
+
"val": {
|
| 133 |
+
"n": 2000,
|
| 134 |
+
"dropped_long": 0,
|
| 135 |
+
"noul_ce": 0.5827071070671082,
|
| 136 |
+
"kt_auc": 0.7610473041998681,
|
| 137 |
+
"kt_acc": 0.7095,
|
| 138 |
+
"kt_f1": 0.7715296893432954,
|
| 139 |
+
"kt_ece": 0.12595178052783013,
|
| 140 |
+
"kt_nll": 0.5827070810943842
|
| 141 |
+
},
|
| 142 |
+
"dropped_long": 0,
|
| 143 |
+
"peak_gib": 18.233368396759033
|
| 144 |
+
}
|
| 145 |
+
}
|
meta_only/podA/e15/F1-pool-64ksched/train_log.jsonl
ADDED
|
@@ -0,0 +1,141 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"time": "15:39:17", "event": "start", "model_dir": "/workspace/jev_kt/models/hf/JEV-9B", "init": "released", "device": "cuda", "load_s": 103.05782461166382, "trainable_lora": 40108032, "head": 98328}
|
| 2 |
+
{"time": "15:39:17", "event": "run", "data": "data/e15/F1-pool", "out": "models/e15/F1-pool"}
|
| 3 |
+
{"time": "15:43:08", "event": "step", "step": 10, "rows_seen": 320, "loss": 0.5303231496363878, "lr": 1.8333333333333333e-05, "tok_per_s": 917.6799223886935}
|
| 4 |
+
{"time": "15:47:03", "event": "step", "step": 20, "rows_seen": 640, "loss": 0.5562503356486559, "lr": 3.5e-05, "tok_per_s": 915.5114575400971}
|
| 5 |
+
{"time": "15:50:52", "event": "step", "step": 30, "rows_seen": 960, "loss": 0.6963973306119442, "lr": 5.166666666666667e-05, "tok_per_s": 915.8647038863378}
|
| 6 |
+
{"time": "15:58:22", "event": "ckpt", "tag": "ckpt-1000", "rows_seen": 1024, "step": 32, "tokens": 678504, "tok_per_s": 593.3689456089205, "val": {"n": 2000, "dropped_long": 0, "noul_ce": 0.5415979027748108, "kt_auc": 0.7622999963156115, "kt_acc": 0.735, "kt_f1": 0.8071324599708879, "kt_ece": 0.03905150377750397, "kt_nll": 0.5415978856135626}, "dropped_long": 0, "peak_gib": 18.233368396759033}
|
| 7 |
+
{"time": "16:01:29", "event": "step", "step": 40, "rows_seen": 1280, "loss": 0.7384315878152847, "lr": 6.833333333333333e-05, "tok_per_s": 639.483023093292}
|
| 8 |
+
{"time": "16:05:18", "event": "step", "step": 50, "rows_seen": 1600, "loss": 0.7585592311806977, "lr": 8.5e-05, "tok_per_s": 681.6037051188571}
|
| 9 |
+
{"time": "16:09:07", "event": "step", "step": 60, "rows_seen": 1920, "loss": 0.6686401646584272, "lr": 0.0001, "tok_per_s": 712.5422011780105}
|
| 10 |
+
{"time": "16:12:53", "event": "step", "step": 70, "rows_seen": 2240, "loss": 0.6079013631679118, "lr": 9.999377197411753e-05, "tok_per_s": 735.6292145623406}
|
| 11 |
+
{"time": "16:16:42", "event": "step", "step": 80, "rows_seen": 2560, "loss": 0.47051791520789266, "lr": 9.997508952966197e-05, "tok_per_s": 755.0812419448204}
|
| 12 |
+
{"time": "16:20:32", "event": "step", "step": 90, "rows_seen": 2880, "loss": 0.6525400904938579, "lr": 9.994395756578056e-05, "tok_per_s": 770.2672138133466}
|
| 13 |
+
{"time": "16:24:19", "event": "step", "step": 100, "rows_seen": 3200, "loss": 0.5345096457749605, "lr": 9.990038424629131e-05, "tok_per_s": 783.6427673037772}
|
| 14 |
+
{"time": "16:28:10", "event": "step", "step": 110, "rows_seen": 3520, "loss": 0.6915148720145226, "lr": 9.984438099754206e-05, "tok_per_s": 795.094549870955}
|
| 15 |
+
{"time": "16:31:49", "event": "step", "step": 120, "rows_seen": 3840, "loss": 0.6409377884119749, "lr": 9.977596250541417e-05, "tok_per_s": 804.0796835017732}
|
| 16 |
+
{"time": "16:40:08", "event": "ckpt", "tag": "ckpt-4000", "rows_seen": 4000, "step": 125, "tokens": 2631181, "tok_per_s": 720.9349008307702, "val": {"n": 2000, "dropped_long": 0, "noul_ce": 0.5827071070671082, "kt_auc": 0.7610473041998681, "kt_acc": 0.7095, "kt_f1": 0.7715296893432954, "kt_ece": 0.12595178052783013, "kt_nll": 0.5827070810943842}, "dropped_long": 0, "peak_gib": 18.233368396759033}
|
| 17 |
+
{"time": "16:42:03", "event": "step", "step": 130, "rows_seen": 4160, "loss": 0.5773828746750951, "lr": 9.969514671147145e-05, "tok_per_s": 727.320631815138}
|
| 18 |
+
{"time": "16:45:51", "event": "step", "step": 140, "rows_seen": 4480, "loss": 0.622537694289349, "lr": 9.96019548082552e-05, "tok_per_s": 738.5132632482621}
|
| 19 |
+
{"time": "16:49:40", "event": "step", "step": 150, "rows_seen": 4800, "loss": 0.46051981835626066, "lr": 9.949641123372692e-05, "tok_per_s": 748.6852342803468}
|
| 20 |
+
{"time": "16:53:30", "event": "step", "step": 160, "rows_seen": 5120, "loss": 0.46602313593029976, "lr": 9.937854366485982e-05, "tok_per_s": 758.1316001590476}
|
| 21 |
+
{"time": "16:57:20", "event": "step", "step": 170, "rows_seen": 5440, "loss": 0.4729933856287971, "lr": 9.924838301038104e-05, "tok_per_s": 766.2714314999588}
|
| 22 |
+
{"time": "17:01:00", "event": "step", "step": 180, "rows_seen": 5760, "loss": 0.5059829540550709, "lr": 9.910596340266635e-05, "tok_per_s": 773.4813770021683}
|
| 23 |
+
{"time": "17:04:44", "event": "step", "step": 190, "rows_seen": 6080, "loss": 0.5682311803102493, "lr": 9.895132218878961e-05, "tok_per_s": 780.5084875820303}
|
| 24 |
+
{"time": "17:08:40", "event": "step", "step": 200, "rows_seen": 6400, "loss": 0.6042291619814932, "lr": 9.8784499920729e-05, "tok_per_s": 786.879533854256}
|
| 25 |
+
{"time": "17:12:29", "event": "step", "step": 210, "rows_seen": 6720, "loss": 0.5507494350895286, "lr": 9.860554034473306e-05, "tok_per_s": 792.5896666116043}
|
| 26 |
+
{"time": "17:16:10", "event": "step", "step": 220, "rows_seen": 7040, "loss": 0.5679560294374824, "lr": 9.841449038984903e-05, "tok_per_s": 797.8685349497988}
|
| 27 |
+
{"time": "17:19:58", "event": "step", "step": 230, "rows_seen": 7360, "loss": 0.6153504848480225, "lr": 9.821140015561645e-05, "tok_per_s": 803.0822876021568}
|
| 28 |
+
{"time": "17:23:38", "event": "step", "step": 240, "rows_seen": 7680, "loss": 0.4862764838617295, "lr": 9.79963228989294e-05, "tok_per_s": 807.4904133296045}
|
| 29 |
+
{"time": "17:27:25", "event": "step", "step": 250, "rows_seen": 8000, "loss": 0.4005183130502701, "lr": 9.776931502007089e-05, "tok_per_s": 811.6711471408112}
|
| 30 |
+
{"time": "17:31:05", "event": "step", "step": 260, "rows_seen": 8320, "loss": 0.4919987157918513, "lr": 9.753043604792285e-05, "tok_per_s": 815.4143511187291}
|
| 31 |
+
{"time": "17:34:52", "event": "step", "step": 270, "rows_seen": 8640, "loss": 0.5269538387656212, "lr": 9.727974862435562e-05, "tok_per_s": 818.8956523756545}
|
| 32 |
+
{"time": "17:38:39", "event": "step", "step": 280, "rows_seen": 8960, "loss": 0.5008812732994556, "lr": 9.701731848780134e-05, "tok_per_s": 822.543867826727}
|
| 33 |
+
{"time": "17:42:29", "event": "step", "step": 290, "rows_seen": 9280, "loss": 0.38873791228979826, "lr": 9.674321445601502e-05, "tok_per_s": 826.0082914816214}
|
| 34 |
+
{"time": "17:46:15", "event": "step", "step": 300, "rows_seen": 9600, "loss": 0.5436608016025275, "lr": 9.645750840802845e-05, "tok_per_s": 828.9043056383329}
|
| 35 |
+
{"time": "17:50:09", "event": "step", "step": 310, "rows_seen": 9920, "loss": 0.7529093250632286, "lr": 9.616027526530109e-05, "tok_per_s": 831.6963838718863}
|
| 36 |
+
{"time": "17:53:55", "event": "step", "step": 320, "rows_seen": 10240, "loss": 0.4479721710085869, "lr": 9.585159297207315e-05, "tok_per_s": 834.3695713094438}
|
| 37 |
+
{"time": "17:57:41", "event": "step", "step": 330, "rows_seen": 10560, "loss": 0.627282902598381, "lr": 9.553154247492623e-05, "tok_per_s": 836.7848249815431}
|
| 38 |
+
{"time": "18:01:33", "event": "step", "step": 340, "rows_seen": 10880, "loss": 0.6014797193929553, "lr": 9.520020770155628e-05, "tok_per_s": 839.3322248042618}
|
| 39 |
+
{"time": "18:05:21", "event": "step", "step": 350, "rows_seen": 11200, "loss": 0.7178940199082717, "lr": 9.485767553876519e-05, "tok_per_s": 841.7427793345576}
|
| 40 |
+
{"time": "18:09:04", "event": "step", "step": 360, "rows_seen": 11520, "loss": 0.484384024515748, "lr": 9.450403580967612e-05, "tok_per_s": 843.8529470258948}
|
| 41 |
+
{"time": "18:12:48", "event": "step", "step": 370, "rows_seen": 11840, "loss": 0.5317176077514887, "lr": 9.413938125017902e-05, "tok_per_s": 846.059576941209}
|
| 42 |
+
{"time": "18:16:35", "event": "step", "step": 380, "rows_seen": 12160, "loss": 0.5715098977088928, "lr": 9.376380748461219e-05, "tok_per_s": 848.1217227650848}
|
| 43 |
+
{"time": "18:20:21", "event": "step", "step": 390, "rows_seen": 12480, "loss": 0.5058213385054842, "lr": 9.337741300068645e-05, "tok_per_s": 850.0723179655464}
|
| 44 |
+
{"time": "18:24:09", "event": "step", "step": 400, "rows_seen": 12800, "loss": 0.4296084127854556, "lr": 9.298029912365855e-05, "tok_per_s": 851.8059748679249}
|
| 45 |
+
{"time": "18:28:01", "event": "step", "step": 410, "rows_seen": 13120, "loss": 0.33691707625985146, "lr": 9.257256998976022e-05, "tok_per_s": 853.4647144936464}
|
| 46 |
+
{"time": "18:31:48", "event": "step", "step": 420, "rows_seen": 13440, "loss": 0.6571972593665123, "lr": 9.215433251889039e-05, "tok_per_s": 855.2346359233384}
|
| 47 |
+
{"time": "18:35:40", "event": "step", "step": 430, "rows_seen": 13760, "loss": 0.48530035838484764, "lr": 9.172569638657718e-05, "tok_per_s": 856.6993497861457}
|
| 48 |
+
{"time": "18:39:22", "event": "step", "step": 440, "rows_seen": 14080, "loss": 0.45016128569841385, "lr": 9.128677399521752e-05, "tok_per_s": 858.2217125178697}
|
| 49 |
+
{"time": "18:43:12", "event": "step", "step": 450, "rows_seen": 14400, "loss": 0.4965642252936959, "lr": 9.083768044460145e-05, "tok_per_s": 859.641563608333}
|
| 50 |
+
{"time": "18:46:58", "event": "step", "step": 460, "rows_seen": 14720, "loss": 0.6798541019670665, "lr": 9.037853350172924e-05, "tok_per_s": 861.0449657342459}
|
| 51 |
+
{"time": "18:50:50", "event": "step", "step": 470, "rows_seen": 15040, "loss": 0.4528823592700064, "lr": 8.990945356992909e-05, "tok_per_s": 862.4460350957266}
|
| 52 |
+
{"time": "18:54:35", "event": "step", "step": 480, "rows_seen": 15360, "loss": 0.5303344270214438, "lr": 8.943056365728335e-05, "tok_per_s": 863.5801301553191}
|
| 53 |
+
{"time": "18:58:22", "event": "step", "step": 490, "rows_seen": 15680, "loss": 0.5809829644858837, "lr": 8.894198934437185e-05, "tok_per_s": 864.9406820574965}
|
| 54 |
+
{"time": "19:02:11", "event": "step", "step": 500, "rows_seen": 16000, "loss": 0.42125075682997704, "lr": 8.844385875134053e-05, "tok_per_s": 866.1408255033542}
|
| 55 |
+
{"time": "19:08:46", "event": "ckpt", "tag": "ckpt-16000", "rows_seen": 16000, "step": 500, "tokens": 10544036, "tok_per_s": 839.0346341733006, "val": {"n": 2000, "dropped_long": 0, "noul_ce": 0.5415951013565063, "kt_auc": 0.7670316444319398, "kt_acc": 0.735, "kt_f1": 0.8067104303428154, "kt_ece": 0.03874686068296433, "kt_nll": 0.5415951018183259}, "dropped_long": 0, "peak_gib": 18.23592233657837}
|
| 56 |
+
{"time": "19:12:36", "event": "step", "step": 510, "rows_seen": 16320, "loss": 0.568250173702836, "lr": 8.793630250430419e-05, "tok_per_s": 840.4565265533621}
|
| 57 |
+
{"time": "19:16:23", "event": "step", "step": 520, "rows_seen": 16640, "loss": 0.45377571345306933, "lr": 8.741945370109202e-05, "tok_per_s": 842.0155779539998}
|
| 58 |
+
{"time": "19:20:11", "event": "step", "step": 530, "rows_seen": 16960, "loss": 0.4178077857941389, "lr": 8.689344787634503e-05, "tok_per_s": 843.519833138732}
|
| 59 |
+
{"time": "19:24:03", "event": "step", "step": 540, "rows_seen": 17280, "loss": 0.42636957624927163, "lr": 8.63584229659744e-05, "tok_per_s": 844.7387319405225}
|
| 60 |
+
{"time": "19:27:53", "event": "step", "step": 550, "rows_seen": 17600, "loss": 0.7405958976596594, "lr": 8.581451927099022e-05, "tok_per_s": 846.011693755548}
|
| 61 |
+
{"time": "19:31:43", "event": "step", "step": 560, "rows_seen": 17920, "loss": 0.5434928685426712, "lr": 8.526187942070997e-05, "tok_per_s": 847.2400496224152}
|
| 62 |
+
{"time": "19:35:29", "event": "step", "step": 570, "rows_seen": 18240, "loss": 0.45691289473325014, "lr": 8.470064833535642e-05, "tok_per_s": 848.4384502963113}
|
| 63 |
+
{"time": "19:39:16", "event": "step", "step": 580, "rows_seen": 18560, "loss": 0.3559294082224369, "lr": 8.413097318805486e-05, "tok_per_s": 849.683480082124}
|
| 64 |
+
{"time": "19:43:08", "event": "step", "step": 590, "rows_seen": 18880, "loss": 0.6441237931139767, "lr": 8.35530033662395e-05, "tok_per_s": 850.8386288421266}
