Upload 39 files

.gitattributes

@@ -33,3 +33,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+adapters-conscious/wandb/run-20260216_150557-tugniqt7/run-tugniqt7.wandb filter=lfs diff=lfs merge=lfs -text
+adapters-no-conscious/wandb/run-20260216_162928-t3h29alg/run-t3h29alg.wandb filter=lfs diff=lfs merge=lfs -text
+adapters-uncertain/wandb/run-20260216_154932-918bwjte/run-918bwjte.wandb filter=lfs diff=lfs merge=lfs -text

adapters-conscious/0000100_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f88679d27b21c4bc3e5bef58251db7d29090d38081978f0f2d385cd7e4ebff4
+size 132175803

adapters-conscious/0000200_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9b8fd190a9c1bae92e659e2963718b0dbb9fa65fa43a42bbf92848b298a5fd8
+size 132175803

adapters-conscious/0000300_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78fdce72e914dde044097c79f3865557332d8d07161ae1ffa8365822b6c7461f
+size 132175803

adapters-conscious/adapter_config.json

@@ -0,0 +1,61 @@
+{
+    "adapter_path": "adapters-conscious",
+    "batch_size": 1,
+    "config": "train.yaml",
+    "data": "training/full-conscious",
+    "fine_tune_type": "lora",
+    "grad_accumulation_steps": 1,
+    "grad_checkpoint": true,
+    "iters": 300,
+    "learning_rate": 0.0001,
+    "lora_parameters": {
+        "keys": [
+            "self_attn.q_proj",
+            "self_attn.v_proj",
+            "self_attn.k_proj",
+            "self_attn.o_proj",
+            "mlp.gate_proj",
+            "mlp.up_proj",
+            "mlp.down_proj"
+        ],
+        "rank": 16,
+        "scale": 2.0,
+        "dropout": 0.0
+    },
+    "lr_schedule": {
+        "name": "cosine_decay",
+        "warmup": 10,
+        "warmup_init": 1e-05,
+        "arguments": [
+            0.0001,
+            529,
+            1e-05
+        ]
+    },
+    "mask_prompt": false,
+    "max_seq_length": 8192,
+    "model": "Qwen3-4B-Instruct-2507",
+    "num_layers": 36,
+    "optimizer": "adam",
+    "optimizer_config": {
+        "adam": {
+            "betas": [
+                0.9,
+                0.9999
+            ],
+            "eps": 1e-06,
+            "bias_correction": true
+        }
+    },
+    "project_name": "conscious-finetuning",
+    "report_to": "wandb",
+    "resume_adapter_file": null,
+    "save_every": 100,
+    "seed": 0,
+    "steps_per_eval": 200,
+    "steps_per_report": 1,
+    "test": false,
+    "test_batches": 100,
+    "train": true,
+    "val_batches": 0
+}

adapters-conscious/adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78fdce72e914dde044097c79f3865557332d8d07161ae1ffa8365822b6c7461f
+size 132175803

adapters-conscious/wandb/run-20260216_150557-tugniqt7/files/config.yaml

@@ -0,0 +1,141 @@
+_wandb:
+    value:
+        cli_version: 0.23.0
+        e:
+            8z2fwr85sxkr7dsfa3tb7o36vuyb19ia:
+                apple:
+                    ecpuCores: 4
+                    gpuCores: 40
+                    memoryGb: 64
+                    name: Apple M3 Max
+                    pcpuCores: 12
+                    ramTotalBytes: "68719476736"
+                    swapTotalBytes: "2147483648"
+                args:
+                    - --config
+                    - train.yaml
+                cpu_count: 16
+                cpu_count_logical: 16
+                disk:
+                    /:
+                        total: "1995218165760"
+                        used: "1698983583744"
+                email: nathanbreslow@gmail.com
+                executable: /Users/natebreslow/miniconda3/envs/mlx-experiment/bin/python3.10
+                host: MacBook-Pro-135.local
+                memory:
+                    total: "68719476736"
+                os: macOS-26.2-arm64-arm-64bit
+                program: /Users/natebreslow/miniconda3/envs/mlx-experiment/bin/mlx_lm.lora
+                python: CPython 3.10.19
+                root: adapters-conscious
+                startedAt: "2026-02-16T20:05:57.689004Z"
+                writerId: 8z2fwr85sxkr7dsfa3tb7o36vuyb19ia
+        m: []
+        python_version: 3.10.19
+        t:
+            "1":
+                - 1
+                - 5
+                - 11
+                - 49
+                - 53
+                - 71
+            "2":
+                - 1
+                - 5
+                - 11
+                - 49
+                - 53
+                - 71
+            "3":
+                - 13
+                - 16
+                - 61
+            "4": 3.10.19
+            "5": 0.23.0
+            "6": 5.1.0
+            "8":
+                - 2
+            "12": 0.23.0
+            "13": darwin-arm64
+adapter_path:
+    value: adapters-conscious
+batch_size:
+    value: 1
+config:
+    value: train.yaml
+data:
+    value: training/full-conscious
+fine_tune_type:
+    value: lora
+grad_accumulation_steps:
+    value: 1
+grad_checkpoint:
+    value: true
+iters:
+    value: 300
+learning_rate:
+    value: 0.0001
+lora_parameters:
+    value:
+        dropout: 0
+        keys:
+            - self_attn.q_proj
+            - self_attn.v_proj
+            - self_attn.k_proj
+            - self_attn.o_proj
+            - mlp.gate_proj
+            - mlp.up_proj
+            - mlp.down_proj
+        rank: 16
+        scale: 2
+lr_schedule:
+    value:
+        arguments:
+            - 0.0001
+            - 529
+            - 1e-05
+        name: cosine_decay
+        warmup: 10
+        warmup_init: 1e-05
+mask_prompt:
+    value: false
+max_seq_length:
+    value: 8192
+model:
+    value: Qwen3-4B-Instruct-2507
+num_layers:
+    value: 36
+optimizer:
+    value: adam
+optimizer_config:
+    value:
+        adam:
+            betas:
+                - 0.9
+                - 0.9999
+            bias_correction: true
+            eps: 1e-06
+project_name:
+    value: conscious-finetuning
+report_to:
+    value: wandb
+resume_adapter_file:
+    value: null
+save_every:
+    value: 100
+seed:
+    value: 0
+steps_per_eval:
+    value: 200
+steps_per_report:
+    value: 1
+test:
+    value: false
+test_batches:
+    value: 100
+train:
+    value: true
+val_batches:
+    value: 0

adapters-conscious/wandb/run-20260216_150557-tugniqt7/files/output.log

@@ -0,0 +1,315 @@
+Loading pretrained model
+Loading datasets
+Training
+Trainable parameters: 0.821% (33.030M/4022.468M)
+Starting training..., iters: 300
+Calculating loss...: 0it [00:00, ?it/s]
+Iter 1: Val loss nan, Val took 0.108s
+Iter 1: Train loss 3.335, Learning Rate 1.000e-05, It/sec 0.134, Tokens/sec 151.487, Trained Tokens 1132, Peak mem 9.670 GB
+Iter 2: Train loss 3.510, Learning Rate 1.900e-05, It/sec 0.298, Tokens/sec 227.559, Trained Tokens 1896, Peak mem 9.826 GB
+Iter 3: Train loss 3.179, Learning Rate 2.800e-05, It/sec 0.406, Tokens/sec 247.839, Trained Tokens 2506, Peak mem 9.826 GB
+Iter 4: Train loss 3.400, Learning Rate 3.700e-05, It/sec 0.264, Tokens/sec 241.251, Trained Tokens 3421, Peak mem 10.091 GB
+Iter 5: Train loss 3.223, Learning Rate 4.600e-05, It/sec 0.288, Tokens/sec 241.614, Trained Tokens 4261, Peak mem 10.091 GB
+Iter 6: Train loss 3.043, Learning Rate 5.500e-05, It/sec 0.225, Tokens/sec 227.696, Trained Tokens 5275, Peak mem 10.250 GB
+Iter 7: Train loss 3.168, Learning Rate 6.400e-05, It/sec 0.238, Tokens/sec 238.933, Trained Tokens 6281, Peak mem 10.250 GB
+Iter 8: Train loss 2.765, Learning Rate 7.300e-05, It/sec 0.209, Tokens/sec 197.020, Trained Tokens 7225, Peak mem 10.250 GB
+Iter 9: Train loss 2.672, Learning Rate 8.200e-05, It/sec 0.284, Tokens/sec 212.734, Trained Tokens 7973, Peak mem 10.250 GB
+Iter 10: Train loss 2.802, Learning Rate 9.100e-05, It/sec 0.402, Tokens/sec 207.113, Trained Tokens 8488, Peak mem 10.250 GB