|
| 65 |
+
{"time": "19:46:51", "event": "step", "step": 600, "rows_seen": 19200, "loss": 0.5293984191957861, "lr": 8.296689043247915e-05, "tok_per_s": 852.1476019148561}
|
| 66 |
+
{"time": "19:50:38", "event": "step", "step": 610, "rows_seen": 19520, "loss": 0.4552152678370476, "lr": 8.237278808473261e-05, "tok_per_s": 853.1887933835146}
|
| 67 |
+
{"time": "19:54:27", "event": "step", "step": 620, "rows_seen": 19840, "loss": 0.4046314130537212, "lr": 8.177085211604405e-05, "tok_per_s": 854.2486299598863}
|
| 68 |
+
{"time": "19:58:10", "event": "step", "step": 630, "rows_seen": 20160, "loss": 0.550866030389443, "lr": 8.116124037368893e-05, "tok_per_s": 855.2622230459343}
|
| 69 |
+
{"time": "20:02:05", "event": "step", "step": 640, "rows_seen": 20480, "loss": 0.5090437447652221, "lr": 8.054411271778132e-05, "tok_per_s": 856.4077629726352}
|
| 70 |
+
{"time": "20:05:54", "event": "step", "step": 650, "rows_seen": 20800, "loss": 0.8178151398897171, "lr": 7.991963097935343e-05, "tok_per_s": 857.410514549224}
|
| 71 |
+
{"time": "20:09:40", "event": "step", "step": 660, "rows_seen": 21120, "loss": 0.44603440910577774, "lr": 7.928795891791805e-05, "tok_per_s": 858.4094964145866}
|
| 72 |
+
{"time": "20:13:21", "event": "step", "step": 670, "rows_seen": 21440, "loss": 0.5785901919007301, "lr": 7.864926217852567e-05, "tok_per_s": 859.1967502950852}
|
| 73 |
+
{"time": "20:17:08", "event": "step", "step": 680, "rows_seen": 21760, "loss": 0.5673699798062444, "lr": 7.800370824832674e-05, "tok_per_s": 860.1660329641811}
|
| 74 |
+
{"time": "20:20:55", "event": "step", "step": 690, "rows_seen": 22080, "loss": 0.5100858034566045, "lr": 7.735146641265118e-05, "tok_per_s": 860.8792926738041}
|
| 75 |
+
{"time": "20:24:42", "event": "step", "step": 700, "rows_seen": 22400, "loss": 0.6086401663487777, "lr": 7.669270771061626e-05, "tok_per_s": 861.7236690519353}
|
| 76 |
+
{"time": "20:28:30", "event": "step", "step": 710, "rows_seen": 22720, "loss": 0.559374613687396, "lr": 7.602760489027442e-05, "tok_per_s": 862.6166361698772}
|
| 77 |
+
{"time": "20:32:23", "event": "step", "step": 720, "rows_seen": 23040, "loss": 0.4983972003683448, "lr": 7.535633236331324e-05, "tok_per_s": 863.2998624635186}
|
| 78 |
+
{"time": "20:36:11", "event": "step", "step": 730, "rows_seen": 23360, "loss": 0.5937356166541576, "lr": 7.467906615931893e-05, "tok_per_s": 864.1571245993105}
|
| 79 |
+
{"time": "20:39:58", "event": "step", "step": 740, "rows_seen": 23680, "loss": 0.48621365893632174, "lr": 7.399598387961561e-05, "tok_per_s": 864.9892139001572}
|
| 80 |
+
{"time": "20:43:48", "event": "step", "step": 750, "rows_seen": 24000, "loss": 0.4657903742045164, "lr": 7.330726465069241e-05, "tok_per_s": 865.7179293945203}
|
| 81 |
+
{"time": "20:47:24", "event": "step", "step": 760, "rows_seen": 24320, "loss": 0.4150607865303755, "lr": 7.26130890772308e-05, "tok_per_s": 866.2976118486816}
|
| 82 |
+
{"time": "20:51:17", "event": "step", "step": 770, "rows_seen": 24640, "loss": 0.34986698254942894, "lr": 7.191363919474393e-05, "tok_per_s": 866.9010986671141}
|
| 83 |
+
{"time": "20:55:11", "event": "step", "step": 780, "rows_seen": 24960, "loss": 0.3827338386327028, "lr": 7.120909842184114e-05, "tok_per_s": 867.7560115192122}
|
| 84 |
+
{"time": "20:59:03", "event": "step", "step": 790, "rows_seen": 25280, "loss": 0.6111974827945232, "lr": 7.049965151212958e-05, "tok_per_s": 868.4870129184216}
|
| 85 |
+
{"time": "21:02:54", "event": "step", "step": 800, "rows_seen": 25600, "loss": 0.6367747157346457, "lr": 6.978548450576568e-05, "tok_per_s": 869.1296069544725}
|
| 86 |
+
{"time": "21:06:47", "event": "step", "step": 810, "rows_seen": 25920, "loss": 0.49950476083904505, "lr": 6.90667846806694e-05, "tok_per_s": 869.8932638249596}
|
| 87 |
+
{"time": "21:10:33", "event": "step", "step": 820, "rows_seen": 26240, "loss": 0.5520738995401189, "lr": 6.834374050341399e-05, "tok_per_s": 870.4447586209693}
|
| 88 |
+
{"time": "21:14:29", "event": "step", "step": 830, "rows_seen": 26560, "loss": 0.6111945323646069, "lr": 6.76165415798036e-05, "tok_per_s": 871.0178030126058}
|
| 89 |
+
{"time": "21:18:17", "event": "step", "step": 840, "rows_seen": 26880, "loss": 0.46462084352970123, "lr": 6.688537860515266e-05, "tok_per_s": 871.6878055594632}
|
| 90 |
+
{"time": "21:22:02", "event": "step", "step": 850, "rows_seen": 27200, "loss": 0.6121984636411071, "lr": 6.615044331427935e-05, "tok_per_s": 872.3494916197046}
|
| 91 |
+
{"time": "21:25:51", "event": "step", "step": 860, "rows_seen": 27520, "loss": 0.44214288191869855, "lr": 6.541192843122628e-05, "tok_per_s": 872.9772054845263}
|
| 92 |
+
{"time": "21:29:43", "event": "step", "step": 870, "rows_seen": 27840, "loss": 0.6299509890377522, "lr": 6.467002761872219e-05, "tok_per_s": 873.6255111775532}
|
| 93 |
+
{"time": "21:33:36", "event": "step", "step": 880, "rows_seen": 28160, "loss": 0.5738023184239864, "lr": 6.392493542739706e-05, "tok_per_s": 874.273161015766}
|
| 94 |
+
{"time": "21:37:18", "event": "step", "step": 890, "rows_seen": 28480, "loss": 0.4683202956803143, "lr": 6.317684724476465e-05, "tok_per_s": 874.7390881864014}
|
| 95 |
+
{"time": "21:41:13", "event": "step", "step": 900, "rows_seen": 28800, "loss": 0.469727270770818, "lr": 6.242595924398558e-05, "tok_per_s": 875.2452938484315}
|
| 96 |
+
{"time": "21:45:04", "event": "step", "step": 910, "rows_seen": 29120, "loss": 0.5859302412718534, "lr": 6.16724683324242e-05, "tok_per_s": 875.7876907104674}
|
| 97 |
+
{"time": "21:48:52", "event": "step", "step": 920, "rows_seen": 29440, "loss": 0.6050049774348736, "lr": 6.091657210001309e-05, "tok_per_s": 876.2969989191104}
|
| 98 |
+
{"time": "21:52:44", "event": "step", "step": 930, "rows_seen": 29760, "loss": 0.5715371587721165, "lr": 6.015846876743859e-05, "tok_per_s": 876.8681743817082}
|
| 99 |
+
{"time": "21:56:40", "event": "step", "step": 940, "rows_seen": 30080, "loss": 0.4671771314460784, "lr": 5.9398357134160796e-05, "tok_per_s": 877.3329122030206}
|
| 100 |
+
{"time": "22:00:30", "event": "step", "step": 950, "rows_seen": 30400, "loss": 0.4529869300313294, "lr": 5.8636436526281816e-05, "tok_per_s": 877.9018591425147}
|
| 101 |
+
{"time": "22:04:26", "event": "step", "step": 960, "rows_seen": 30720, "loss": 0.41454673279076815, "lr": 5.7872906744276015e-05, "tok_per_s": 878.3612537181989}
|
| 102 |
+
{"time": "22:08:12", "event": "step", "step": 970, "rows_seen": 31040, "loss": 0.3616260652197525, "lr": 5.7107968010595795e-05, "tok_per_s": 878.8574154893333}
|
| 103 |
+
{"time": "22:12:00", "event": "step", "step": 980, "rows_seen": 31360, "loss": 0.42127921991050243, "lr": 5.634182091716671e-05, "tok_per_s": 879.3245961286045}
|
| 104 |
+
{"time": "22:15:48", "event": "step", "step": 990, "rows_seen": 31680, "loss": 0.48763304576277733, "lr": 5.5574666372785714e-05, "tok_per_s": 879.6622376699371}
|
| 105 |
+
{"time": "22:19:33", "event": "step", "step": 1000, "rows_seen": 32000, "loss": 0.39415604900568724, "lr": 5.48067055504363e-05, "tok_per_s": 880.144365974715}
|
| 106 |
+
{"time": "22:23:18", "event": "step", "step": 1010, "rows_seen": 32320, "loss": 0.5953622036613524, "lr": 5.403813983453441e-05, "tok_per_s": 880.6091040762847}
|
| 107 |
+
{"time": "22:27:02", "event": "step", "step": 1020, "rows_seen": 32640, "loss": 0.5723254401236773, "lr": 5.3269170768118824e-05, "tok_per_s": 880.9888997861627}
|
| 108 |
+
{"time": "22:30:46", "event": "step", "step": 1030, "rows_seen": 32960, "loss": 0.48154027154669166, "lr": 5.25e-05, "tok_per_s": 881.4756558311823}
|
| 109 |
+
{"time": "22:34:33", "event": "step", "step": 1040, "rows_seen": 33280, "loss": 0.5911916547920555, "lr": 5.173082923188119e-05, "tok_per_s": 881.8793810802554}
|
| 110 |
+
{"time": "22:38:23", "event": "step", "step": 1050, "rows_seen": 33600, "loss": 0.40235981717705727, "lr": 5.0961860165465604e-05, "tok_per_s": 882.1918668407012}
|
| 111 |
+
{"time": "22:42:14", "event": "step", "step": 1060, "rows_seen": 33920, "loss": 0.48941397696034983, "lr": 5.019329444956372e-05, "tok_per_s": 882.6023178228943}
|
| 112 |
+
{"time": "22:45:56", "event": "step", "step": 1070, "rows_seen": 34240, "loss": 0.5359806548804045, "lr": 4.9425333627214304e-05, "tok_per_s": 882.9507920233431}
|
| 113 |
+
{"time": "22:49:48", "event": "step", "step": 1080, "rows_seen": 34560, "loss": 0.5615057991817594, "lr": 4.865817908283331e-05, "tok_per_s": 883.3701355496231}
|
| 114 |
+
{"time": "22:53:35", "event": "step", "step": 1090, "rows_seen": 34880, "loss": 0.3740476332604885, "lr": 4.7892031989404217e-05, "tok_per_s": 883.6806431050793}
|
| 115 |
+
{"time": "22:57:26", "event": "step", "step": 1100, "rows_seen": 35200, "loss": 0.5550510231405497, "lr": 4.7127093255723996e-05, "tok_per_s": 884.0670581582208}
|
| 116 |
+
{"time": "23:01:22", "event": "step", "step": 1110, "rows_seen": 35520, "loss": 0.4037408595904708, "lr": 4.636356347371821e-05, "tok_per_s": 884.4985633971859}
|
| 117 |
+
{"time": "23:05:07", "event": "step", "step": 1120, "rows_seen": 35840, "loss": 0.38779116701334715, "lr": 4.560164286583922e-05, "tok_per_s": 884.8658508521748}
|
| 118 |
+
{"time": "23:08:54", "event": "step", "step": 1130, "rows_seen": 36160, "loss": 0.5320682032033801, "lr": 4.4841531232561406e-05, "tok_per_s": 885.2586423224216}
|
| 119 |
+
{"time": "23:12:48", "event": "step", "step": 1140, "rows_seen": 36480, "loss": 0.5520844999700785, "lr": 4.4083427899986926e-05, "tok_per_s": 885.5301693552192}
|
| 120 |
+
{"time": "23:16:33", "event": "step", "step": 1150, "rows_seen": 36800, "loss": 0.6835330203175545, "lr": 4.3327531667575824e-05, "tok_per_s": 885.8355958066563}
|
| 121 |
+
{"time": "23:20:23", "event": "step", "step": 1160, "rows_seen": 37120, "loss": 0.41280198795720935, "lr": 4.257404075601443e-05, "tok_per_s": 886.1520477304778}
|
| 122 |
+
{"time": "23:24:16", "event": "step", "step": 1170, "rows_seen": 37440, "loss": 0.6245222613215446, "lr": 4.182315275523537e-05, "tok_per_s": 886.3943568226827}
|
| 123 |
+
{"time": "23:28:07", "event": "step", "step": 1180, "rows_seen": 37760, "loss": 0.4812806900590658, "lr": 4.1075064572602964e-05, "tok_per_s": 886.7769082886637}
|
| 124 |
+
{"time": "23:31:56", "event": "step", "step": 1190, "rows_seen": 38080, "loss": 0.5893618543632329, "lr": 4.0329972381277805e-05, "tok_per_s": 887.0910034016722}
|
| 125 |
+
{"time": "23:35:46", "event": "step", "step": 1200, "rows_seen": 38400, "loss": 0.3978904075920582, "lr": 3.958807156877371e-05, "tok_per_s": 887.3882293667284}
|
| 126 |
+
{"time": "23:39:31", "event": "step", "step": 1210, "rows_seen": 38720, "loss": 0.38529140362516046, "lr": 3.884955668572066e-05, "tok_per_s": 887.7042056519199}
|
| 127 |
+
{"time": "23:43:19", "event": "step", "step": 1220, "rows_seen": 39040, "loss": 0.4402267914265394, "lr": 3.811462139484735e-05, "tok_per_s": 887.9619960888213}
|
| 128 |
+
{"time": "23:47:09", "event": "step", "step": 1230, "rows_seen": 39360, "loss": 0.42451312858611345, "lr": 3.738345842019643e-05, "tok_per_s": 888.2876609433782}
|
| 129 |
+
{"time": "23:50:56", "event": "step", "step": 1240, "rows_seen": 39680, "loss": 0.49415042996406555, "lr": 3.665625949658603e-05, "tok_per_s": 888.5749530891543}
|
| 130 |
+
{"time": "23:54:54", "event": "step", "step": 1250, "rows_seen": 40000, "loss": 0.473277322947979, "lr": 3.5933215319330585e-05, "tok_per_s": 888.8087293598423}
|
| 131 |
+
{"time": "23:58:41", "event": "step", "step": 1260, "rows_seen": 40320, "loss": 0.3329443195834756, "lr": 3.521451549423435e-05, "tok_per_s": 889.0324983994182}
|
| 132 |
+
{"time": "00:02:35", "event": "step", "step": 1270, "rows_seen": 40640, "loss": 0.6091567252296954, "lr": 3.450034848787043e-05, "tok_per_s": 889.267645497651}
|
| 133 |
+
{"time": "00:06:22", "event": "step", "step": 1280, "rows_seen": 40960, "loss": 0.531061683781445, "lr": 3.379090157815887e-05, "tok_per_s": 889.5916202515181}
|
| 134 |
+