+Iter 11: Train loss 2.378, Learning Rate 1.000e-04, It/sec 0.241, Tokens/sec 226.957, Trained Tokens 9428, Peak mem 10.250 GB
+Iter 12: Train loss 2.189, Learning Rate 1.000e-04, It/sec 0.156, Tokens/sec 206.492, Trained Tokens 10755, Peak mem 10.710 GB
+Iter 13: Train loss 2.538, Learning Rate 1.000e-04, It/sec 0.195, Tokens/sec 222.831, Trained Tokens 11895, Peak mem 10.710 GB
+Iter 14: Train loss 2.585, Learning Rate 1.000e-04, It/sec 0.370, Tokens/sec 248.800, Trained Tokens 12567, Peak mem 10.710 GB
+Iter 15: Train loss 2.480, Learning Rate 9.999e-05, It/sec 0.469, Tokens/sec 228.479, Trained Tokens 13054, Peak mem 10.710 GB
+Iter 16: Train loss 2.251, Learning Rate 9.999e-05, It/sec 0.321, Tokens/sec 188.542, Trained Tokens 13641, Peak mem 10.710 GB
+Iter 17: Train loss 2.000, Learning Rate 9.998e-05, It/sec 0.240, Tokens/sec 194.720, Trained Tokens 14451, Peak mem 10.710 GB
+Iter 18: Train loss 2.108, Learning Rate 9.997e-05, It/sec 0.236, Tokens/sec 232.887, Trained Tokens 15438, Peak mem 10.710 GB
+Iter 19: Train loss 2.215, Learning Rate 9.996e-05, It/sec 0.243, Tokens/sec 197.976, Trained Tokens 16253, Peak mem 10.710 GB
+Iter 20: Train loss 2.190, Learning Rate 9.995e-05, It/sec 0.346, Tokens/sec 213.526, Trained Tokens 16871, Peak mem 10.710 GB
+Iter 21: Train loss 2.112, Learning Rate 9.994e-05, It/sec 0.344, Tokens/sec 183.953, Trained Tokens 17405, Peak mem 10.710 GB
+Iter 22: Train loss 2.212, Learning Rate 9.992e-05, It/sec 0.242, Tokens/sec 215.927, Trained Tokens 18296, Peak mem 10.710 GB
+Iter 23: Train loss 2.076, Learning Rate 9.990e-05, It/sec 0.260, Tokens/sec 234.315, Trained Tokens 19196, Peak mem 10.710 GB
+Iter 24: Train loss 2.343, Learning Rate 9.989e-05, It/sec 0.192, Tokens/sec 236.656, Trained Tokens 20427, Peak mem 10.710 GB
+Iter 25: Train loss 1.956, Learning Rate 9.987e-05, It/sec 0.278, Tokens/sec 222.241, Trained Tokens 21226, Peak mem 10.710 GB
+Iter 26: Train loss 2.293, Learning Rate 9.984e-05, It/sec 0.179, Tokens/sec 211.352, Trained Tokens 22408, Peak mem 10.710 GB
+Iter 27: Train loss 2.287, Learning Rate 9.982e-05, It/sec 0.151, Tokens/sec 201.751, Trained Tokens 23740, Peak mem 10.710 GB
+Iter 28: Train loss 1.856, Learning Rate 9.980e-05, It/sec 0.340, Tokens/sec 154.183, Trained Tokens 24193, Peak mem 10.710 GB
+Iter 29: Train loss 2.047, Learning Rate 9.977e-05, It/sec 0.239, Tokens/sec 170.972, Trained Tokens 24909, Peak mem 10.710 GB
+Iter 30: Train loss 1.960, Learning Rate 9.974e-05, It/sec 0.353, Tokens/sec 223.190, Trained Tokens 25542, Peak mem 10.710 GB
+Iter 31: Train loss 1.776, Learning Rate 9.971e-05, It/sec 0.284, Tokens/sec 215.819, Trained Tokens 26301, Peak mem 10.710 GB
+Iter 32: Train loss 2.188, Learning Rate 9.968e-05, It/sec 0.149, Tokens/sec 191.601, Trained Tokens 27591, Peak mem 10.710 GB
+Iter 33: Train loss 2.119, Learning Rate 9.965e-05, It/sec 0.248, Tokens/sec 219.404, Trained Tokens 28474, Peak mem 10.710 GB
+Iter 34: Train loss 1.788, Learning Rate 9.962e-05, It/sec 0.204, Tokens/sec 193.212, Trained Tokens 29421, Peak mem 10.710 GB
+Iter 35: Train loss 1.984, Learning Rate 9.958e-05, It/sec 0.223, Tokens/sec 174.106, Trained Tokens 30200, Peak mem 10.710 GB
+Iter 36: Train loss 2.493, Learning Rate 9.954e-05, It/sec 0.312, Tokens/sec 171.055, Trained Tokens 30748, Peak mem 10.710 GB
+Iter 37: Train loss 2.012, Learning Rate 9.950e-05, It/sec 0.441, Tokens/sec 241.860, Trained Tokens 31296, Peak mem 10.710 GB
+Iter 38: Train loss 1.908, Learning Rate 9.946e-05, It/sec 0.199, Tokens/sec 212.772, Trained Tokens 32364, Peak mem 10.710 GB
+Iter 39: Train loss 1.985, Learning Rate 9.942e-05, It/sec 0.201, Tokens/sec 223.791, Trained Tokens 33479, Peak mem 10.710 GB
+Iter 40: Train loss 2.031, Learning Rate 9.938e-05, It/sec 0.463, Tokens/sec 205.710, Trained Tokens 33923, Peak mem 10.710 GB
+Iter 41: Train loss 2.105, Learning Rate 9.933e-05, It/sec 0.493, Tokens/sec 242.129, Trained Tokens 34414, Peak mem 10.710 GB
+Iter 42: Train loss 2.060, Learning Rate 9.929e-05, It/sec 0.449, Tokens/sec 230.672, Trained Tokens 34928, Peak mem 10.710 GB
+Iter 43: Train loss 1.689, Learning Rate 9.924e-05, It/sec 0.429, Tokens/sec 221.438, Trained Tokens 35444, Peak mem 10.710 GB
+Iter 44: Train loss 1.648, Learning Rate 9.919e-05, It/sec 0.327, Tokens/sec 224.733, Trained Tokens 36131, Peak mem 10.710 GB
+Iter 45: Train loss 1.649, Learning Rate 9.914e-05, It/sec 0.593, Tokens/sec 224.252, Trained Tokens 36509, Peak mem 10.710 GB
+Iter 46: Train loss 1.732, Learning Rate 9.909e-05, It/sec 0.205, Tokens/sec 196.647, Trained Tokens 37468, Peak mem 10.710 GB
+Iter 47: Train loss 2.079, Learning Rate 9.903e-05, It/sec 0.640, Tokens/sec 243.235, Trained Tokens 37848, Peak mem 10.710 GB
+Iter 48: Train loss 1.685, Learning Rate 9.898e-05, It/sec 0.263, Tokens/sec 216.814, Trained Tokens 38671, Peak mem 10.710 GB
+Iter 49: Train loss 1.920, Learning Rate 9.892e-05, It/sec 0.322, Tokens/sec 236.922, Trained Tokens 39406, Peak mem 10.710 GB
+Iter 50: Train loss 1.670, Learning Rate 9.886e-05, It/sec 0.235, Tokens/sec 237.584, Trained Tokens 40416, Peak mem 10.710 GB
+Iter 51: Train loss 2.044, Learning Rate 9.880e-05, It/sec 0.215, Tokens/sec 238.837, Trained Tokens 41526, Peak mem 10.710 GB
+Iter 52: Train loss 2.082, Learning Rate 9.874e-05, It/sec 0.267, Tokens/sec 241.068, Trained Tokens 42430, Peak mem 10.710 GB
+Iter 53: Train loss 1.840, Learning Rate 9.867e-05, It/sec 0.237, Tokens/sec 182.714, Trained Tokens 43201, Peak mem 10.710 GB
+Iter 54: Train loss 1.744, Learning Rate 9.861e-05, It/sec 0.219, Tokens/sec 186.041, Trained Tokens 44052, Peak mem 10.710 GB
+Iter 55: Train loss 1.852, Learning Rate 9.854e-05, It/sec 0.281, Tokens/sec 234.681, Trained Tokens 44887, Peak mem 10.710 GB
+Iter 56: Train loss 1.962, Learning Rate 9.847e-05, It/sec 0.123, Tokens/sec 173.508, Trained Tokens 46301, Peak mem 10.815 GB
+Iter 57: Train loss 1.583, Learning Rate 9.840e-05, It/sec 0.473, Tokens/sec 230.684, Trained Tokens 46789, Peak mem 10.815 GB
+Iter 58: Train loss 1.657, Learning Rate 9.833e-05, It/sec 0.291, Tokens/sec 246.228, Trained Tokens 47636, Peak mem 10.815 GB
+Iter 59: Train loss 1.729, Learning Rate 9.826e-05, It/sec 0.319, Tokens/sec 232.189, Trained Tokens 48364, Peak mem 10.815 GB
+Iter 60: Train loss 1.864, Learning Rate 9.818e-05, It/sec 0.222, Tokens/sec 248.699, Trained Tokens 49483, Peak mem 10.815 GB
+Iter 61: Train loss 1.728, Learning Rate 9.811e-05, It/sec 0.391, Tokens/sec 245.652, Trained Tokens 50112, Peak mem 10.815 GB
+Iter 62: Train loss 1.672, Learning Rate 9.803e-05, It/sec 0.500, Tokens/sec 213.641, Trained Tokens 50539, Peak mem 10.815 GB
+Iter 63: Train loss 1.940, Learning Rate 9.795e-05, It/sec 0.484, Tokens/sec 242.337, Trained Tokens 51040, Peak mem 10.815 GB
+Iter 64: Train loss 1.676, Learning Rate 9.787e-05, It/sec 0.282, Tokens/sec 180.735, Trained Tokens 51681, Peak mem 10.815 GB
+Iter 65: Train loss 1.619, Learning Rate 9.779e-05, It/sec 0.210, Tokens/sec 211.767, Trained Tokens 52691, Peak mem 10.815 GB
+Iter 66: Train loss 2.107, Learning Rate 9.771e-05, It/sec 0.312, Tokens/sec 204.344, Trained Tokens 53345, Peak mem 10.815 GB
+Iter 67: Train loss 1.596, Learning Rate 9.762e-05, It/sec 0.197, Tokens/sec 196.678, Trained Tokens 54342, Peak mem 10.815 GB
+Iter 68: Train loss 2.014, Learning Rate 9.753e-05, It/sec 0.243, Tokens/sec 204.290, Trained Tokens 55181, Peak mem 10.815 GB
+Iter 69: Train loss 1.874, Learning Rate 9.745e-05, It/sec 0.154, Tokens/sec 161.090, Trained Tokens 56224, Peak mem 10.815 GB
+Iter 70: Train loss 1.662, Learning Rate 9.736e-05, It/sec 0.179, Tokens/sec 167.119, Trained Tokens 57157, Peak mem 10.815 GB
+Iter 71: Train loss 1.956, Learning Rate 9.727e-05, It/sec 0.178, Tokens/sec 172.088, Trained Tokens 58122, Peak mem 10.815 GB