{"time": "00:10:18", "event": "step", "step": 1290, "rows_seen": 41280, "loss": 0.4837521743029356, "lr": 3.3086360805256096e-05, "tok_per_s": 889.8400434571761}
|
| 135 |
+
{"time": "00:14:08", "event": "step", "step": 1300, "rows_seen": 41600, "loss": 0.5356471491977572, "lr": 3.238691092276923e-05, "tok_per_s": 890.1081757998043}
|
| 136 |
+
{"time": "00:17:52", "event": "step", "step": 1310, "rows_seen": 41920, "loss": 0.5196949392557144, "lr": 3.169273534930759e-05, "tok_per_s": 890.3170668490965}
|
| 137 |
+
{"time": "00:21:38", "event": "step", "step": 1320, "rows_seen": 42240, "loss": 0.7474268823862076, "lr": 3.10040161203844e-05, "tok_per_s": 890.5464844827695}
|
| 138 |
+
{"time": "00:25:29", "event": "step", "step": 1330, "rows_seen": 42560, "loss": 0.4453593622893095, "lr": 3.032093384068107e-05, "tok_per_s": 890.7989624938164}
|
| 139 |
+
{"time": "00:29:22", "event": "step", "step": 1340, "rows_seen": 42880, "loss": 0.3571602231822908, "lr": 2.9643667636686757e-05, "tok_per_s": 890.9791258810741}
|
| 140 |
+
{"time": "00:33:08", "event": "step", "step": 1350, "rows_seen": 43200, "loss": 0.6555291656404734, "lr": 2.8972395109725607e-05, "tok_per_s": 891.2636423360315}
|
| 141 |
+
{"time": "00:36:53", "event": "step", "step": 1360, "rows_seen": 43520, "loss": 0.4141939878463745, "lr": 2.8307292289383776e-05, "tok_per_s": 891.4690158154687}
|
meta_only/podA/e15/F1-pool-ts1/ckpt-1000/adapter/adapter_config.json
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B",
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 32,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.05,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 16,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"up_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"o_proj",
|
| 41 |
+
"q_proj",
|
| 42 |
+
"k_proj",
|
| 43 |
+
"in_proj_qkv",
|
| 44 |
+
"out_proj",
|
| 45 |
+
"in_proj_z",
|
| 46 |
+
"down_proj",
|
| 47 |
+
"v_proj"
|
| 48 |
+
],
|
| 49 |
+
"target_parameters": null,
|
| 50 |
+
"task_type": null,
|
| 51 |
+
"trainable_token_indices": null,
|
| 52 |
+
"use_bdlora": null,
|
| 53 |
+
"use_dora": false,
|
| 54 |
+
"use_qalora": false,
|
| 55 |
+
"use_rslora": false,
|
| 56 |
+
"velora_config": null
|
| 57 |
+
}
|
meta_only/podA/e15/F1-pool-ts1/ckpt-1000/adapter_vllm/adapter_config.json
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": null,
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 64.0,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.0,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 32,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"down_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"in_proj_qkv",
|
| 41 |
+
"in_proj_z",
|
| 42 |
+
"k_proj",
|
| 43 |
+
"lm_head",
|
| 44 |
+
"o_proj",
|
| 45 |
+
"out_proj",
|
| 46 |
+
"q_proj",
|
| 47 |
+
"up_proj",
|
| 48 |
+
"v_proj"
|
| 49 |
+
],
|
| 50 |
+
"target_parameters": null,
|
| 51 |
+
"task_type": null,
|
| 52 |
+
"trainable_token_indices": null,
|
| 53 |
+
"use_bdlora": null,
|
| 54 |
+
"use_dora": false,
|
| 55 |
+
"use_qalora": false,
|
| 56 |
+
"use_rslora": false,
|
| 57 |
+
"velora_config": null
|
| 58 |
+
}
|
meta_only/podA/e15/F1-pool-ts1/ckpt-1000/adapter_vllm/decision_head.json
ADDED
|
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bias": [
|
| 3 |
+
3.946308424929157e-05,
|
| 4 |
+
-3.9710779674351215e-05,
|
| 5 |
+
-0.0025878134183585644,
|
| 6 |
+
-0.0012615364976227283,
|
| 7 |
+
-0.00025007399381138384,
|
| 8 |
+
0.0019558551721274853,
|
| 9 |
+
0.0013559526996687055,
|
| 10 |
+
-0.0023348911199718714,
|
| 11 |
+
-0.0004082925443071872,
|
| 12 |
+
-0.0002078723773593083,
|
| 13 |
+
-1.2957353646925185e-05,
|
| 14 |
+
0.0001676026004133746,
|
| 15 |
+
0.0005569425993598998,
|
| 16 |
+
-0.001412136247381568,
|
| 17 |
+
-0.0028663526754826307,
|
| 18 |
+
-0.005061559844762087,
|
| 19 |
+
-0.00593328895047307,
|
| 20 |
+
-0.004319756291806698,
|
| 21 |
+
-0.004922541789710522,
|
| 22 |
+
-0.0049067651852965355,
|
| 23 |
+
-0.003943523392081261,
|
| 24 |
+
-0.003319827374070883,
|
| 25 |
+
-0.005581587553024292,
|
| 26 |
+
-0.0034480984322726727
|
| 27 |
+
],
|
| 28 |
+
"verbalizer_ids": [
|
| 29 |
+
3721,
|
| 30 |
+
1802,
|
| 31 |
+
15,
|
| 32 |
+
16,
|
| 33 |
+
17,
|
| 34 |
+
18,
|
| 35 |
+
19,
|
| 36 |
+
20,
|
| 37 |
+
32,
|
| 38 |
+
33,
|
| 39 |
+
34,
|
| 40 |
+
35,
|
| 41 |
+
36,
|
| 42 |
+
37,
|
| 43 |
+
38,
|
| 44 |
+
39,
|
| 45 |
+
40,
|
| 46 |
+
41,
|
| 47 |
+
42,
|
| 48 |
+
43,
|
| 49 |
+
44,
|
| 50 |
+
45,
|
| 51 |
+
46,
|
| 52 |
+
47
|
| 53 |
+
],
|
| 54 |
+
"slots": {
|
| 55 |
+
"num_slots": 24,
|
| 56 |
+
"ranges": {
|
| 57 |
+
"noul": [
|
| 58 |
+
0,
|
| 59 |
+
2
|
| 60 |
+
],
|
| 61 |
+
"score": [
|
| 62 |
+
2,
|
| 63 |
+
8
|
| 64 |
+
],
|
| 65 |
+
"choice": [
|
| 66 |
+
8,
|
| 67 |
+
24
|
| 68 |
+
]
|
| 69 |
+
},
|
| 70 |
+
"verbalizers": [
|
| 71 |
+
"false",
|
| 72 |
+
"true",
|
| 73 |
+
"0",
|
| 74 |
+
"1",
|
| 75 |
+
"2",
|
| 76 |
+
"3",
|
| 77 |
+
"4",
|
| 78 |
+
"5",
|
| 79 |
+
"A",
|
| 80 |
+
"B",
|
| 81 |
+
"C",
|
| 82 |
+
"D",
|
| 83 |
+
"E",
|
| 84 |
+
"F",
|
| 85 |
+
"G",
|
| 86 |
+
"H",
|
| 87 |
+
"I",
|
| 88 |
+
"J",
|
| 89 |
+
"K",
|
| 90 |
+
"L",
|
| 91 |
+
"M",
|
| 92 |
+
"N",
|
| 93 |
+
"O",
|
| 94 |
+
"P"
|
| 95 |
+
],
|
| 96 |
+
"template_version": "bare-v1"
|
| 97 |
+
},
|
| 98 |
+
"note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature"
|
| 99 |
+
}
|
meta_only/podA/e15/F1-pool-ts1/ckpt-1000/calibration.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"version": 1,
|
| 3 |
+
"per_kind": {
|
| 4 |
+
"noul": 1.8453449550299923,
|
| 5 |
+
"choice": 1.0,
|
| 6 |
+
"score": 1.0
|
| 7 |
+
},
|
| 8 |
+
"fit": {
|
| 9 |
+
"noul": {
|
| 10 |
+
"n": 2000,
|
| 11 |
+
"T": 1.8453449550299923,
|
| 12 |
+
"ce_before": 0.6150162816047668,
|
| 13 |
+
"ce_after": 0.5616257786750793
|
| 14 |
+
}
|
| 15 |
+
},
|
| 16 |
+
"source": "models/e15/F1-pool-ts1/ckpt-1000"
|
| 17 |
+
}
|
meta_only/podA/e15/F1-pool-ts1/ckpt-1000/judge_config.json
ADDED
|
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"base_model_path": "/root/models/Qwen3.5-9B",
|
| 3 |
+
"hidden_size": 4096,
|
| 4 |
+
"slots": {
|
| 5 |
+
"num_slots": 24,
|
| 6 |
+
"ranges": {
|
| 7 |
+
"noul": [
|
| 8 |
+
0,
|
| 9 |
+
2
|
| 10 |
+
],
|
| 11 |
+
"score": [
|
| 12 |
+
2,
|
| 13 |
+
8
|
| 14 |
+
],
|
| 15 |
+
"choice": [
|
| 16 |
+
8,
|
| 17 |
+
24
|
| 18 |
+
]
|
| 19 |
+
},
|
| 20 |
+
"verbalizers": [
|
| 21 |
+
"false",
|
| 22 |
+
"true",
|
| 23 |
+
"0",
|
| 24 |
+
"1",
|
| 25 |
+
"2",
|
| 26 |
+
"3",
|
| 27 |
+
"4",
|
| 28 |
+
"5",
|
| 29 |
+
"A",
|
| 30 |
+
"B",
|
| 31 |
+
"C",
|
| 32 |
+
"D",
|
| 33 |
+
"E",
|
| 34 |
+
"F",
|
| 35 |
+
"G",
|
| 36 |
+
"H",
|
| 37 |
+
"I",
|
| 38 |
+
"J",
|
| 39 |
+
"K",
|
| 40 |
+
"L",
|
| 41 |
+
"M",
|
| 42 |
+
"N",
|
| 43 |
+
"O",
|
| 44 |
+
"P"
|
| 45 |
+
],
|
| 46 |
+
"template_version": "bare-v1"
|
| 47 |
+
},
|
| 48 |
+
"verbalizer_ids": [
|
| 49 |
+
3721,
|
| 50 |
+
1802,
|
| 51 |
+
15,
|
| 52 |
+
16,
|
| 53 |
+
17,
|
| 54 |
+
18,
|
| 55 |
+
19,
|
| 56 |
+
20,
|
| 57 |
+
32,
|
| 58 |
+
33,
|
| 59 |
+
34,
|
| 60 |
+
35,
|
| 61 |
+
36,
|
| 62 |
+
37,
|
| 63 |
+
38,
|
| 64 |
+
39,
|
| 65 |
+
40,
|
| 66 |
+
41,
|
| 67 |
+
42,
|
| 68 |
+
43,
|
| 69 |
+
44,
|
| 70 |
+
45,
|
| 71 |
+
46,
|
| 72 |
+
47
|
| 73 |
+
],
|
| 74 |
+
"kinds": [
|
| 75 |
+
"noul",
|
| 76 |
+
"choice",
|
| 77 |
+
"score"
|
| 78 |
+
],
|
| 79 |
+
"model_name": "jev-judge-qwen35-9b",
|
| 80 |
+
"model_version": "0.8.0",
|
| 81 |
+
"lm_head_included": true,
|
| 82 |
+
"weights_mode": "unmerged",
|
| 83 |
+
"adapter_subfolder": "adapter",
|
| 84 |
+
"lora": {
|
| 85 |
+
"r": 16,
|
| 86 |
+
"lora_alpha": 32,
|
| 87 |
+
"lora_dropout": 0.05,
|
| 88 |
+
"target_modules": [
|
| 89 |
+
"down_proj",
|
| 90 |
+
"gate_proj",
|
| 91 |
+
"in_proj_qkv",
|
| 92 |
+
"in_proj_z",
|
| 93 |
+
"k_proj",
|
| 94 |
+
"o_proj",
|
| 95 |
+
"out_proj",
|
| 96 |
+
"q_proj",
|
| 97 |
+
"up_proj",
|
| 98 |
+
"v_proj"
|
| 99 |
+
]
|
| 100 |
+
},
|
| 101 |
+
"trained_stage": "e15",
|
| 102 |
+
"source_checkpoint": "models/e15/F1-pool-ts1/ckpt-1000",
|
| 103 |
+
"softcap": null,
|
| 104 |
+
"e15": {
|
| 105 |
+
"data": "data/e15/F1-pool",
|
| 106 |
+
"data_manifest": {
|
| 107 |
+
"arm": "F1-pool",
|
| 108 |
+
"name": "F1-pool",
|
| 109 |
+
"train_sha256": "dad16f94a5659768b254a6c12538c01f9c64cd43d574a3435bcc70c9c1bac130",
|
| 110 |
+
"n_train": 64000
|
| 111 |
+
},
|
| 112 |
+
"init": "released",
|
| 113 |
+
"lr": 0.0001,
|
| 114 |
+
"lr_head": 0.0002,
|
| 115 |
+
"rows_per_step": 32,
|
| 116 |
+
"skip_rows": 0,
|
| 117 |
+
"epochs": 1,
|
| 118 |
+
"max_len": 3072,
|
| 119 |
+
"schedule": "warmup 0.03 + cosine to x0.05 over 500 steps",
|
| 120 |
+
"total_rows": 16000,
|
| 121 |
+
"lora": {
|
| 122 |
+
"r": 16,
|
| 123 |
+
"lora_alpha": 32,
|
| 124 |
+
"lora_dropout": 0.05
|
| 125 |
+
},
|
| 126 |
+
"train_seed": 1,
|
| 127 |
+
"event": "ckpt",
|
| 128 |
+
"tag": "ckpt-1000",
|
| 129 |
+
"rows_seen": 1024,
|
| 130 |
+
"step": 32,
|
| 131 |
+
"tokens": 684623,
|
| 132 |
+
"tok_per_s": 604.1670308900905,
|
| 133 |
+
"val": {
|
| 134 |
+
"n": 2000,
|
| 135 |
+
"dropped_long": 0,
|
| 136 |
+
"noul_ce": 0.6150162816047668,
|
| 137 |
+
"kt_auc": 0.7537879422243212,
|
| 138 |
+
"kt_acc": 0.694,
|
| 139 |
+
"kt_f1": 0.8083907326236693,
|
| 140 |
+
"kt_ece": 0.14204001927375792,
|
| 141 |
+
"kt_nll": 0.6150163156474592
|
| 142 |
+
},
|
| 143 |
+
"dropped_long": 0,
|
| 144 |
+
"peak_gib": 18.229214191436768
|
| 145 |
+
}
|
| 146 |
+
}
|
meta_only/podA/e15/F1-pool-ts1/ckpt-16000/adapter/adapter_config.json
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B",
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 32,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.05,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 16,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"up_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"o_proj",
|
| 41 |
+
"q_proj",
|
| 42 |
+
"k_proj",
|
| 43 |
+
"in_proj_qkv",
|
| 44 |
+
"out_proj",
|
| 45 |
+
"in_proj_z",
|
| 46 |
+
"down_proj",
|
| 47 |
+
"v_proj"
|
| 48 |
+
],
|
| 49 |
+
"target_parameters": null,
|
| 50 |
+
"task_type": null,
|
| 51 |
+
"trainable_token_indices": null,
|
| 52 |
+
"use_bdlora": null,
|
| 53 |
+
"use_dora": false,
|
| 54 |
+
"use_qalora": false,
|
| 55 |
+
"use_rslora": false,
|
| 56 |
+
"velora_config": null
|
| 57 |
+
}
|
meta_only/podA/e15/F1-pool-ts1/ckpt-16000/adapter_vllm/adapter_config.json
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": null,
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 64.0,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.0,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 32,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"down_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"in_proj_qkv",
|
| 41 |
+
"in_proj_z",
|
| 42 |
+
"k_proj",
|
| 43 |
+
"lm_head",
|
| 44 |
+
"o_proj",
|
| 45 |
+
"out_proj",
|
| 46 |
+
"q_proj",
|
| 47 |
+
"up_proj",
|
| 48 |
+
"v_proj"
|
| 49 |
+
],
|
| 50 |
+
"target_parameters": null,
|
| 51 |
+
"task_type": null,
|
| 52 |
+
"trainable_token_indices": null,
|
| 53 |
+
"use_bdlora": null,
|
| 54 |
+
"use_dora": false,
|
| 55 |
+
"use_qalora": false,
|
| 56 |
+
"use_rslora": false,
|
| 57 |
+
"velora_config": null
|
| 58 |
+
}
|
meta_only/podA/e15/F1-pool-ts1/ckpt-16000/adapter_vllm/decision_head.json
ADDED
|
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bias": [
|
| 3 |
+
-5.97268262936268e-05,
|
| 4 |
+
5.9471665736055e-05,
|
| 5 |
+
-0.0025878134183585644,
|
| 6 |
+
-0.0012615364976227283,
|
| 7 |
+
-0.00025007399381138384,
|
| 8 |
+
0.0019558551721274853,
|
| 9 |
+
0.0013559526996687055,
|
| 10 |
+
-0.0023348911199718714,
|
| 11 |
+
-0.0004082925443071872,
|
| 12 |
+
-0.0002078723773593083,
|
| 13 |
+
-1.2957353646925185e-05,
|
| 14 |
+
0.0001676026004133746,
|
| 15 |
+
0.0005569425993598998,
|
| 16 |
+
-0.001412136247381568,
|
| 17 |
+
-0.0028663526754826307,
|
| 18 |
+
-0.005061559844762087,
|
| 19 |
+
-0.00593328895047307,
|
| 20 |
+
-0.004319756291806698,
|
| 21 |
+
-0.004922541789710522,
|
| 22 |
+
-0.0049067651852965355,
|
| 23 |
+
-0.003943523392081261,
|
| 24 |
+
-0.003319827374070883,
|
| 25 |
+
-0.005581587553024292,
|
| 26 |
+
-0.0034480984322726727
|
| 27 |
+
],
|
| 28 |
+
"verbalizer_ids": [
|
| 29 |
+
3721,
|
| 30 |
+
1802,
|
| 31 |
+
15,
|
| 32 |
+
16,
|
| 33 |
+
17,
|
| 34 |
+
18,
|
| 35 |
+
19,
|
| 36 |
+
20,
|
| 37 |
+
32,
|
| 38 |
+
33,
|
| 39 |
+
34,
|
| 40 |
+
35,
|
| 41 |
+
36,
|
| 42 |
+
37,
|
| 43 |
+
38,
|
| 44 |
+
39,
|
| 45 |
+
40,
|
| 46 |
+
41,
|
| 47 |
+
42,
|
| 48 |
+
43,
|
| 49 |
+
44,
|
| 50 |
+
45,
|
| 51 |
+
46,
|
| 52 |
+
47
|
| 53 |
+
],
|
| 54 |
+
"slots": {
|
| 55 |
+
"num_slots": 24,
|
| 56 |
+
"ranges": {
|
| 57 |
+
"noul": [
|
| 58 |
+
0,
|
| 59 |
+
2
|
| 60 |
+
],
|
| 61 |
+
"score": [
|
| 62 |
+
2,
|
| 63 |
+
8
|
| 64 |
+
],
|
| 65 |
+
"choice": [
|
| 66 |
+
8,
|
| 67 |
+
24
|
| 68 |
+
]
|
| 69 |
+
},
|
| 70 |
+
"verbalizers": [
|
| 71 |
+
"false",
|
| 72 |
+
"true",
|
| 73 |
+
"0",
|
| 74 |
+
"1",
|
| 75 |
+
"2",
|
| 76 |
+
"3",
|
| 77 |
+
"4",
|
| 78 |
+
"5",
|
| 79 |
+
"A",
|
| 80 |
+
"B",
|
| 81 |
+
"C",
|
| 82 |
+
"D",
|
| 83 |
+
"E",
|
| 84 |
+
"F",
|
| 85 |
+
"G",
|
| 86 |
+
"H",
|
| 87 |
+
"I",
|
| 88 |
+
"J",
|
| 89 |
+
"K",
|
| 90 |
+
"L",
|
| 91 |
+
"M",
|
| 92 |
+
"N",
|
| 93 |
+
"O",
|
| 94 |
+
"P"
|
| 95 |
+
],
|
| 96 |
+
"template_version": "bare-v1"
|
| 97 |
+
},
|
| 98 |
+
"note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature"
|
| 99 |
+
}
|
meta_only/podA/e15/F1-pool-ts1/ckpt-16000/calibration.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"version": 1,
|
| 3 |
+
"per_kind": {
|
| 4 |
+
"noul": 1.126429616320892,
|
| 5 |
+
"choice": 1.0,
|
| 6 |
+
"score": 1.0
|
| 7 |
+
},
|
| 8 |
+
"fit": {
|
| 9 |
+
"noul": {
|
| 10 |
+
"n": 2000,
|
| 11 |
+
"T": 1.126429616320892,
|
| 12 |
+
"ce_before": 0.5292231440544128,
|
| 13 |
+
"ce_after": 0.5275862812995911
|
| 14 |
+
}
|
| 15 |
+
},
|
| 16 |
+
"source": "models/e15/F1-pool-ts1/ckpt-16000"
|
| 17 |
+
}
|
meta_only/podA/e15/F1-pool-ts1/ckpt-16000/judge_config.json
ADDED
|
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"base_model_path": "/root/models/Qwen3.5-9B",
|
| 3 |
+
"hidden_size": 4096,
|
| 4 |
+
"slots": {
|
| 5 |
+
"num_slots": 24,
|
| 6 |
+
"ranges": {
|
| 7 |
+
"noul": [
|
| 8 |
+
0,
|
| 9 |
+
2
|
| 10 |
+
],
|
| 11 |
+
"score": [
|
| 12 |
+
2,
|
| 13 |
+
8
|
| 14 |
+
],
|
| 15 |
+
"choice": [
|
| 16 |
+
8,
|
| 17 |
+
24
|
| 18 |
+
]
|
| 19 |
+
},
|
| 20 |
+
"verbalizers": [
|
| 21 |
+
"false",
|
| 22 |
+
"true",
|
| 23 |
+
"0",
|
| 24 |
+
"1",
|
| 25 |
+
"2",
|
| 26 |
+
"3",
|
| 27 |
+
"4",
|
| 28 |
+
"5",
|
| 29 |
+
"A",
|
| 30 |
+
"B",
|
| 31 |
+
"C",
|
| 32 |
+
"D",
|
| 33 |
+
"E",
|
| 34 |
+
"F",
|
| 35 |
+
"G",
|
| 36 |
+
"H",
|
| 37 |
+
"I",
|
| 38 |
+
"J",
|
| 39 |
+
"K",
|
| 40 |
+
"L",
|
| 41 |
+
"M",
|
| 42 |
+
"N",
|
| 43 |
+
"O",
|
| 44 |
+
"P"
|
| 45 |
+
],
|
| 46 |
+
"template_version": "bare-v1"
|
| 47 |
+
},
|
| 48 |
+
"verbalizer_ids": [
|
| 49 |
+
3721,
|
| 50 |
+
1802,
|
| 51 |
+
15,
|
| 52 |
+
16,
|
| 53 |
+
17,
|
| 54 |
+
18,
|
| 55 |
+
19,
|
| 56 |
+
20,
|
| 57 |
+
32,
|
| 58 |
+
33,
|
| 59 |
+
34,
|
| 60 |
+
35,
|
| 61 |
+
36,
|
| 62 |
+
37,
|
| 63 |
+
38,
|
| 64 |
+
39,
|
| 65 |
+
40,
|
| 66 |
+
41,
|
| 67 |
+
42,
|
| 68 |
+
43,
|
| 69 |
+
44,
|
| 70 |
+
45,
|
| 71 |
+
46,
|
| 72 |
+
47
|
| 73 |
+
],
|
| 74 |
+
"kinds": [
|
| 75 |
+
"noul",
|
| 76 |
+
"choice",
|
| 77 |
+
"score"
|
| 78 |
+
],
|
| 79 |
+
"model_name": "jev-judge-qwen35-9b",
|
| 80 |
+
"model_version": "0.8.0",
|
| 81 |
+
"lm_head_included": true,
|
| 82 |
+
"weights_mode": "unmerged",
|
| 83 |
+
"adapter_subfolder": "adapter",
|
| 84 |
+
"lora": {
|
| 85 |
+
"r": 16,
|
| 86 |
+
"lora_alpha": 32,
|
| 87 |
+
"lora_dropout": 0.05,
|
| 88 |
+
"target_modules": [
|
| 89 |
+
"down_proj",
|
| 90 |
+
"gate_proj",
|
| 91 |
+
"in_proj_qkv",
|
| 92 |
+
"in_proj_z",
|
| 93 |
+
"k_proj",
|
| 94 |
+
"o_proj",
|
| 95 |
+
"out_proj",
|
| 96 |
+
"q_proj",
|
| 97 |
+
"up_proj",
|
| 98 |
+
"v_proj"
|
| 99 |
+
]
|
| 100 |
+
},
|
| 101 |
+
"trained_stage": "e15",
|
| 102 |
+
"source_checkpoint": "models/e15/F1-pool-ts1/ckpt-16000",
|
| 103 |
+
"softcap": null,
|
| 104 |
+
"e15": {
|
| 105 |
+
"data": "data/e15/F1-pool",
|
| 106 |
+
"data_manifest": {
|
| 107 |
+
"arm": "F1-pool",
|
| 108 |
+
"name": "F1-pool",
|
| 109 |
+
"train_sha256": "dad16f94a5659768b254a6c12538c01f9c64cd43d574a3435bcc70c9c1bac130",
|
| 110 |
+
"n_train": 64000
|
| 111 |
+
},
|
| 112 |
+
"init": "released",
|
| 113 |
+
"lr": 0.0001,
|
| 114 |
+
"lr_head": 0.0002,
|
| 115 |
+
"rows_per_step": 32,
|
| 116 |
+
"skip_rows": 0,
|
| 117 |
+
"epochs": 1,
|
| 118 |
+
"max_len": 3072,
|
| 119 |
+
"schedule": "warmup 0.03 + cosine to x0.05 over 500 steps",
|
| 120 |
+
"total_rows": 16000,
|
| 121 |
+
"lora": {
|
| 122 |
+
"r": 16,
|
| 123 |
+
"lora_alpha": 32,
|
| 124 |
+
"lora_dropout": 0.05
|
| 125 |
+
},
|
| 126 |
+
"train_seed": 1,
|
| 127 |
+
"event": "ckpt",
|
| 128 |
+
"tag": "ckpt-16000",
|
| 129 |
+
"rows_seen": 16000,
|
| 130 |
+
"step": 500,
|
| 131 |
+
"tokens": 10544036,
|
| 132 |
+
"tok_per_s": 840.4803245529926,
|
| 133 |
+
"val": {
|
| 134 |
+
"n": 2000,
|
| 135 |
+
"dropped_long": 0,
|
| 136 |
+
"noul_ce": 0.5292231440544128,
|
| 137 |
+
"kt_auc": 0.7788161054672977,
|
| 138 |
+
"kt_acc": 0.734,
|
| 139 |
+
"kt_f1": 0.8044117647058824,
|
| 140 |
+
"kt_ece": 0.04880642385780811,
|
| 141 |
+
"kt_nll": 0.5292231264216826
|
| 142 |
+
},
|
| 143 |
+
"dropped_long": 0,
|
| 144 |
+
"peak_gib": 18.23512029647827
|
| 145 |
+
}
|
| 146 |
+
}
|
meta_only/podA/e15/F1-pool-ts1/ckpt-4000/adapter/adapter_config.json
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B",
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 32,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.05,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 16,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"up_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"o_proj",
|
| 41 |
+
"q_proj",
|
| 42 |
+
"k_proj",
|
| 43 |
+
"in_proj_qkv",
|
| 44 |
+
"out_proj",
|
| 45 |
+
"in_proj_z",
|
| 46 |
+
"down_proj",
|
| 47 |
+
"v_proj"
|
| 48 |
+
],
|
| 49 |
+
"target_parameters": null,
|
| 50 |
+
"task_type": null,
|
| 51 |
+
"trainable_token_indices": null,
|
| 52 |
+
"use_bdlora": null,
|
| 53 |
+
"use_dora": false,
|
| 54 |
+
"use_qalora": false,
|
| 55 |
+
"use_rslora": false,
|
| 56 |
+
"velora_config": null
|
| 57 |
+
}
|
meta_only/podA/e15/F1-pool-ts1/ckpt-4000/adapter_vllm/adapter_config.json
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": null,
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 64.0,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.0,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 32,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"down_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"in_proj_qkv",
|
| 41 |
+
"in_proj_z",
|
| 42 |
+
"k_proj",
|
| 43 |
+
"lm_head",
|
| 44 |
+
"o_proj",
|
| 45 |
+
"out_proj",
|
| 46 |
+
"q_proj",
|
| 47 |
+
"up_proj",
|
| 48 |
+
"v_proj"
|
| 49 |
+
],
|
| 50 |
+
"target_parameters": null,
|
| 51 |
+
"task_type": null,
|
| 52 |
+
"trainable_token_indices": null,
|
| 53 |
+
"use_bdlora": null,
|
| 54 |
+
"use_dora": false,
|
| 55 |
+
"use_qalora": false,
|
| 56 |
+
"use_rslora": false,
|
| 57 |
+
"velora_config": null
|
| 58 |
+
}
|
meta_only/podA/e15/F1-pool-ts1/ckpt-4000/adapter_vllm/decision_head.json
ADDED
|
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bias": [
|
| 3 |
+
0.00010020808986155316,
|
| 4 |
+
-0.0001004588048090227,
|
| 5 |
+
-0.0025878134183585644,
|
| 6 |
+
-0.0012615364976227283,
|
| 7 |
+
-0.00025007399381138384,
|
| 8 |
+
0.0019558551721274853,
|
| 9 |
+
0.0013559526996687055,
|
| 10 |
+
-0.0023348911199718714,
|
| 11 |
+
-0.0004082925443071872,
|
| 12 |
+
-0.0002078723773593083,
|
| 13 |
+
-1.2957353646925185e-05,
|
| 14 |
+
0.0001676026004133746,
|
| 15 |
+
0.0005569425993598998,
|
| 16 |
+
-0.001412136247381568,
|
| 17 |
+
-0.0028663526754826307,
|
| 18 |
+
-0.005061559844762087,
|
| 19 |
+
-0.00593328895047307,
|
| 20 |
+
-0.004319756291806698,
|
| 21 |
+
-0.004922541789710522,
|
| 22 |
+
-0.0049067651852965355,
|
| 23 |
+
-0.003943523392081261,
|
| 24 |
+
-0.003319827374070883,
|
| 25 |
+
-0.005581587553024292,
|
| 26 |
+
-0.0034480984322726727
|
| 27 |
+
],
|
| 28 |
+
"verbalizer_ids": [
|
| 29 |
+
3721,
|
| 30 |
+
1802,
|
| 31 |
+
15,
|
| 32 |
+
16,
|
| 33 |
+
17,
|
| 34 |
+
18,
|
| 35 |
+
19,
|
| 36 |
+
20,
|
| 37 |
+
32,
|
| 38 |
+
33,
|
| 39 |
+
34,
|
| 40 |
+
35,
|
| 41 |
+
36,
|
| 42 |
+
37,
|
| 43 |
+
38,
|
| 44 |
+
39,
|
| 45 |
+
40,
|
| 46 |
+
41,
|
| 47 |
+
42,
|
| 48 |
+
43,
|
| 49 |
+
44,
|
| 50 |
+
45,
|
| 51 |
+
46,
|
| 52 |
+
47
|
| 53 |
+
],
|
| 54 |
+
"slots": {
|
| 55 |
+
"num_slots": 24,
|
| 56 |
+
"ranges": {
|
| 57 |
+
"noul": [
|
| 58 |
+
0,
|
| 59 |
+
2
|
| 60 |
+
],
|
| 61 |
+
"score": [
|
| 62 |
+
2,
|
| 63 |
+
8
|
| 64 |
+
],
|
| 65 |
+
"choice": [
|
| 66 |
+
8,
|
| 67 |
+
24
|
| 68 |
+
]
|
| 69 |
+
},
|
| 70 |
+
"verbalizers": [
|
| 71 |
+
"false",
|
| 72 |
+
"true",
|
| 73 |
+
"0",
|
| 74 |
+
"1",
|
| 75 |
+
"2",
|
| 76 |
+
"3",
|
| 77 |
+
"4",
|
| 78 |
+
"5",
|
| 79 |
+
"A",
|
| 80 |
+
"B",
|
| 81 |
+
"C",
|
| 82 |
+
"D",
|
| 83 |
+
"E",
|
| 84 |
+
"F",
|
| 85 |
+
"G",
|
| 86 |
+
"H",
|
| 87 |
+
"I",
|
| 88 |
+
"J",
|
| 89 |
+
"K",
|
| 90 |
+
"L",
|
| 91 |
+
"M",
|
| 92 |
+
"N",
|
| 93 |
+
"O",
|
| 94 |
+
"P"
|
| 95 |
+
],
|
| 96 |
+
"template_version": "bare-v1"
|
| 97 |
+
},
|
| 98 |
+
"note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature"
|
| 99 |
+
}
|
meta_only/podA/e15/F1-pool-ts1/ckpt-4000/calibration.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"version": 1,
|
| 3 |
+
"per_kind": {
|
| 4 |
+
"noul": 1.0395277106542806,
|
| 5 |
+
"choice": 1.0,
|
| 6 |
+
"score": 1.0
|
| 7 |
+
},
|
| 8 |
+
"fit": {
|
| 9 |
+
"noul": {
|
| 10 |
+
"n": 2000,
|
| 11 |
+
"T": 1.0395277106542806,
|
| 12 |
+
"ce_before": 0.5319774150848389,
|
| 13 |
+
"ce_after": 0.5318149328231812
|
| 14 |
+
}
|
| 15 |
+
},
|
| 16 |
+
"source": "models/e15/F1-pool-ts1/ckpt-4000"
|
| 17 |
+
}
|
meta_only/podA/e15/F1-pool-ts1/ckpt-4000/judge_config.json
ADDED
|
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"base_model_path": "/root/models/Qwen3.5-9B",
|
| 3 |
+
"hidden_size": 4096,
|
| 4 |
+
"slots": {
|
| 5 |
+
"num_slots": 24,
|
| 6 |
+
"ranges": {
|
| 7 |
+
"noul": [
|
| 8 |
+
0,
|
| 9 |
+
2
|
| 10 |
+
],
|
| 11 |
+
"score": [
|
| 12 |
+
2,
|
| 13 |
+
8
|
| 14 |
+
],
|
| 15 |
+
"choice": [
|
| 16 |
+
8,
|
| 17 |
+
24
|
| 18 |
+
]
|
| 19 |
+
},
|
| 20 |
+
"verbalizers": [
|
| 21 |
+
"false",
|
| 22 |
+
"true",
|
| 23 |
+
"0",
|
| 24 |
+
"1",
|
| 25 |
+
"2",
|
| 26 |
+
"3",
|
| 27 |
+
"4",
|
| 28 |
+
"5",
|
| 29 |
+
"A",
|
| 30 |
+
"B",
|
| 31 |
+
"C",
|
| 32 |
+
"D",
|
| 33 |
+
"E",
|
| 34 |
+
"F",
|
| 35 |
+
"G",
|
| 36 |
+
"H",
|
| 37 |
+
"I",
|
| 38 |
+
"J",
|
| 39 |
+
"K",
|
| 40 |
+
"L",
|
| 41 |
+
"M",
|
| 42 |
+
"N",
|
| 43 |
+
"O",
|
| 44 |
+
"P"
|
| 45 |
+
],
|
| 46 |
+
"template_version": "bare-v1"
|
| 47 |
+
},
|
| 48 |
+
"verbalizer_ids": [
|
| 49 |
+
3721,
|
| 50 |
+
1802,
|
| 51 |
+
15,
|
| 52 |
+
16,
|
| 53 |
+
17,
|
| 54 |
+
18,
|
| 55 |
+
19,
|
| 56 |
+
20,
|
| 57 |
+
32,
|
| 58 |
+
33,
|
| 59 |
+
34,
|
| 60 |
+
35,
|
| 61 |
+
36,
|
| 62 |
+
37,
|
| 63 |
+
38,
|
| 64 |
+
39,
|
| 65 |
+
40,
|
| 66 |
+
41,
|
| 67 |
+
42,
|
| 68 |
+
43,
|
| 69 |
+
44,
|
| 70 |
+
45,
|
| 71 |
+
46,
|
| 72 |
+
47
|
| 73 |
+
],
|
| 74 |
+
"kinds": [
|
| 75 |
+
"noul",
|
| 76 |
+
"choice",
|
| 77 |
+
"score"
|
| 78 |
+
],
|
| 79 |
+
"model_name": "jev-judge-qwen35-9b",
|
| 80 |
+
"model_version": "0.8.0",
|
| 81 |
+
"lm_head_included": true,
|
| 82 |
+
"weights_mode": "unmerged",
|
| 83 |
+
"adapter_subfolder": "adapter",
|
| 84 |
+
"lora": {
|
| 85 |
+
"r": 16,
|
| 86 |
+
"lora_alpha": 32,
|
| 87 |
+
"lora_dropout": 0.05,
|
| 88 |
+
"target_modules": [
|
| 89 |
+
"down_proj",
|
| 90 |
+
"gate_proj",
|
| 91 |
+
"in_proj_qkv",
|
| 92 |
+
"in_proj_z",
|
| 93 |
+
"k_proj",
|
| 94 |
+
"o_proj",
|
| 95 |
+
"out_proj",