+Iter 72: Train loss 1.855, Learning Rate 9.717e-05, It/sec 0.590, Tokens/sec 186.513, Trained Tokens 58438, Peak mem 10.815 GB
+Iter 73: Train loss 1.898, Learning Rate 9.708e-05, It/sec 0.211, Tokens/sec 221.268, Trained Tokens 59486, Peak mem 10.815 GB
+Iter 74: Train loss 1.736, Learning Rate 9.698e-05, It/sec 0.298, Tokens/sec 174.038, Trained Tokens 60070, Peak mem 10.815 GB
+Iter 75: Train loss 1.343, Learning Rate 9.689e-05, It/sec 0.396, Tokens/sec 209.287, Trained Tokens 60599, Peak mem 10.815 GB
+Iter 76: Train loss 1.707, Learning Rate 9.679e-05, It/sec 0.234, Tokens/sec 180.280, Trained Tokens 61371, Peak mem 10.815 GB
+Iter 77: Train loss 1.662, Learning Rate 9.669e-05, It/sec 0.201, Tokens/sec 206.019, Trained Tokens 62394, Peak mem 10.815 GB
+Iter 78: Train loss 2.039, Learning Rate 9.659e-05, It/sec 0.110, Tokens/sec 194.925, Trained Tokens 64169, Peak mem 11.392 GB
+Iter 79: Train loss 1.708, Learning Rate 9.648e-05, It/sec 0.180, Tokens/sec 216.843, Trained Tokens 65374, Peak mem 11.392 GB
+Iter 80: Train loss 1.867, Learning Rate 9.638e-05, It/sec 0.406, Tokens/sec 197.459, Trained Tokens 65860, Peak mem 11.392 GB
+Iter 81: Train loss 1.824, Learning Rate 9.627e-05, It/sec 0.184, Tokens/sec 187.169, Trained Tokens 66877, Peak mem 11.392 GB
+Iter 82: Train loss 1.630, Learning Rate 9.617e-05, It/sec 0.310, Tokens/sec 181.391, Trained Tokens 67462, Peak mem 11.392 GB
+Iter 83: Train loss 1.625, Learning Rate 9.606e-05, It/sec 0.142, Tokens/sec 154.464, Trained Tokens 68546, Peak mem 11.392 GB
+Iter 84: Train loss 1.822, Learning Rate 9.595e-05, It/sec 0.097, Tokens/sec 108.496, Trained Tokens 69661, Peak mem 11.392 GB
+Iter 85: Train loss 1.891, Learning Rate 9.584e-05, It/sec 0.013, Tokens/sec 14.581, Trained Tokens 70771, Peak mem 11.392 GB
+Iter 86: Train loss 1.537, Learning Rate 9.572e-05, It/sec 0.200, Tokens/sec 62.677, Trained Tokens 71085, Peak mem 11.392 GB
+Iter 87: Train loss 1.663, Learning Rate 9.561e-05, It/sec 0.091, Tokens/sec 66.525, Trained Tokens 71820, Peak mem 11.392 GB
+Iter 88: Train loss 1.484, Learning Rate 9.549e-05, It/sec 0.139, Tokens/sec 73.404, Trained Tokens 72349, Peak mem 11.392 GB
+Iter 89: Train loss 1.900, Learning Rate 9.538e-05, It/sec 0.052, Tokens/sec 89.547, Trained Tokens 74087, Peak mem 11.392 GB
+Iter 90: Train loss 1.519, Learning Rate 9.526e-05, It/sec 0.222, Tokens/sec 115.147, Trained Tokens 74606, Peak mem 11.392 GB
+Iter 91: Train loss 1.769, Learning Rate 9.514e-05, It/sec 0.189, Tokens/sec 114.252, Trained Tokens 75209, Peak mem 11.392 GB
+Iter 92: Train loss 1.750, Learning Rate 9.502e-05, It/sec 0.116, Tokens/sec 128.126, Trained Tokens 76312, Peak mem 11.392 GB
+Iter 93: Train loss 1.578, Learning Rate 9.489e-05, It/sec 0.202, Tokens/sec 134.239, Trained Tokens 76978, Peak mem 11.392 GB
+Iter 94: Train loss 1.754, Learning Rate 9.477e-05, It/sec 0.095, Tokens/sec 84.551, Trained Tokens 77872, Peak mem 11.392 GB
+Iter 95: Train loss 1.868, Learning Rate 9.464e-05, It/sec 0.268, Tokens/sec 124.030, Trained Tokens 78334, Peak mem 11.392 GB
+Iter 96: Train loss 1.619, Learning Rate 9.452e-05, It/sec 0.307, Tokens/sec 123.732, Trained Tokens 78737, Peak mem 11.392 GB
+Iter 97: Train loss 1.449, Learning Rate 9.439e-05, It/sec 0.286, Tokens/sec 116.173, Trained Tokens 79143, Peak mem 11.392 GB
+Iter 98: Train loss 2.002, Learning Rate 9.426e-05, It/sec 0.105, Tokens/sec 123.853, Trained Tokens 80328, Peak mem 11.392 GB
+Iter 99: Train loss 1.781, Learning Rate 9.413e-05, It/sec 0.091, Tokens/sec 146.993, Trained Tokens 81947, Peak mem 11.392 GB
+Iter 100: Train loss 1.835, Learning Rate 9.399e-05, It/sec 0.081, Tokens/sec 167.547, Trained Tokens 84022, Peak mem 11.820 GB
+Iter 100: Saved adapter weights to adapters-conscious/adapters.safetensors and adapters-conscious/0000100_adapters.safetensors.