|
| 96 |
+
"q_proj",
|
| 97 |
+
"up_proj",
|
| 98 |
+
"v_proj"
|
| 99 |
+
]
|
| 100 |
+
},
|
| 101 |
+
"trained_stage": "e15",
|
| 102 |
+
"source_checkpoint": "models/e15/F1-pool-ts1/ckpt-4000",
|
| 103 |
+
"softcap": null,
|
| 104 |
+
"e15": {
|
| 105 |
+
"data": "data/e15/F1-pool",
|
| 106 |
+
"data_manifest": {
|
| 107 |
+
"arm": "F1-pool",
|
| 108 |
+
"name": "F1-pool",
|
| 109 |
+
"train_sha256": "dad16f94a5659768b254a6c12538c01f9c64cd43d574a3435bcc70c9c1bac130",
|
| 110 |
+
"n_train": 64000
|
| 111 |
+
},
|
| 112 |
+
"init": "released",
|
| 113 |
+
"lr": 0.0001,
|
| 114 |
+
"lr_head": 0.0002,
|
| 115 |
+
"rows_per_step": 32,
|
| 116 |
+
"skip_rows": 0,
|
| 117 |
+
"epochs": 1,
|
| 118 |
+
"max_len": 3072,
|
| 119 |
+
"schedule": "warmup 0.03 + cosine to x0.05 over 500 steps",
|
| 120 |
+
"total_rows": 16000,
|
| 121 |
+
"lora": {
|
| 122 |
+
"r": 16,
|
| 123 |
+
"lora_alpha": 32,
|
| 124 |
+
"lora_dropout": 0.05
|
| 125 |
+
},
|
| 126 |
+
"train_seed": 1,
|
| 127 |
+
"event": "ckpt",
|
| 128 |
+
"tag": "ckpt-4000",
|
| 129 |
+
"rows_seen": 4000,
|
| 130 |
+
"step": 125,
|
| 131 |
+
"tokens": 2634924,
|
| 132 |
+
"tok_per_s": 725.6523934927723,
|
| 133 |
+
"val": {
|
| 134 |
+
"n": 2000,
|
| 135 |
+
"dropped_long": 0,
|
| 136 |
+
"noul_ce": 0.5319774150848389,
|
| 137 |
+
"kt_auc": 0.7736847570145734,
|
| 138 |
+
"kt_acc": 0.7315,
|
| 139 |
+
"kt_f1": 0.8035126234906695,
|
| 140 |
+
"kt_ece": 0.036713274300098415,
|
| 141 |
+
"kt_nll": 0.5319773827847093
|
| 142 |
+
},
|
| 143 |
+
"dropped_long": 0,
|
| 144 |
+
"peak_gib": 18.233368396759033
|
| 145 |
+
}
|
| 146 |
+
}
|
meta_only/podA/e15/F1-pool-ts1/train_log.jsonl
ADDED
|
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"time": "20:33:08", "event": "start", "model_dir": "/workspace/jev_kt/models/hf/JEV-9B", "init": "released", "device": "cuda", "load_s": 102.64752459526062, "trainable_lora": 40108032, "head": 98328}
|
| 2 |
+
{"time": "20:33:08", "event": "run", "data": "data/e15/F1-pool", "out": "models/e15/F1-pool-ts1"}
|
| 3 |
+
{"time": "20:37:00", "event": "step", "step": 10, "rows_seen": 320, "loss": 0.45051592838717625, "lr": 7.333333333333333e-05, "tok_per_s": 919.539078877024}
|
| 4 |
+
{"time": "20:40:48", "event": "step", "step": 20, "rows_seen": 640, "loss": 0.43134723883122206, "lr": 9.997508952966197e-05, "tok_per_s": 923.764315435565}
|
| 5 |
+
{"time": "20:44:39", "event": "step", "step": 30, "rows_seen": 960, "loss": 0.4748785784468055, "lr": 9.977596250541417e-05, "tok_per_s": 928.1489288872847}
|
| 6 |
+
{"time": "20:52:03", "event": "ckpt", "tag": "ckpt-1000", "rows_seen": 1024, "step": 32, "tokens": 684623, "tok_per_s": 604.1670308900905, "val": {"n": 2000, "dropped_long": 0, "noul_ce": 0.6150162816047668, "kt_auc": 0.7537879422243212, "kt_acc": 0.694, "kt_f1": 0.8083907326236693, "kt_ece": 0.14204001927375792, "kt_nll": 0.6150163156474592}, "dropped_long": 0, "peak_gib": 18.229214191436768}
|
| 7 |
+
{"time": "20:55:05", "event": "step", "step": 40, "rows_seen": 1280, "loss": 0.7064548693597317, "lr": 9.937854366485982e-05, "tok_per_s": 648.5008595879774}
|
| 8 |
+
{"time": "20:58:43", "event": "step", "step": 50, "rows_seen": 1600, "loss": 0.6333137471228838, "lr": 9.8784499920729e-05, "tok_per_s": 687.4788491743927}
|
| 9 |
+
{"time": "21:02:22", "event": "step", "step": 60, "rows_seen": 1920, "loss": 0.45100219175219536, "lr": 9.79963228989294e-05, "tok_per_s": 718.1501770376802}
|
| 10 |
+
{"time": "21:06:10", "event": "step", "step": 70, "rows_seen": 2240, "loss": 0.6964020365849137, "lr": 9.701731848780134e-05, "tok_per_s": 742.9895911338607}
|
| 11 |
+
{"time": "21:09:56", "event": "step", "step": 80, "rows_seen": 2560, "loss": 0.5738761178217828, "lr": 9.585159297207315e-05, "tok_per_s": 761.0523220361788}
|
| 12 |
+
{"time": "21:13:43", "event": "step", "step": 90, "rows_seen": 2880, "loss": 0.7960144877433777, "lr": 9.450403580967612e-05, "tok_per_s": 775.4950693956374}
|
| 13 |
+
{"time": "21:17:33", "event": "step", "step": 100, "rows_seen": 3200, "loss": 0.5445596191566437, "lr": 9.298029912365855e-05, "tok_per_s": 787.9801010184477}
|
| 14 |
+
{"time": "21:21:18", "event": "step", "step": 110, "rows_seen": 3520, "loss": 0.5469026081264019, "lr": 9.128677399521752e-05, "tok_per_s": 799.4004888378769}
|
| 15 |
+
{"time": "21:25:10", "event": "step", "step": 120, "rows_seen": 3840, "loss": 0.6618786524049938, "lr": 8.943056365728335e-05, "tok_per_s": 809.4959356233697}
|
| 16 |
+
{"time": "21:33:41", "event": "ckpt", "tag": "ckpt-4000", "rows_seen": 4000, "step": 125, "tokens": 2634924, "tok_per_s": 725.6523934927723, "val": {"n": 2000, "dropped_long": 0, "noul_ce": 0.5319774150848389, "kt_auc": 0.7736847570145734, "kt_acc": 0.7315, "kt_f1": 0.8035126234906695, "kt_ece": 0.036713274300098415, "kt_nll": 0.5319773827847093}, "dropped_long": 0, "peak_gib": 18.233368396759033}
|
| 17 |
+
{"time": "21:35:33", "event": "step", "step": 130, "rows_seen": 4160, "loss": 0.3973365316633135, "lr": 8.741945370109202e-05, "tok_per_s": 731.3993931035988}
|
| 18 |
+
{"time": "21:39:21", "event": "step", "step": 140, "rows_seen": 4480, "loss": 0.46528124809265137, "lr": 8.526187942070997e-05, "tok_per_s": 742.3691403894654}
|
| 19 |
+
{"time": "21:43:12", "event": "step", "step": 150, "rows_seen": 4800, "loss": 0.5147384945303202, "lr": 8.296689043247915e-05, "tok_per_s": 752.5023062816863}
|
| 20 |
+
{"time": "21:47:01", "event": "step", "step": 160, "rows_seen": 5120, "loss": 0.5509811383672059, "lr": 8.054411271778132e-05, "tok_per_s": 761.7635619060982}
|
| 21 |
+
{"time": "21:50:54", "event": "step", "step": 170, "rows_seen": 5440, "loss": 0.5749849639832973, "lr": 7.800370824832674e-05, "tok_per_s": 769.639156845915}
|
| 22 |
+
{"time": "21:54:45", "event": "step", "step": 180, "rows_seen": 5760, "loss": 0.6066065253689885, "lr": 7.535633236331324e-05, "tok_per_s": 777.2214625198451}
|
| 23 |
+
{"time": "21:58:32", "event": "step", "step": 190, "rows_seen": 6080, "loss": 0.5618765652179718, "lr": 7.26130890772308e-05, "tok_per_s": 784.447168278369}
|
| 24 |
+
{"time": "22:02:12", "event": "step", "step": 200, "rows_seen": 6400, "loss": 0.5502310581505299, "lr": 6.978548450576568e-05, "tok_per_s": 790.1599212422134}
|
| 25 |
+
{"time": "22:05:53", "event": "step", "step": 210, "rows_seen": 6720, "loss": 0.57252436876297, "lr": 6.688537860515266e-05, "tok_per_s": 795.9475022133475}
|
| 26 |
+
{"time": "22:09:37", "event": "step", "step": 220, "rows_seen": 7040, "loss": 0.36584951635450125, "lr": 6.392493542739706e-05, "tok_per_s": 800.5920666218658}
|
| 27 |
+
{"time": "22:13:17", "event": "step", "step": 230, "rows_seen": 7360, "loss": 0.8337856869911775, "lr": 6.091657210001309e-05, "tok_per_s": 805.1741604861654}
|
| 28 |
+
{"time": "22:17:01", "event": "step", "step": 240, "rows_seen": 7680, "loss": 0.6638323441147804, "lr": 5.7872906744276015e-05, "tok_per_s": 809.425301418376}
|
| 29 |
+
{"time": "22:20:54", "event": "step", "step": 250, "rows_seen": 8000, "loss": 0.4929341431707144, "lr": 5.48067055504363e-05, "tok_per_s": 813.7053041692927}
|
| 30 |
+
{"time": "22:24:39", "event": "step", "step": 260, "rows_seen": 8320, "loss": 0.5219788514077663, "lr": 5.173082923188119e-05, "tok_per_s": 817.7422170106252}
|
| 31 |
+
{"time": "22:28:37", "event": "step", "step": 270, "rows_seen": 8640, "loss": 0.6068589836359024, "lr": 4.865817908283331e-05, "tok_per_s": 821.3578963276943}
|
| 32 |
+
{"time": "22:32:24", "event": "step", "step": 280, "rows_seen": 8960, "loss": 0.5025558490306139, "lr": 4.560164286583922e-05, "tok_per_s": 824.494419323579}
|
| 33 |
+
{"time": "22:36:10", "event": "step", "step": 290, "rows_seen": 9280, "loss": 0.5852738108951598, "lr": 4.257404075601443e-05, "tok_per_s": 827.5627468267276}
|
| 34 |
+
{"time": "22:39:57", "event": "step", "step": 300, "rows_seen": 9600, "loss": 0.45266949478536844, "lr": 3.958807156877371e-05, "tok_per_s": 830.7154937202417}
|
| 35 |
+
{"time": "22:43:33", "event": "step", "step": 310, "rows_seen": 9920, "loss": 0.4867369693238288, "lr": 3.665625949658603e-05, "tok_per_s": 833.3502588249851}
|
| 36 |
+
{"time": "22:47:15", "event": "step", "step": 320, "rows_seen": 10240, "loss": 0.5434638634324074, "lr": 3.379090157815887e-05, "tok_per_s": 835.9340570527845}
|
| 37 |
+
{"time": "22:51:02", "event": "step", "step": 330, "rows_seen": 10560, "loss": 0.474417710211128, "lr": 3.10040161203844e-05, "tok_per_s": 838.5074372044842}
|
| 38 |
+
{"time": "22:54:52", "event": "step", "step": 340, "rows_seen": 10880, "loss": 0.36452322266995907, "lr": 2.8307292289383776e-05, "tok_per_s": 840.722595539541}
|
| 39 |
+
{"time": "22:58:41", "event": "step", "step": 350, "rows_seen": 11200, "loss": 0.5055600453633815, "lr": 2.571204108208194e-05, "tok_per_s": 843.006768122122}
|
| 40 |
+
{"time": "23:02:29", "event": "step", "step": 360, "rows_seen": 11520, "loss": 0.4144336171448231, "lr": 2.322914788395595e-05, "tok_per_s": 845.3261040912352}
|
| 41 |
+
{"time": "23:06:08", "event": "step", "step": 370, "rows_seen": 11840, "loss": 0.5108116087503731, "lr": 2.0869026811945157e-05, "tok_per_s": 847.4822648312784}
|
| 42 |
+
{"time": "23:09:59", "event": "step", "step": 380, "rows_seen": 12160, "loss": 0.5361540941521525, "lr": 1.8641577034025627e-05, "tok_per_s": 849.4573283326233}
|
| 43 |
+
{"time": "23:13:46", "event": "step", "step": 390, "rows_seen": 12480, "loss": 0.5968066081404686, "lr": 1.6556141248659472e-05, "tok_per_s": 851.7579833231458}
|
| 44 |
+
{"time": "23:17:40", "event": "step", "step": 400, "rows_seen": 12800, "loss": 0.5564720034599304, "lr": 1.4621466498270773e-05, "tok_per_s": 853.4727288473566}
|
| 45 |
+
{"time": "23:21:30", "event": "step", "step": 410, "rows_seen": 13120, "loss": 0.5119495200924575, "lr": 1.284566748110963e-05, "tok_per_s": 854.9119117829106}
|
| 46 |
+
{"time": "23:25:19", "event": "step", "step": 420, "rows_seen": 13440, "loss": 0.499324104283005, "lr": 1.1236192515387821e-05, "tok_per_s": 856.6259480887267}
|
| 47 |
+
{"time": "23:29:06", "event": "step", "step": 430, "rows_seen": 13760, "loss": 0.39226976077770814, "lr": 9.799792298443728e-06, "tok_per_s": 858.2744816821305}
|
| 48 |
+
{"time": "23:32:55", "event": "step", "step": 440, "rows_seen": 14080, "loss": 0.7001174576580524, "lr": 8.542491591971554e-06, "tok_per_s": 859.9450850246118}
|
| 49 |
+
{"time": "23:36:46", "event": "step", "step": 450, "rows_seen": 14400, "loss": 0.6050784513354301, "lr": 7.469563952077155e-06, "tok_per_s": 861.3215771805565}
|
| 50 |
+
{"time": "23:40:37", "event": "step", "step": 460, "rows_seen": 14720, "loss": 0.6117255276767537, "lr": 6.585509610150972e-06, "tok_per_s": 862.8705987518805}
|
| 51 |
+
{"time": "23:44:19", "event": "step", "step": 470, "rows_seen": 15040, "loss": 0.4003585472237319, "lr": 5.894036597333646e-06, "tok_per_s": 864.0987724045657}
|
| 52 |
+
{"time": "23:48:11", "event": "step", "step": 480, "rows_seen": 15360, "loss": 0.6546561303548515, "lr": 5.398045191744809e-06, "tok_per_s": 865.1820534663028}
|
| 53 |
+
{"time": "23:51:55", "event": "step", "step": 490, "rows_seen": 15680, "loss": 0.522389056161046, "lr": 5.099615753708697e-06, "tok_per_s": 866.4499695757102}
|
| 54 |
+
{"time": "23:55:41", "event": "step", "step": 500, "rows_seen": 16000, "loss": 0.5392244383692741, "lr": 5e-06, "tok_per_s": 867.6505154736941}
|
| 55 |
+
{"time": "00:02:15", "event": "ckpt", "tag": "ckpt-16000", "rows_seen": 16000, "step": 500, "tokens": 10544036, "tok_per_s": 840.4803245529926, "val": {"n": 2000, "dropped_long": 0, "noul_ce": 0.5292231440544128, "kt_auc": 0.7788161054672977, "kt_acc": 0.734, "kt_f1": 0.8044117647058824, "kt_ece": 0.04880642385780811, "kt_nll": 0.5292231264216826}, "dropped_long": 0, "peak_gib": 18.23512029647827}
|
meta_only/podA/e15/F1-pool-ts2/ckpt-1000/adapter/adapter_config.json
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B",
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 32,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.05,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 16,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"q_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"o_proj",