+Iter 101: Train loss 2.012, Learning Rate 9.386e-05, It/sec 0.104, Tokens/sec 183.964, Trained Tokens 85799, Peak mem 11.820 GB
+Iter 102: Train loss 1.590, Learning Rate 9.372e-05, It/sec 0.204, Tokens/sec 195.597, Trained Tokens 86760, Peak mem 11.820 GB
+Iter 103: Train loss 1.802, Learning Rate 9.359e-05, It/sec 0.395, Tokens/sec 191.385, Trained Tokens 87245, Peak mem 11.820 GB
+Iter 104: Train loss 1.638, Learning Rate 9.345e-05, It/sec 0.182, Tokens/sec 202.104, Trained Tokens 88354, Peak mem 11.820 GB
+Iter 105: Train loss 1.574, Learning Rate 9.331e-05, It/sec 0.496, Tokens/sec 203.163, Trained Tokens 88764, Peak mem 11.820 GB
+Iter 106: Train loss 1.657, Learning Rate 9.317e-05, It/sec 0.201, Tokens/sec 195.999, Trained Tokens 89738, Peak mem 11.820 GB
+Iter 107: Train loss 1.725, Learning Rate 9.303e-05, It/sec 0.291, Tokens/sec 209.459, Trained Tokens 90459, Peak mem 11.820 GB
+Iter 108: Train loss 1.836, Learning Rate 9.288e-05, It/sec 0.207, Tokens/sec 196.198, Trained Tokens 91409, Peak mem 11.820 GB
+Iter 109: Train loss 1.619, Learning Rate 9.274e-05, It/sec 0.203, Tokens/sec 205.953, Trained Tokens 92422, Peak mem 11.820 GB
+Iter 110: Train loss 1.762, Learning Rate 9.259e-05, It/sec 0.435, Tokens/sec 219.421, Trained Tokens 92926, Peak mem 11.820 GB
+Iter 111: Train loss 1.611, Learning Rate 9.244e-05, It/sec 0.181, Tokens/sec 187.917, Trained Tokens 93964, Peak mem 11.820 GB
+Iter 112: Train loss 1.572, Learning Rate 9.230e-05, It/sec 0.165, Tokens/sec 201.447, Trained Tokens 95186, Peak mem 11.820 GB
+Iter 113: Train loss 1.766, Learning Rate 9.214e-05, It/sec 0.267, Tokens/sec 204.239, Trained Tokens 95950, Peak mem 11.820 GB
+Iter 114: Train loss 1.511, Learning Rate 9.199e-05, It/sec 0.457, Tokens/sec 160.986, Trained Tokens 96302, Peak mem 11.820 GB
+Iter 115: Train loss 1.864, Learning Rate 9.184e-05, It/sec 0.115, Tokens/sec 184.740, Trained Tokens 97903, Peak mem 11.820 GB
+Iter 116: Train loss 1.711, Learning Rate 9.169e-05, It/sec 0.214, Tokens/sec 200.400, Trained Tokens 98840, Peak mem 11.820 GB
+Iter 117: Train loss 1.732, Learning Rate 9.153e-05, It/sec 0.271, Tokens/sec 204.038, Trained Tokens 99593, Peak mem 11.820 GB
+Iter 118: Train loss 1.416, Learning Rate 9.137e-05, It/sec 0.264, Tokens/sec 214.608, Trained Tokens 100405, Peak mem 11.820 GB
+Iter 119: Train loss 1.603, Learning Rate 9.122e-05, It/sec 0.224, Tokens/sec 205.712, Trained Tokens 101323, Peak mem 11.820 GB
+Iter 120: Train loss 2.044, Learning Rate 9.106e-05, It/sec 0.094, Tokens/sec 181.083, Trained Tokens 103247, Peak mem 11.820 GB
+Iter 121: Train loss 1.574, Learning Rate 9.090e-05, It/sec 0.258, Tokens/sec 206.953, Trained Tokens 104050, Peak mem 11.820 GB
+Iter 122: Train loss 1.573, Learning Rate 9.073e-05, It/sec 0.347, Tokens/sec 191.431, Trained Tokens 104602, Peak mem 11.820 GB
+Iter 123: Train loss 1.477, Learning Rate 9.057e-05, It/sec 0.193, Tokens/sec 197.456, Trained Tokens 105625, Peak mem 11.820 GB
+Iter 124: Train loss 1.889, Learning Rate 9.041e-05, It/sec 0.120, Tokens/sec 171.125, Trained Tokens 107049, Peak mem 11.820 GB
+Iter 125: Train loss 1.651, Learning Rate 9.024e-05, It/sec 0.139, Tokens/sec 122.182, Trained Tokens 107925, Peak mem 11.820 GB
+Iter 126: Train loss 1.436, Learning Rate 9.008e-05, It/sec 0.180, Tokens/sec 149.931, Trained Tokens 108756, Peak mem 11.820 GB
+Iter 127: Train loss 1.453, Learning Rate 8.991e-05, It/sec 0.144, Tokens/sec 99.762, Trained Tokens 109447, Peak mem 11.820 GB
+Iter 128: Train loss 1.740, Learning Rate 8.974e-05, It/sec 0.097, Tokens/sec 139.837, Trained Tokens 110896, Peak mem 11.820 GB
+Iter 129: Train loss 1.829, Learning Rate 8.957e-05, It/sec 0.240, Tokens/sec 159.669, Trained Tokens 111560, Peak mem 11.820 GB
+Iter 130: Train loss 1.707, Learning Rate 8.940e-05, It/sec 0.307, Tokens/sec 141.373, Trained Tokens 112021, Peak mem 11.820 GB
+Iter 131: Train loss 1.680, Learning Rate 8.922e-05, It/sec 0.283, Tokens/sec 166.546, Trained Tokens 112609, Peak mem 11.820 GB