|
| 41 |
+
"v_proj",
|
| 42 |
+
"down_proj",
|
| 43 |
+
"k_proj",
|
| 44 |
+
"up_proj",
|
| 45 |
+
"in_proj_z",
|
| 46 |
+
"out_proj",
|
| 47 |
+
"in_proj_qkv"
|
| 48 |
+
],
|
| 49 |
+
"target_parameters": null,
|
| 50 |
+
"task_type": null,
|
| 51 |
+
"trainable_token_indices": null,
|
| 52 |
+
"use_bdlora": null,
|
| 53 |
+
"use_dora": false,
|
| 54 |
+
"use_qalora": false,
|
| 55 |
+
"use_rslora": false,
|
| 56 |
+
"velora_config": null
|
| 57 |
+
}
|
meta_only/podA/e15/F1-pool-ts2/ckpt-1000/adapter_vllm/adapter_config.json
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": null,
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 64.0,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.0,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 32,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"down_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"in_proj_qkv",
|
| 41 |
+
"in_proj_z",
|
| 42 |
+
"k_proj",
|
| 43 |
+
"lm_head",
|
| 44 |
+
"o_proj",
|
| 45 |
+
"out_proj",
|
| 46 |
+
"q_proj",
|
| 47 |
+
"up_proj",
|
| 48 |
+
"v_proj"
|
| 49 |
+
],
|
| 50 |
+
"target_parameters": null,
|
| 51 |
+
"task_type": null,
|
| 52 |
+
"trainable_token_indices": null,
|
| 53 |
+
"use_bdlora": null,
|
| 54 |
+
"use_dora": false,
|
| 55 |
+
"use_qalora": false,
|
| 56 |
+
"use_rslora": false,
|
| 57 |
+
"velora_config": null
|
| 58 |
+
}
|
meta_only/podA/e15/F1-pool-ts2/ckpt-1000/adapter_vllm/decision_head.json
ADDED
|
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bias": [
|
| 3 |
+
0.00030852665076963603,
|
| 4 |
+
-0.00030877452809363604,
|
| 5 |
+
-0.0025878134183585644,
|
| 6 |
+
-0.0012615364976227283,
|
| 7 |
+
-0.00025007399381138384,
|
| 8 |
+
0.0019558551721274853,
|
| 9 |
+
0.0013559526996687055,
|
| 10 |
+
-0.0023348911199718714,
|
| 11 |
+
-0.0004082925443071872,
|
| 12 |
+
-0.0002078723773593083,
|
| 13 |
+
-1.2957353646925185e-05,
|
| 14 |
+
0.0001676026004133746,
|
| 15 |
+
0.0005569425993598998,
|
| 16 |
+
-0.001412136247381568,
|
| 17 |
+
-0.0028663526754826307,
|
| 18 |
+
-0.005061559844762087,
|
| 19 |
+
-0.00593328895047307,
|
| 20 |
+
-0.004319756291806698,
|
| 21 |
+
-0.004922541789710522,
|
| 22 |
+
-0.0049067651852965355,
|
| 23 |
+
-0.003943523392081261,
|
| 24 |
+
-0.003319827374070883,
|
| 25 |
+
-0.005581587553024292,
|
| 26 |
+
-0.0034480984322726727
|
| 27 |
+
],
|
| 28 |
+
"verbalizer_ids": [
|
| 29 |
+
3721,
|
| 30 |
+
1802,
|
| 31 |
+
15,
|
| 32 |
+
16,
|
| 33 |
+
17,
|
| 34 |
+
18,
|
| 35 |
+
19,
|
| 36 |
+
20,
|
| 37 |
+
32,
|
| 38 |
+
33,
|
| 39 |
+
34,
|
| 40 |
+
35,
|
| 41 |
+
36,
|
| 42 |
+
37,
|
| 43 |
+
38,
|
| 44 |
+
39,
|
| 45 |
+
40,
|
| 46 |
+
41,
|
| 47 |
+
42,
|
| 48 |
+
43,
|
| 49 |
+
44,
|
| 50 |
+
45,
|
| 51 |
+
46,
|
| 52 |
+
47
|
| 53 |
+
],
|
| 54 |
+
"slots": {
|
| 55 |
+
"num_slots": 24,
|
| 56 |
+
"ranges": {
|
| 57 |
+
"noul": [
|
| 58 |
+
0,
|
| 59 |
+
2
|
| 60 |
+
],
|
| 61 |
+
"score": [
|
| 62 |
+
2,
|
| 63 |
+
8
|
| 64 |
+
],
|
| 65 |
+
"choice": [
|
| 66 |
+
8,
|
| 67 |
+
24
|
| 68 |
+
]
|
| 69 |
+
},
|
| 70 |
+
"verbalizers": [
|
| 71 |
+
"false",
|
| 72 |
+
"true",
|
| 73 |
+
"0",
|
| 74 |
+
"1",
|
| 75 |
+
"2",
|
| 76 |
+
"3",
|
| 77 |
+
"4",
|
| 78 |
+
"5",
|
| 79 |
+
"A",
|
| 80 |
+
"B",
|
| 81 |
+
"C",
|
| 82 |
+
"D",
|
| 83 |
+
"E",
|
| 84 |
+
"F",
|
| 85 |
+
"G",
|
| 86 |
+
"H",
|
| 87 |
+
"I",
|
| 88 |
+
"J",
|
| 89 |
+
"K",
|
| 90 |
+
"L",
|
| 91 |
+
"M",
|
| 92 |
+
"N",
|
| 93 |
+
"O",
|
| 94 |
+
"P"
|
| 95 |
+
],
|
| 96 |
+
"template_version": "bare-v1"
|
| 97 |
+
},
|
| 98 |
+
"note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature"
|
| 99 |
+
}
|
meta_only/podA/e15/F1-pool-ts2/ckpt-1000/calibration.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"version": 1,
|
| 3 |
+
"per_kind": {
|
| 4 |
+
"noul": 0.8774969833752866,
|
| 5 |
+
"choice": 1.0,
|
| 6 |
+
"score": 1.0
|
| 7 |
+
},
|
| 8 |
+
"fit": {
|
| 9 |
+
"noul": {
|
| 10 |
+
"n": 2000,
|
| 11 |
+
"T": 0.8774969833752866,
|
| 12 |
+
"ce_before": 0.5506089925765991,
|
| 13 |
+
"ce_after": 0.5488398671150208
|
| 14 |
+
}
|
| 15 |
+
},
|
| 16 |
+
"source": "models/e15/F1-pool-ts2/ckpt-1000"
|
| 17 |
+
}
|
meta_only/podA/e15/F1-pool-ts2/ckpt-1000/judge_config.json
ADDED
|
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"base_model_path": "/root/models/Qwen3.5-9B",
|
| 3 |
+
"hidden_size": 4096,
|
| 4 |
+
"slots": {
|
| 5 |
+
"num_slots": 24,
|
| 6 |
+
"ranges": {
|
| 7 |
+
"noul": [
|
| 8 |
+
0,
|
| 9 |
+
2
|
| 10 |
+
],
|
| 11 |
+
"score": [
|
| 12 |
+
2,
|
| 13 |
+
8
|
| 14 |
+
],
|
| 15 |
+
"choice": [
|
| 16 |
+
8,
|
| 17 |
+
24
|
| 18 |
+
]
|
| 19 |
+
},
|
| 20 |
+
"verbalizers": [
|
| 21 |
+
"false",
|
| 22 |
+
"true",
|
| 23 |
+
"0",
|
| 24 |
+
"1",
|
| 25 |
+
"2",
|
| 26 |
+
"3",
|
| 27 |
+
"4",
|
| 28 |
+
"5",
|
| 29 |
+
"A",
|
| 30 |
+
"B",
|
| 31 |
+
"C",
|
| 32 |
+
"D",
|
| 33 |
+
"E",
|
| 34 |
+
"F",
|
| 35 |
+
"G",
|
| 36 |
+
"H",
|
| 37 |
+
"I",
|
| 38 |
+
"J",
|
| 39 |
+
"K",
|
| 40 |
+
"L",
|
| 41 |
+
"M",
|
| 42 |
+
"N",
|
| 43 |
+
"O",
|
| 44 |
+
"P"
|
| 45 |
+
],
|
| 46 |
+
"template_version": "bare-v1"
|
| 47 |
+
},
|
| 48 |
+
"verbalizer_ids": [
|
| 49 |
+
3721,
|
| 50 |
+
1802,
|
| 51 |
+
15,
|
| 52 |
+
16,
|
| 53 |
+
17,
|
| 54 |
+
18,
|
| 55 |
+
19,
|
| 56 |
+
20,
|
| 57 |
+
32,
|
| 58 |
+
33,
|
| 59 |
+
34,
|
| 60 |
+
35,
|
| 61 |
+
36,
|
| 62 |
+
37,
|
| 63 |
+
38,
|
| 64 |
+
39,
|
| 65 |
+
40,
|
| 66 |
+
41,
|
| 67 |
+
42,
|
| 68 |
+
43,
|
| 69 |
+
44,
|
| 70 |
+
45,
|
| 71 |
+
46,
|
| 72 |
+
47
|
| 73 |
+
],
|
| 74 |
+
"kinds": [
|
| 75 |
+
"noul",
|
| 76 |
+
"choice",
|
| 77 |
+
"score"
|
| 78 |
+
],
|
| 79 |
+
"model_name": "jev-judge-qwen35-9b",
|
| 80 |
+
"model_version": "0.8.0",
|
| 81 |
+
"lm_head_included": true,
|
| 82 |
+
"weights_mode": "unmerged",
|
| 83 |
+
"adapter_subfolder": "adapter",
|
| 84 |
+
"lora": {
|
| 85 |
+
"r": 16,
|
| 86 |
+
"lora_alpha": 32,
|
| 87 |
+
"lora_dropout": 0.05,
|
| 88 |
+
"target_modules": [
|
| 89 |
+
"down_proj",
|
| 90 |
+
"gate_proj",
|
| 91 |
+
"in_proj_qkv",
|
| 92 |
+
"in_proj_z",
|
| 93 |
+
"k_proj",
|
| 94 |
+
"o_proj",
|
| 95 |
+
"out_proj",
|
| 96 |
+
"q_proj",
|
| 97 |
+
"up_proj",
|
| 98 |
+
"v_proj"
|
| 99 |
+
]
|
| 100 |
+
},
|
| 101 |
+
"trained_stage": "e15",
|
| 102 |
+
"source_checkpoint": "models/e15/F1-pool-ts2/ckpt-1000",
|
| 103 |
+
"softcap": null,
|
| 104 |
+
"e15": {
|
| 105 |
+
"data": "data/e15/F1-pool",
|
| 106 |
+
"data_manifest": {
|
| 107 |
+
"arm": "F1-pool",
|
| 108 |
+
"name": "F1-pool",
|
| 109 |
+
"train_sha256": "dad16f94a5659768b254a6c12538c01f9c64cd43d574a3435bcc70c9c1bac130",
|
| 110 |
+
"n_train": 64000
|
| 111 |
+
},
|
| 112 |
+
"init": "released",
|
| 113 |
+
"lr": 0.0001,
|
| 114 |
+
"lr_head": 0.0002,
|
| 115 |
+
"rows_per_step": 32,
|
| 116 |
+
"skip_rows": 0,
|
| 117 |
+
"epochs": 1,
|
| 118 |
+
"max_len": 3072,
|
| 119 |
+
"schedule": "warmup 0.03 + cosine to x0.05 over 500 steps",
|
| 120 |
+
"total_rows": 16000,
|
| 121 |
+
"lora": {
|
| 122 |
+
"r": 16,
|
| 123 |
+
"lora_alpha": 32,
|
| 124 |
+
"lora_dropout": 0.05
|
| 125 |
+
},
|
| 126 |
+
"train_seed": 2,
|
| 127 |
+
"event": "ckpt",
|
| 128 |
+
"tag": "ckpt-1000",
|
| 129 |
+
"rows_seen": 1024,
|
| 130 |
+
"step": 32,
|
| 131 |
+
"tokens": 669196,
|
| 132 |
+
"tok_per_s": 597.7086495887113,
|
| 133 |
+
"val": {
|
| 134 |
+
"n": 2000,
|
| 135 |
+
"dropped_long": 0,
|
| 136 |
+
"noul_ce": 0.5506089925765991,
|
| 137 |
+
"kt_auc": 0.7614994791614331,
|
| 138 |
+
"kt_acc": 0.733,
|
| 139 |
+
"kt_f1": 0.8065217391304348,
|
| 140 |
+
"kt_ece": 0.060642471492290496,
|
| 141 |
+
"kt_nll": 0.5506090411692858
|
| 142 |
+
},
|
| 143 |
+
"dropped_long": 0,
|
| 144 |
+
"peak_gib": 18.23512029647827
|
| 145 |
+
}
|
| 146 |
+
}
|
meta_only/podA/e15/F1-pool-ts2/ckpt-16000/adapter/adapter_config.json
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B",
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 32,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.05,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 16,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"q_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"o_proj",
|
| 41 |
+
"v_proj",
|
| 42 |
+
"down_proj",
|
| 43 |
+
"k_proj",
|
| 44 |
+
"up_proj",
|
| 45 |
+
"in_proj_z",
|
| 46 |
+
"out_proj",
|
| 47 |
+
"in_proj_qkv"
|
| 48 |
+
],
|
| 49 |
+
"target_parameters": null,
|
| 50 |
+
"task_type": null,
|
| 51 |
+
"trainable_token_indices": null,
|
| 52 |
+
"use_bdlora": null,
|
| 53 |
+
"use_dora": false,
|
| 54 |
+
"use_qalora": false,
|
| 55 |
+
"use_rslora": false,
|
| 56 |
+
"velora_config": null
|
| 57 |
+
}
|
meta_only/podA/e15/F1-pool-ts2/ckpt-16000/adapter_vllm/adapter_config.json
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": null,
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 64.0,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.0,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 32,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"down_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"in_proj_qkv",
|
| 41 |
+
"in_proj_z",
|
| 42 |
+
"k_proj",
|
| 43 |
+
"lm_head",
|
| 44 |
+
"o_proj",
|
| 45 |
+
"out_proj",
|
| 46 |
+
"q_proj",
|
| 47 |
+
"up_proj",
|
| 48 |
+
"v_proj"
|
| 49 |
+
],
|
| 50 |
+
"target_parameters": null,
|
| 51 |
+
"task_type": null,
|
| 52 |
+
"trainable_token_indices": null,
|
| 53 |
+
"use_bdlora": null,
|
| 54 |
+
"use_dora": false,
|
| 55 |
+
"use_qalora": false,
|
| 56 |
+
"use_rslora": false,
|
| 57 |
+
"velora_config": null
|
| 58 |
+
}
|
meta_only/podA/e15/F1-pool-ts2/ckpt-16000/adapter_vllm/decision_head.json
ADDED
|
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bias": [
|
| 3 |
+
-3.5705408663488925e-05,
|
| 4 |
+
3.544979335856624e-05,
|
| 5 |
+
-0.0025878134183585644,
|
| 6 |
+
-0.0012615364976227283,
|
| 7 |
+
-0.00025007399381138384,
|
| 8 |
+
0.0019558551721274853,
|
| 9 |
+
0.0013559526996687055,
|
| 10 |
+
-0.0023348911199718714,
|
| 11 |
+
-0.0004082925443071872,
|
| 12 |
+
-0.0002078723773593083,
|
| 13 |
+
-1.2957353646925185e-05,
|
| 14 |
+
0.0001676026004133746,
|
| 15 |
+
0.0005569425993598998,
|
| 16 |
+
-0.001412136247381568,
|
| 17 |
+
-0.0028663526754826307,
|
| 18 |
+
-0.005061559844762087,
|
| 19 |
+
-0.00593328895047307,
|
| 20 |
+
-0.004319756291806698,
|
| 21 |
+
-0.004922541789710522,
|
| 22 |
+
-0.0049067651852965355,
|
| 23 |
+
-0.003943523392081261,
|
| 24 |
+
-0.003319827374070883,
|
| 25 |
+
-0.005581587553024292,
|
| 26 |
+
-0.0034480984322726727
|
| 27 |
+
],
|
| 28 |
+
"verbalizer_ids": [
|
| 29 |
+
3721,
|
| 30 |
+
1802,
|
| 31 |
+
15,
|
| 32 |
+
16,
|
| 33 |
+
17,
|
| 34 |
+
18,
|
| 35 |
+
19,
|
| 36 |
+
20,
|
| 37 |
+
32,
|
| 38 |
+
33,
|
| 39 |
+
34,
|
| 40 |
+
35,
|
| 41 |
+
36,
|
| 42 |
+
37,
|
| 43 |
+
38,
|
| 44 |
+
39,
|
| 45 |
+
40,
|
| 46 |
+
41,
|
| 47 |
+
42,
|
| 48 |
+
43,
|
| 49 |
+
44,
|
| 50 |
+
45,
|
| 51 |
+
46,
|
| 52 |
+
47
|
| 53 |
+
],
|
| 54 |
+
"slots": {
|
| 55 |
+
"num_slots": 24,
|
| 56 |
+
"ranges": {
|
| 57 |
+
"noul": [
|
| 58 |
+
0,
|
| 59 |
+
2
|
| 60 |
+
],
|
| 61 |
+
"score": [
|
| 62 |
+
2,
|
| 63 |
+
8
|
| 64 |
+
],
|
| 65 |
+
"choice": [
|
| 66 |
+
8,
|
| 67 |
+
24
|
| 68 |
+
]
|
| 69 |
+
},
|
| 70 |
+
"verbalizers": [
|
| 71 |
+
"false",
|
| 72 |
+
"true",
|
| 73 |
+
"0",
|
| 74 |
+
"1",
|
| 75 |
+
"2",
|
| 76 |
+
"3",
|
| 77 |
+
"4",
|
| 78 |
+
"5",
|
| 79 |
+
"A",
|
| 80 |
+
"B",
|
| 81 |
+
"C",
|
| 82 |
+
"D",
|
| 83 |
+
"E",
|
| 84 |
+
"F",
|
| 85 |
+
"G",
|
| 86 |
+
"H",
|
| 87 |
+
"I",
|
| 88 |
+
"J",
|
| 89 |
+
"K",
|
| 90 |
+
"L",
|
| 91 |
+
"M",
|
| 92 |
+
"N",
|
| 93 |
+
"O",
|
| 94 |
+
"P"
|
| 95 |
+
],
|
| 96 |
+
"template_version": "bare-v1"
|
| 97 |
+
},
|
| 98 |
+
"note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature"
|
| 99 |
+
}
|
meta_only/podA/e15/F1-pool-ts2/ckpt-16000/calibration.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"version": 1,
|
| 3 |
+
"per_kind": {
|
| 4 |