+Iter 132: Train loss 1.522, Learning Rate 8.905e-05, It/sec 0.075, Tokens/sec 77.919, Trained Tokens 113646, Peak mem 11.820 GB
+Iter 133: Train loss 1.364, Learning Rate 8.887e-05, It/sec 0.197, Tokens/sec 101.780, Trained Tokens 114162, Peak mem 11.820 GB
+Iter 134: Train loss 1.609, Learning Rate 8.870e-05, It/sec 0.083, Tokens/sec 92.891, Trained Tokens 115279, Peak mem 11.820 GB
+Iter 135: Train loss 1.648, Learning Rate 8.852e-05, It/sec 0.056, Tokens/sec 83.222, Trained Tokens 116761, Peak mem 11.820 GB
+Iter 136: Train loss 1.850, Learning Rate 8.834e-05, It/sec 0.086, Tokens/sec 57.123, Trained Tokens 117424, Peak mem 11.820 GB
+Iter 137: Train loss 1.868, Learning Rate 8.816e-05, It/sec 0.035, Tokens/sec 49.049, Trained Tokens 118819, Peak mem 11.820 GB
+Iter 138: Train loss 1.664, Learning Rate 8.798e-05, It/sec 0.032, Tokens/sec 45.821, Trained Tokens 120238, Peak mem 11.820 GB
+Iter 139: Train loss 1.497, Learning Rate 8.780e-05, It/sec 0.077, Tokens/sec 62.477, Trained Tokens 121052, Peak mem 11.820 GB
+Iter 140: Train loss 1.602, Learning Rate 8.761e-05, It/sec 0.122, Tokens/sec 55.672, Trained Tokens 121510, Peak mem 11.820 GB
+Iter 141: Train loss 1.410, Learning Rate 8.743e-05, It/sec 0.061, Tokens/sec 65.152, Trained Tokens 122571, Peak mem 11.820 GB
+Iter 142: Train loss 1.371, Learning Rate 8.724e-05, It/sec 0.076, Tokens/sec 57.939, Trained Tokens 123338, Peak mem 11.820 GB
+Iter 143: Train loss 1.350, Learning Rate 8.706e-05, It/sec 0.117, Tokens/sec 61.701, Trained Tokens 123866, Peak mem 11.820 GB
+Iter 144: Train loss 1.645, Learning Rate 8.687e-05, It/sec 0.233, Tokens/sec 96.797, Trained Tokens 124281, Peak mem 11.820 GB
+Iter 145: Train loss 1.586, Learning Rate 8.668e-05, It/sec 0.084, Tokens/sec 67.882, Trained Tokens 125087, Peak mem 11.820 GB
+Iter 146: Train loss 1.564, Learning Rate 8.649e-05, It/sec 0.097, Tokens/sec 96.270, Trained Tokens 126075, Peak mem 11.820 GB
+Iter 147: Train loss 1.623, Learning Rate 8.630e-05, It/sec 0.076, Tokens/sec 92.571, Trained Tokens 127286, Peak mem 11.820 GB
+Iter 148: Train loss 1.606, Learning Rate 8.610e-05, It/sec 0.066, Tokens/sec 94.580, Trained Tokens 128721, Peak mem 11.820 GB
+Iter 149: Train loss 1.564, Learning Rate 8.591e-05, It/sec 0.133, Tokens/sec 135.051, Trained Tokens 129740, Peak mem 11.820 GB
+Iter 150: Train loss 1.759, Learning Rate 8.571e-05, It/sec 0.099, Tokens/sec 103.457, Trained Tokens 130781, Peak mem 11.820 GB
+Iter 151: Train loss 1.587, Learning Rate 8.552e-05, It/sec 0.212, Tokens/sec 166.329, Trained Tokens 131566, Peak mem 11.820 GB
+Iter 152: Train loss 1.540, Learning Rate 8.532e-05, It/sec 0.091, Tokens/sec 119.728, Trained Tokens 132878, Peak mem 11.820 GB
+Iter 153: Train loss 1.689, Learning Rate 8.512e-05, It/sec 0.157, Tokens/sec 125.928, Trained Tokens 133680, Peak mem 11.820 GB
+Iter 154: Train loss 1.781, Learning Rate 8.493e-05, It/sec 0.081, Tokens/sec 108.726, Trained Tokens 135027, Peak mem 11.820 GB
+Iter 155: Train loss 1.673, Learning Rate 8.472e-05, It/sec 0.149, Tokens/sec 182.359, Trained Tokens 136248, Peak mem 11.820 GB
+Iter 156: Train loss 1.825, Learning Rate 8.452e-05, It/sec 0.292, Tokens/sec 188.467, Trained Tokens 136894, Peak mem 11.820 GB
+Iter 157: Train loss 1.497, Learning Rate 8.432e-05, It/sec 0.334, Tokens/sec 180.120, Trained Tokens 137434, Peak mem 11.820 GB
+Iter 158: Train loss 1.596, Learning Rate 8.412e-05, It/sec 0.368, Tokens/sec 176.641, Trained Tokens 137914, Peak mem 11.820 GB
+Iter 159: Train loss 1.679, Learning Rate 8.391e-05, It/sec 0.199, Tokens/sec 171.155, Trained Tokens 138774, Peak mem 11.820 GB
+Iter 160: Train loss 1.514, Learning Rate 8.371e-05, It/sec 0.193, Tokens/sec 182.780, Trained Tokens 139722, Peak mem 11.820 GB
+Iter 161: Train loss 1.432, Learning Rate 8.350e-05, It/sec 0.303, Tokens/sec 194.772, Trained Tokens 140365, Peak mem 11.820 GB