+
"noul": 1.0486965745037178,
|
| 5 |
+
"choice": 1.0,
|
| 6 |
+
"score": 1.0
|
| 7 |
+
},
|
| 8 |
+
"fit": {
|
| 9 |
+
"noul": {
|
| 10 |
+
"n": 2000,
|
| 11 |
+
"T": 1.0486965745037178,
|
| 12 |
+
"ce_before": 0.5264289975166321,
|
| 13 |
+
"ce_after": 0.5261688232421875
|
| 14 |
+
}
|
| 15 |
+
},
|
| 16 |
+
"source": "models/e15/F1-pool-ts2/ckpt-16000"
|
| 17 |
+
}
|
meta_only/podA/e15/F1-pool-ts2/ckpt-16000/judge_config.json
ADDED
|
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"base_model_path": "/root/models/Qwen3.5-9B",
|
| 3 |
+
"hidden_size": 4096,
|
| 4 |
+
"slots": {
|
| 5 |
+
"num_slots": 24,
|
| 6 |
+
"ranges": {
|
| 7 |
+
"noul": [
|
| 8 |
+
0,
|
| 9 |
+
2
|
| 10 |
+
],
|
| 11 |
+
"score": [
|
| 12 |
+
2,
|
| 13 |
+
8
|
| 14 |
+
],
|
| 15 |
+
"choice": [
|
| 16 |
+
8,
|
| 17 |
+
24
|
| 18 |
+
]
|
| 19 |
+
},
|
| 20 |
+
"verbalizers": [
|
| 21 |
+
"false",
|
| 22 |
+
"true",
|
| 23 |
+
"0",
|
| 24 |
+
"1",
|
| 25 |
+
"2",
|
| 26 |
+
"3",
|
| 27 |
+
"4",
|
| 28 |
+
"5",
|
| 29 |
+
"A",
|
| 30 |
+
"B",
|
| 31 |
+
"C",
|
| 32 |
+
"D",
|
| 33 |
+
"E",
|
| 34 |
+
"F",
|
| 35 |
+
"G",
|
| 36 |
+
"H",
|
| 37 |
+
"I",
|
| 38 |
+
"J",
|
| 39 |
+
"K",
|
| 40 |
+
"L",
|
| 41 |
+
"M",
|
| 42 |
+
"N",
|
| 43 |
+
"O",
|
| 44 |
+
"P"
|
| 45 |
+
],
|
| 46 |
+
"template_version": "bare-v1"
|
| 47 |
+
},
|
| 48 |
+
"verbalizer_ids": [
|
| 49 |
+
3721,
|
| 50 |
+
1802,
|
| 51 |
+
15,
|
| 52 |
+
16,
|
| 53 |
+
17,
|
| 54 |
+
18,
|
| 55 |
+
19,
|
| 56 |
+
20,
|
| 57 |
+
32,
|
| 58 |
+
33,
|
| 59 |
+
34,
|
| 60 |
+
35,
|
| 61 |
+
36,
|
| 62 |
+
37,
|
| 63 |
+
38,
|
| 64 |
+
39,
|
| 65 |
+
40,
|
| 66 |
+
41,
|
| 67 |
+
42,
|
| 68 |
+
43,
|
| 69 |
+
44,
|
| 70 |
+
45,
|
| 71 |
+
46,
|
| 72 |
+
47
|
| 73 |
+
],
|
| 74 |
+
"kinds": [
|
| 75 |
+
"noul",
|
| 76 |
+
"choice",
|
| 77 |
+
"score"
|
| 78 |
+
],
|
| 79 |
+
"model_name": "jev-judge-qwen35-9b",
|
| 80 |
+
"model_version": "0.8.0",
|
| 81 |
+
"lm_head_included": true,
|
| 82 |
+
"weights_mode": "unmerged",
|
| 83 |
+
"adapter_subfolder": "adapter",
|
| 84 |
+
"lora": {
|
| 85 |
+
"r": 16,
|
| 86 |
+
"lora_alpha": 32,
|
| 87 |
+
"lora_dropout": 0.05,
|
| 88 |
+
"target_modules": [
|
| 89 |
+
"down_proj",
|
| 90 |
+
"gate_proj",
|
| 91 |
+
"in_proj_qkv",
|
| 92 |
+
"in_proj_z",
|
| 93 |
+
"k_proj",
|
| 94 |
+
"o_proj",
|
| 95 |
+
"out_proj",
|
| 96 |
+
"q_proj",
|
| 97 |
+
"up_proj",
|
| 98 |
+
"v_proj"
|
| 99 |
+
]
|
| 100 |
+
},
|
| 101 |
+
"trained_stage": "e15",
|
| 102 |
+
"source_checkpoint": "models/e15/F1-pool-ts2/ckpt-16000",
|
| 103 |
+
"softcap": null,
|
| 104 |
+
"e15": {
|
| 105 |
+
"data": "data/e15/F1-pool",
|
| 106 |
+
"data_manifest": {
|
| 107 |
+
"arm": "F1-pool",
|
| 108 |
+
"name": "F1-pool",
|
| 109 |
+
"train_sha256": "dad16f94a5659768b254a6c12538c01f9c64cd43d574a3435bcc70c9c1bac130",
|
| 110 |
+
"n_train": 64000
|
| 111 |
+
},
|
| 112 |
+
"init": "released",
|
| 113 |
+
"lr": 0.0001,
|
| 114 |
+
"lr_head": 0.0002,
|
| 115 |
+
"rows_per_step": 32,
|
| 116 |
+
"skip_rows": 0,
|
| 117 |
+
"epochs": 1,
|
| 118 |
+
"max_len": 3072,
|
| 119 |
+
"schedule": "warmup 0.03 + cosine to x0.05 over 500 steps",
|
| 120 |
+
"total_rows": 16000,
|
| 121 |
+
"lora": {
|
| 122 |
+
"r": 16,
|
| 123 |
+
"lora_alpha": 32,
|
| 124 |
+
"lora_dropout": 0.05
|
| 125 |
+
},
|
| 126 |
+
"train_seed": 2,
|
| 127 |
+
"event": "ckpt",
|
| 128 |
+
"tag": "ckpt-16000",
|
| 129 |
+
"rows_seen": 16000,
|
| 130 |
+
"step": 500,
|
| 131 |
+
"tokens": 10544036,
|
| 132 |
+
"tok_per_s": 839.5485261528129,
|
| 133 |
+
"val": {
|
| 134 |
+
"n": 2000,
|
| 135 |
+
"dropped_long": 0,
|
| 136 |
+
"noul_ce": 0.5264289975166321,
|
| 137 |
+
"kt_auc": 0.778155148486442,
|
| 138 |
+
"kt_acc": 0.734,
|
| 139 |
+
"kt_f1": 0.8098641887062187,
|
| 140 |
+
"kt_ece": 0.031232122391462322,
|
| 141 |
+
"kt_nll": 0.526428999545984
|
| 142 |
+
},
|
| 143 |
+
"dropped_long": 0,
|
| 144 |
+
"peak_gib": 18.237350940704346
|
| 145 |
+
}
|
| 146 |
+
}
|
meta_only/podA/e15/F1-pool-ts2/ckpt-4000/adapter/adapter_config.json
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B",
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 32,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.05,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 16,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"q_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"o_proj",
|
| 41 |
+
"v_proj",
|
| 42 |
+
"down_proj",
|
| 43 |
+
"k_proj",
|
| 44 |
+
"up_proj",
|
| 45 |
+
"in_proj_z",
|
| 46 |
+
"out_proj",
|
| 47 |
+
"in_proj_qkv"
|
| 48 |
+
],
|
| 49 |
+
"target_parameters": null,
|
| 50 |
+
"task_type": null,
|
| 51 |
+
"trainable_token_indices": null,
|
| 52 |
+
"use_bdlora": null,
|
| 53 |
+
"use_dora": false,
|
| 54 |
+
"use_qalora": false,
|
| 55 |
+
"use_rslora": false,
|
| 56 |
+
"velora_config": null
|
| 57 |
+
}
|
meta_only/podA/e15/F1-pool-ts2/ckpt-4000/adapter_vllm/adapter_config.json
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": null,
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 64.0,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.0,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 32,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"down_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"in_proj_qkv",
|
| 41 |
+
"in_proj_z",
|
| 42 |
+
"k_proj",
|
| 43 |
+
"lm_head",
|
| 44 |
+
"o_proj",
|
| 45 |
+
"out_proj",
|
| 46 |
+
"q_proj",
|
| 47 |
+
"up_proj",
|
| 48 |
+
"v_proj"
|
| 49 |
+
],
|
| 50 |
+
"target_parameters": null,
|
| 51 |
+
"task_type": null,
|
| 52 |
+
"trainable_token_indices": null,
|
| 53 |
+
"use_bdlora": null,
|
| 54 |
+
"use_dora": false,
|
| 55 |
+
"use_qalora": false,
|
| 56 |
+
"use_rslora": false,
|
| 57 |
+
"velora_config": null
|
| 58 |
+
}
|
meta_only/podA/e15/F1-pool-ts2/ckpt-4000/adapter_vllm/decision_head.json
ADDED
|
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bias": [
|
| 3 |
+
5.387282726587728e-06,
|
| 4 |
+
-5.637648428091779e-06,
|
| 5 |
+
-0.0025878134183585644,
|
| 6 |
+
-0.0012615364976227283,
|
| 7 |
+
-0.00025007399381138384,
|
| 8 |
+
0.0019558551721274853,
|
| 9 |
+
0.0013559526996687055,
|
| 10 |
+
-0.0023348911199718714,
|
| 11 |
+
-0.0004082925443071872,
|
| 12 |
+
-0.0002078723773593083,
|
| 13 |
+
-1.2957353646925185e-05,
|
| 14 |
+
0.0001676026004133746,
|
| 15 |
+
0.0005569425993598998,
|
| 16 |
+
-0.001412136247381568,
|
| 17 |
+
-0.0028663526754826307,
|
| 18 |
+
-0.005061559844762087,
|
| 19 |
+
-0.00593328895047307,
|
| 20 |
+
-0.004319756291806698,
|
| 21 |
+
-0.004922541789710522,
|
| 22 |
+
-0.0049067651852965355,
|
| 23 |
+
-0.003943523392081261,
|
| 24 |
+
-0.003319827374070883,
|
| 25 |
+
-0.005581587553024292,
|
| 26 |
+
-0.0034480984322726727
|
| 27 |
+
],
|
| 28 |
+
"verbalizer_ids": [
|
| 29 |
+
3721,
|
| 30 |
+
1802,
|
| 31 |
+
15,
|
| 32 |
+
16,
|
| 33 |
+
17,
|
| 34 |
+
18,
|
| 35 |
+
19,
|
| 36 |
+
20,
|
| 37 |
+
32,
|
| 38 |
+
33,
|
| 39 |
+
34,
|
| 40 |
+
35,
|
| 41 |
+
36,
|
| 42 |
+
37,
|
| 43 |
+
38,
|
| 44 |
+
39,
|
| 45 |
+
40,
|
| 46 |
+
41,
|
| 47 |
+
42,
|
| 48 |
+
43,
|
| 49 |
+
44,
|
| 50 |
+
45,
|
| 51 |
+
46,
|
| 52 |
+
47
|
| 53 |
+
],
|
| 54 |
+
"slots": {
|
| 55 |
+
"num_slots": 24,
|
| 56 |
+
"ranges": {
|
| 57 |
+
"noul": [
|
| 58 |
+
0,
|
| 59 |
+
2
|
| 60 |
+
],
|
| 61 |
+
"score": [
|
| 62 |
+
2,
|
| 63 |
+
8
|
| 64 |
+
],
|
| 65 |
+
"choice": [
|
| 66 |
+
8,
|
| 67 |
+
24
|
| 68 |
+
]
|
| 69 |
+
},
|
| 70 |
+
"verbalizers": [
|
| 71 |
+
"false",
|
| 72 |
+
"true",
|
| 73 |
+
"0",
|
| 74 |
+
"1",
|
| 75 |
+
"2",
|
| 76 |
+
"3",
|
| 77 |
+
"4",
|
| 78 |
+
"5",
|
| 79 |
+
"A",
|
| 80 |
+
"B",
|
| 81 |
+
"C",
|
| 82 |
+
"D",
|
| 83 |
+
"E",
|
| 84 |
+
"F",
|
| 85 |
+
"G",
|
| 86 |
+
"H",
|
| 87 |
+
"I",
|
| 88 |
+
"J",
|
| 89 |
+
"K",
|
| 90 |
+
"L",
|
| 91 |
+
"M",
|
| 92 |
+
"N",
|
| 93 |
+
"O",
|
| 94 |
+
"P"
|
| 95 |
+
],
|
| 96 |
+
"template_version": "bare-v1"
|
| 97 |
+
},
|
| 98 |
+
"note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature"
|
| 99 |
+
}
|
meta_only/podA/e15/F1-pool-ts2/ckpt-4000/calibration.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"version": 1,
|
| 3 |
+
"per_kind": {
|
| 4 |
+
"noul": 1.3328384732866634,
|
| 5 |
+
"choice": 1.0,
|
| 6 |
+
"score": 1.0
|
| 7 |
+
},
|
| 8 |
+
"fit": {
|
| 9 |
+
"noul": {
|
| 10 |
+
"n": 2000,
|
| 11 |
+
"T": 1.3328384732866634,
|
| 12 |
+
"ce_before": 0.5441238880157471,
|
| 13 |
+
"ce_after": 0.5343976020812988
|
| 14 |
+
}
|
| 15 |
+
},
|
| 16 |
+
"source": "models/e15/F1-pool-ts2/ckpt-4000"
|
| 17 |
+
}
|
meta_only/podA/e15/F1-pool-ts2/ckpt-4000/judge_config.json
ADDED
|
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"base_model_path": "/root/models/Qwen3.5-9B",
|
| 3 |
+
"hidden_size": 4096,
|
| 4 |
+
"slots": {
|
| 5 |
+
"num_slots": 24,
|
| 6 |
+
"ranges": {
|
| 7 |
+
"noul": [
|
| 8 |
+
0,
|
| 9 |
+
2
|
| 10 |
+
],
|
| 11 |
+
"score": [
|
| 12 |
+
2,
|
| 13 |
+
8
|
| 14 |
+
],
|
| 15 |
+
"choice": [
|
| 16 |
+
8,
|
| 17 |
+
24
|
| 18 |
+
]
|
| 19 |
+
},
|
| 20 |
+
"verbalizers": [
|
| 21 |
+
"false",
|
| 22 |
+
"true",
|
| 23 |
+
"0",
|
| 24 |
+
"1",
|
| 25 |
+
"2",
|
| 26 |
+
"3",
|
| 27 |
+
"4",
|
| 28 |
+
"5",
|
| 29 |
+
"A",
|
| 30 |
+
"B",
|
| 31 |
+
"C",
|
| 32 |
+
"D",
|
| 33 |
+
"E",
|
| 34 |
+
"F",
|
| 35 |
+
"G",
|
| 36 |
+
"H",
|
| 37 |
+
"I",
|
| 38 |
+
"J",
|
| 39 |
+
"K",
|
| 40 |
+
"L",
|
| 41 |
+
"M",
|
| 42 |
+
"N",
|
| 43 |
+
"O",
|
| 44 |
+
"P"
|
| 45 |
+
],
|
| 46 |
+
"template_version": "bare-v1"
|
| 47 |
+
},
|
| 48 |
+
"verbalizer_ids": [
|
| 49 |
+
3721,
|
| 50 |
+
1802,
|
| 51 |
+
15,
|
| 52 |
+
16,
|
| 53 |
+
17,
|
| 54 |
+
18,
|
| 55 |
+
19,
|
| 56 |
+
20,
|
| 57 |
+
32,
|
| 58 |
+
33,
|
| 59 |
+
34,
|
| 60 |
+
35,
|
| 61 |
+
36,
|
| 62 |
+
37,
|
| 63 |
+
38,
|
| 64 |
+
39,
|
| 65 |
+
40,
|
| 66 |
+
41,
|
| 67 |
+
42,
|
| 68 |
+
43,
|
| 69 |
+
44,
|
| 70 |
+
45,
|
| 71 |
+
46,
|
| 72 |
+
47
|
| 73 |
+
],
|
| 74 |
+
"kinds": [
|
| 75 |
+
"noul",
|
| 76 |
+
"choice",
|
| 77 |
+
"score"
|
| 78 |
+
],
|
| 79 |
+
"model_name": "jev-judge-qwen35-9b",
|
| 80 |
+
"model_version": "0.8.0",
|
| 81 |
+
"lm_head_included": true,
|
| 82 |
+
"weights_mode": "unmerged",
|
| 83 |
+
"adapter_subfolder": "adapter",
|
| 84 |
+
"lora": {
|
| 85 |
+
"r": 16,
|
| 86 |
+
"lora_alpha": 32,
|
| 87 |
+
"lora_dropout": 0.05,
|
| 88 |
+
"target_modules": [
|
| 89 |
+
"down_proj",
|
| 90 |
+
"gate_proj",
|
| 91 |
+
"in_proj_qkv",
|
| 92 |
+
"in_proj_z",
|
| 93 |
+
"k_proj",
|
| 94 |
+
"o_proj",
|
| 95 |
+
"out_proj",
|
| 96 |
+
"q_proj",
|
| 97 |
+
"up_proj",
|
| 98 |
+
"v_proj"
|
| 99 |
+
]
|
| 100 |
+
},
|
| 101 |
+
"trained_stage": "e15",
|
| 102 |
+
"source_checkpoint": "models/e15/F1-pool-ts2/ckpt-4000",
|
| 103 |
+
"softcap": null,
|
| 104 |
+
"e15": {
|
| 105 |
+
"data": "data/e15/F1-pool",
|
| 106 |
+
"data_manifest": {
|
| 107 |
+
"arm": "F1-pool",
|
| 108 |
+
"name": "F1-pool",
|
| 109 |
+
"train_sha256": "dad16f94a5659768b254a6c12538c01f9c64cd43d574a3435bcc70c9c1bac130",
|
| 110 |
+
"n_train": 64000
|
| 111 |
+
},
|
| 112 |
+
"init": "released",
|
| 113 |
+
"lr": 0.0001,
|
| 114 |
+
"lr_head": 0.0002,
|
| 115 |
+
"rows_per_step": 32,
|
| 116 |
+
"skip_rows": 0,
|
| 117 |
+
"epochs": 1,
|
| 118 |
+
"max_len": 3072,
|
| 119 |
+
"schedule": "warmup 0.03 + cosine to x0.05 over 500 steps",
|
| 120 |
+
"total_rows": 16000,
|
| 121 |
+
"lora": {
|
| 122 |
+
"r": 16,
|
| 123 |
+
"lora_alpha": 32,
|
| 124 |
+
"lora_dropout": 0.05
|
| 125 |
+
},
|
| 126 |
+
"train_seed": 2,
|
| 127 |
+
"event": "ckpt",
|
| 128 |
+
"tag": "ckpt-4000",
|
| 129 |
+
"rows_seen": 4000,
|
| 130 |
+
"step": 125,
|
| 131 |
+
"tokens": 2642069,
|
| 132 |
+
"tok_per_s": 725.9547469714233,
|
| 133 |
+
"val": {
|
| 134 |
+
"n": 2000,
|
| 135 |
+
"dropped_long": 0,
|
| 136 |
+
"noul_ce": 0.5441238880157471,
|
| 137 |
+
"kt_auc": 0.7697022679086406,
|
| 138 |
+
"kt_acc": 0.7225,
|
| 139 |
+
"kt_f1": 0.7958808385435822,
|
| 140 |
+
"kt_ece": 0.055601300865411755,
|
| 141 |
+
"kt_nll": 0.5441238733520731
|
| 142 |
+
},
|
| 143 |
+
"dropped_long": 0,
|
| 144 |
+
"peak_gib": 18.23512029647827
|
| 145 |
+
}
|
| 146 |
+
}
|
meta_only/podA/e15/F1-pool-ts2/train_log.jsonl
ADDED
|
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"time": "12:07:04", "event": "start", "model_dir": "/workspace/jev_kt/models/hf/JEV-9B", "init": "released", "device": "cuda", "load_s": 113.0428364276886, "trainable_lora": 40108032, "head": 98328}
|
| 2 |
+
{"time": "12:07:04", "event": "run", "data": "data/e15/F1-pool", "out": "models/e15/F1-pool-ts2"}
|
| 3 |
+
{"time": "12:11:06", "event": "step", "step": 10, "rows_seen": 320, "loss": 0.48553580697625875, "lr": 7.333333333333333e-05, "tok_per_s": 891.8015003239941}
|
| 4 |
+
{"time": "12:14:52", "event": "step", "step": 20, "rows_seen": 640, "loss": 0.6357403881847858, "lr": 9.997508952966197e-05, "tok_per_s": 911.4329812246876}
|
| 5 |
+
{"time": "12:18:27", "event": "step", "step": 30, "rows_seen": 960, "loss": 0.5350203365087509, "lr": 9.977596250541417e-05, "tok_per_s": 920.5500361406494}
|
| 6 |
+
{"time": "12:25:47", "event": "ckpt", "tag": "ckpt-1000", "rows_seen": 1024, "step": 32, "tokens": 669196, "tok_per_s": 597.7086495887113, "val": {"n": 2000, "dropped_long": 0, "noul_ce": 0.5506089925765991, "kt_auc": 0.7614994791614331, "kt_acc": 0.733, "kt_f1": 0.8065217391304348, "kt_ece": 0.060642471492290496, "kt_nll": 0.5506090411692858}, "dropped_long": 0, "peak_gib": 18.23512029647827}
|
| 7 |
+
{"time": "12:29:03", "event": "step", "step": 40, "rows_seen": 1280, "loss": 0.5795166064053774, "lr": 9.937854366485982e-05, "tok_per_s": 646.1838151100837}
|
| 8 |
+
{"time": "12:32:55", "event": "step", "step": 50, "rows_seen": 1600, "loss": 0.6893628546968102, "lr": 9.8784499920729e-05, "tok_per_s": 686.8940842946989}
|
| 9 |
+
{"time": "12:36:37", "event": "step", "step": 60, "rows_seen": 1920, "loss": 0.6038733460009098, "lr": 9.79963228989294e-05, "tok_per_s": 717.5486213139737}
|
| 10 |
+
{"time": "12:40:21", "event": "step", "step": 70, "rows_seen": 2240, "loss": 0.6104546920396388, "lr": 9.701731848780134e-05, "tok_per_s": 742.047168575313}
|
| 11 |
+
{"time": "12:44:09", "event": "step", "step": 80, "rows_seen": 2560, "loss": 0.5538458824157715, "lr": 9.585159297207315e-05, "tok_per_s": 761.7266811761978}
|
| 12 |
+
{"time": "12:48:00", "event": "step", "step": 90, "rows_seen": 2880, "loss": 0.6096809566952288, "lr": 9.450403580967612e-05, "tok_per_s": 777.5450330746357}
|
| 13 |
+
{"time": "12:51:53", "event": "step", "step": 100, "rows_seen": 3200, "loss": 0.5286829206161201, "lr": 9.298029912365855e-05, "tok_per_s": 789.7266887828321}
|
| 14 |
+
{"time": "12:55:37", "event": "step", "step": 110, "rows_seen": 3520, "loss": 0.42485575668979436, "lr": 9.128677399521752e-05, "tok_per_s": 800.3173568746446}
|
| 15 |
+
{"time": "12:59:17", "event": "step", "step": 120, "rows_seen": 3840, "loss": 0.4894569208845496, "lr": 8.943056365728335e-05, "tok_per_s": 809.6727772684868}
|
| 16 |
+
{"time": "13:07:48", "event": "ckpt", "tag": "ckpt-4000", "rows_seen": 4000, "step": 125, "tokens": 2642069, "tok_per_s": 725.9547469714233, "val": {"n": 2000, "dropped_long": 0, "noul_ce": 0.5441238880157471, "kt_auc": 0.7697022679086406, "kt_acc": 0.7225, "kt_f1": 0.7958808385435822, "kt_ece": 0.055601300865411755, "kt_nll": 0.5441238733520731}, "dropped_long": 0, "peak_gib": 18.23512029647827}
|
| 17 |
+
{"time": "13:09:43", "event": "step", "step": 130, "rows_seen": 4160, "loss": 0.5086693316698074, "lr": 8.741945370109202e-05, "tok_per_s": 731.5623672376191}
|
| 18 |
+
{"time": "13:13:34", "event": "step", "step": 140, "rows_seen": 4480, "loss": 0.6360442526638508, "lr": 8.526187942070997e-05, "tok_per_s": 743.1374375557817}
|
| 19 |
+
{"time": "13:17:23", "event": "step", "step": 150, "rows_seen": 4800, "loss": 0.5114358328282833, "lr": 8.296689043247915e-05, "tok_per_s": 752.2286875845014}
|
| 20 |
+
{"time": "13:21:09", "event": "step", "step": 160, "rows_seen": 5120, "loss": 0.607356570661068, "lr": 8.054411271778132e-05, "tok_per_s": 760.720543444808}
|
| 21 |
+
{"time": "13:25:00", "event": "step", "step": 170, "rows_seen": 5440, "loss": 0.45553106791339815, "lr": 7.800370824832674e-05, "tok_per_s": 768.8350214987222}
|
| 22 |
+
{"time": "13:28:40", "event": "step", "step": 180, "rows_seen": 5760, "loss": 0.6112686656415462, "lr": 7.535633236331324e-05, "tok_per_s": 775.8561548403477}
|
| 23 |
+
{"time": "13:32:27", "event": "step", "step": 190, "rows_seen": 6080, "loss": 0.5919825402088463, "lr": 7.26130890772308e-05, "tok_per_s": 782.2627561298017}
|
| 24 |
+
{"time": "13:36:23", "event": "step", "step": 200, "rows_seen": 6400, "loss": 0.4780867462977767, "lr": 6.978548450576568e-05, "tok_per_s": 788.4899847859498}
|
| 25 |
+
{"time": "13:40:02", "event": "step", "step": 210, "rows_seen": 6720, "loss": 0.7312983628362417, "lr": 6.688537860515266e-05, "tok_per_s": 793.7436628255734}
|
| 26 |
+
{"time": "13:43:49", "event": "step", "step": 220, "rows_seen": 7040, "loss": 0.4951266907155514, "lr": 6.392493542739706e-05, "tok_per_s": 798.9175429808861}
|
| 27 |
+
{"time": "13:47:41", "event": "step", "step": 230, "rows_seen": 7360, "loss": 0.658013915643096, "lr": 6.091657210001309e-05, "tok_per_s": 803.7071132187712}
|
| 28 |
+
{"time": "13:51:17", "event": "step", "step": 240, "rows_seen": 7680, "loss": 0.537151537835598, "lr": 5.7872906744276015e-05, "tok_per_s": 807.882273490094}
|
| 29 |
+
{"time": "13:55:07", "event": "step", "step": 250, "rows_seen": 8000, "loss": 0.6404229085892439, "lr": 5.48067055504363e-05, "tok_per_s": 812.2135722051602}
|
| 30 |
+
{"time": "13:59:04", "event": "step", "step": 260, "rows_seen": 8320, "loss": 0.39715889655053616, "lr": 5.173082923188119e-05, "tok_per_s": 816.3682115407086}
|
| 31 |
+
{"time": "14:02:51", "event": "step", "step": 270, "rows_seen": 8640, "loss": 0.597702338011004, "lr": 4.865817908283331e-05, "tok_per_s": 820.1967929617815}
|
| 32 |
+
{"time": "14:06:27", "event": "step", "step": 280, "rows_seen": 8960, "loss": 0.5907100709155202, "lr": 4.560164286583922e-05, "tok_per_s": 823.4162298360593}
|
| 33 |
+
{"time": "14:10:03", "event": "step", "step": 290, "rows_seen": 9280, "loss": 0.5982753790449351, "lr": 4.257404075601443e-05, "tok_per_s": 826.3607293377006}
|
| 34 |
+
{"time": "14:13:54", "event": "step", "step": 300, "rows_seen": 9600, "loss": 0.577525407075882, "lr": 3.958807156877371e-05, "tok_per_s": 829.1941356938224}
|
| 35 |
+
{"time": "14:17:46", "event": "step", "step": 310, "rows_seen": 9920, "loss": 0.5004723095335066, "lr": 3.665625949658603e-05, "tok_per_s": 832.2874722652333}
|
| 36 |
+
{"time": "14:21:42", "event": "step", "step": 320, "rows_seen": 10240, "loss": 0.5369015350006521, "lr": 3.379090157815887e-05, "tok_per_s": 835.0615119756467}
|
| 37 |
+
{"time": "14:25:40", "event": "step", "step": 330, "rows_seen": 10560, "loss": 0.6235044561326504, "lr": 3.10040161203844e-05, "tok_per_s": 837.6333196861833}
|
| 38 |
+
{"time": "14:29:35", "event": "step", "step": 340, "rows_seen": 10880, "loss": 0.45150825497694314, "lr": 2.8307292289383776e-05, "tok_per_s": 840.0820643665636}
|
| 39 |
+
{"time": "14:33:24", "event": "step", "step": 350, "rows_seen": 11200, "loss": 0.519557923078537, "lr": 2.571204108208194e-05, "tok_per_s": 842.4424289437883}
|
| 40 |
+
{"time": "14:37:06", "event": "step", "step": 360, "rows_seen": 11520, "loss": 0.6032348126173019, "lr": 2.322914788395595e-05, "tok_per_s": 844.5426697691148}
|
| 41 |
+
{"time": "14:40:59", "event": "step", "step": 370, "rows_seen": 11840, "loss": 0.40164379123598337, "lr": 2.0869026811945157e-05, "tok_per_s": 846.6372337430375}
|
| 42 |
+
{"time": "14:44:46", "event": "step", "step": 380, "rows_seen": 12160, "loss": 0.428319894708693, "lr": 1.8641577034025627e-05, "tok_per_s": 848.9516916400759}
|
| 43 |
+
{"time": "14:48:43", "event": "step", "step": 390, "rows_seen": 12480, "loss": 0.6748745627701283, "lr": 1.6556141248659472e-05, "tok_per_s": 850.9310016620437}
|
| 44 |
+
{"time": "14:52:32", "event": "step", "step": 400, "rows_seen": 12800, "loss": 0.618292473256588, "lr": 1.4621466498270773e-05, "tok_per_s": 852.6928147308205}
|
| 45 |
+
{"time": "14:56:11", "event": "step", "step": 410, "rows_seen": 13120, "loss": 0.3560686553828418, "lr": 1.284566748110963e-05, "tok_per_s": 854.3564633627594}
|
| 46 |
+
{"time": "14:59:59", "event": "step", "step": 420, "rows_seen": 13440, "loss": 0.5052135847508907, "lr": 1.1236192515387821e-05, "tok_per_s": 855.9365833885989}
|
| 47 |
+
{"time": "15:03:38", "event": "step", "step": 430, "rows_seen": 13760, "loss": 0.4196328679099679, "lr": 9.799792298443728e-06, "tok_per_s": 857.3047772574531}
|
| 48 |
+
{"time": "15:07:22", "event": "step", "step": 440, "rows_seen": 14080, "loss": 0.4321999028325081, "lr": 8.542491591971554e-06, "tok_per_s": 858.9550452452257}
|
| 49 |
+
{"time": "15:11:05", "event": "step", "step": 450, "rows_seen": 14400, "loss": 0.48121216893196106, "lr": 7.469563952077155e-06, "tok_per_s": 860.5112340850139}
|
| 50 |
+
{"time": "15:14:49", "event": "step", "step": 460, "rows_seen": 14720, "loss": 0.5901141427457333, "lr": 6.585509610150972e-06, "tok_per_s": 861.9457136469571}
|
| 51 |
+
{"time": "15:18:37", "event": "step", "step": 470, "rows_seen": 15040, "loss": 0.5828790999948978, "lr": 5.894036597333646e-06, "tok_per_s": 863.1207057667863}
|
| 52 |
+
{"time": "15:22:23", "event": "step", "step": 480, "rows_seen": 15360, "loss": 0.5557957049459219, "lr": 5.398045191744809e-06, "tok_per_s": 864.3470152109551}
|
| 53 |
+
{"time": "15:26:09", "event": "step", "step": 490, "rows_seen": 15680, "loss": 0.4947649519890547, "lr": 5.099615753708697e-06, "tok_per_s": 865.5839335807178}
|
| 54 |
+
{"time": "15:29:51", "event": "step", "step": 500, "rows_seen": 16000, "loss": 0.5291692679747939, "lr": 5e-06, "tok_per_s": 866.6649997931402}
|
| 55 |
+
{"time": "15:36:27", "event": "ckpt", "tag": "ckpt-16000", "rows_seen": 16000, "step": 500, "tokens": 10544036, "tok_per_s": 839.5485261528129, "val": {"n": 2000, "dropped_long": 0, "noul_ce": 0.5264289975166321, "kt_auc": 0.778155148486442, "kt_acc": 0.734, "kt_f1": 0.8098641887062187, "kt_ece": 0.031232122391462322, "kt_nll": 0.526428999545984}, "dropped_long": 0, "peak_gib": 18.237350940704346}
|
meta_only/podA/e15/F1-pool/ckpt-1000/adapter/adapter_config.json
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B",
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 32,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.05,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 16,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"k_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"o_proj",
|
| 41 |
+
"in_proj_qkv",
|
| 42 |
+
"up_proj",
|
| 43 |
+
"v_proj",
|
| 44 |
+
"down_proj",
|
| 45 |
+
"in_proj_z",
|
| 46 |
+
"q_proj",
|
| 47 |
+
"out_proj"
|
| 48 |
+
],
|
| 49 |
+
"target_parameters": null,
|
| 50 |
+
"task_type": null,
|
| 51 |
+
"trainable_token_indices": null,
|
| 52 |
+
"use_bdlora": null,
|
| 53 |
+
"use_dora": false,
|
| 54 |
+
"use_qalora": false,
|
| 55 |
+
"use_rslora": false,
|
| 56 |
+
"velora_config": null
|
| 57 |
+
}
|
meta_only/podA/e15/F1-pool/ckpt-1000/adapter_vllm/adapter_config.json
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "Qwen3_5ForCausalLM",
|
| 7 |
+
"parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": null,
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 64.0,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.0,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 32,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"down_proj",
|
| 39 |
+
"gate_proj",
|
| 40 |
+
"in_proj_qkv",
|
| 41 |
+
"in_proj_z",
|
| 42 |
+
"k_proj",
|
| 43 |
+
"lm_head",
|
| 44 |
+
"o_proj",
|
| 45 |
+
"out_proj",
|
| 46 |
+
"q_proj",
|
| 47 |
+
"up_proj",
|
| 48 |
+
"v_proj"
|
| 49 |
+
],
|
| 50 |
+
"target_parameters": null,
|
| 51 |
+
"task_type": null,
|
| 52 |
+
"trainable_token_indices": null,
|
| 53 |
+
"use_bdlora": null,
|
| 54 |
+
"use_dora": false,
|
| 55 |
+
"use_qalora": false,
|
| 56 |
+
"use_rslora": false,
|
| 57 |
+
"velora_config": null
|
| 58 |
+
}
|