+Iter 162: Train loss 2.050, Learning Rate 8.330e-05, It/sec 0.105, Tokens/sec 164.724, Trained Tokens 141937, Peak mem 11.820 GB
+Iter 163: Train loss 1.919, Learning Rate 8.309e-05, It/sec 0.143, Tokens/sec 162.492, Trained Tokens 143075, Peak mem 11.820 GB
+Iter 164: Train loss 1.551, Learning Rate 8.288e-05, It/sec 0.285, Tokens/sec 126.121, Trained Tokens 143518, Peak mem 11.820 GB
+Iter 165: Train loss 1.431, Learning Rate 8.267e-05, It/sec 0.242, Tokens/sec 179.960, Trained Tokens 144262, Peak mem 11.820 GB
+Iter 166: Train loss 1.819, Learning Rate 8.246e-05, It/sec 0.232, Tokens/sec 159.555, Trained Tokens 144951, Peak mem 11.820 GB
+Iter 167: Train loss 1.871, Learning Rate 8.224e-05, It/sec 0.151, Tokens/sec 176.992, Trained Tokens 146123, Peak mem 11.820 GB
+Iter 168: Train loss 1.393, Learning Rate 8.203e-05, It/sec 0.323, Tokens/sec 144.483, Trained Tokens 146571, Peak mem 11.820 GB
+Iter 169: Train loss 1.864, Learning Rate 8.182e-05, It/sec 0.127, Tokens/sec 145.733, Trained Tokens 147721, Peak mem 11.820 GB
+Iter 170: Train loss 1.534, Learning Rate 8.160e-05, It/sec 0.140, Tokens/sec 153.570, Trained Tokens 148821, Peak mem 11.820 GB
+Iter 171: Train loss 1.757, Learning Rate 8.139e-05, It/sec 0.208, Tokens/sec 154.295, Trained Tokens 149563, Peak mem 11.820 GB
+Iter 172: Train loss 1.809, Learning Rate 8.117e-05, It/sec 0.039, Tokens/sec 45.915, Trained Tokens 150740, Peak mem 11.820 GB
+Iter 173: Train loss 1.486, Learning Rate 8.095e-05, It/sec 0.102, Tokens/sec 109.155, Trained Tokens 151805, Peak mem 11.820 GB
+Iter 174: Train loss 1.214, Learning Rate 8.073e-05, It/sec 0.334, Tokens/sec 163.511, Trained Tokens 152295, Peak mem 11.820 GB
+Iter 175: Train loss 1.746, Learning Rate 8.051e-05, It/sec 0.337, Tokens/sec 179.316, Trained Tokens 152827, Peak mem 11.820 GB
+Iter 176: Train loss 1.728, Learning Rate 8.029e-05, It/sec 0.125, Tokens/sec 174.942, Trained Tokens 154225, Peak mem 11.820 GB
+Iter 177: Train loss 1.883, Learning Rate 8.007e-05, It/sec 0.088, Tokens/sec 159.737, Trained Tokens 156050, Peak mem 11.820 GB
+Iter 178: Train loss 1.596, Learning Rate 7.985e-05, It/sec 0.323, Tokens/sec 176.292, Trained Tokens 156596, Peak mem 11.820 GB
+Iter 179: Train loss 1.331, Learning Rate 7.962e-05, It/sec 0.438, Tokens/sec 176.757, Trained Tokens 157000, Peak mem 11.820 GB
+Iter 180: Train loss 1.533, Learning Rate 7.940e-05, It/sec 0.193, Tokens/sec 175.520, Trained Tokens 157911, Peak mem 11.820 GB
+Iter 181: Train loss 1.702, Learning Rate 7.918e-05, It/sec 0.399, Tokens/sec 181.867, Trained Tokens 158367, Peak mem 11.820 GB
+Iter 182: Train loss 1.467, Learning Rate 7.895e-05, It/sec 0.144, Tokens/sec 151.373, Trained Tokens 159419, Peak mem 11.820 GB
+Iter 183: Train loss 1.502, Learning Rate 7.872e-05, It/sec 0.210, Tokens/sec 120.807, Trained Tokens 159994, Peak mem 11.820 GB
+Iter 184: Train loss 1.432, Learning Rate 7.850e-05, It/sec 0.108, Tokens/sec 131.116, Trained Tokens 161212, Peak mem 11.820 GB
+Iter 185: Train loss 1.423, Learning Rate 7.827e-05, It/sec 0.376, Tokens/sec 186.293, Trained Tokens 161707, Peak mem 11.820 GB
+Iter 186: Train loss 1.537, Learning Rate 7.804e-05, It/sec 0.103, Tokens/sec 115.227, Trained Tokens 162825, Peak mem 11.820 GB
+Iter 187: Train loss 1.812, Learning Rate 7.781e-05, It/sec 0.164, Tokens/sec 115.836, Trained Tokens 163531, Peak mem 11.820 GB
+Iter 188: Train loss 1.440, Learning Rate 7.758e-05, It/sec 0.118, Tokens/sec 121.534, Trained Tokens 164559, Peak mem 11.820 GB
+Iter 189: Train loss 1.786, Learning Rate 7.735e-05, It/sec 0.199, Tokens/sec 114.110, Trained Tokens 165132, Peak mem 11.820 GB
+Iter 190: Train loss 1.399, Learning Rate 7.711e-05, It/sec 0.308, Tokens/sec 143.171, Trained Tokens 165597, Peak mem 11.820 GB
+Iter 191: Train loss 1.683, Learning Rate 7.688e-05, It/sec 0.151, Tokens/sec 164.176, Trained Tokens 166682, Peak mem 11.820 GB
+Iter 192: Train loss 1.772, Learning Rate 7.665e-05, It/sec 0.073, Tokens/sec 105.064, Trained Tokens 168120, Peak mem 11.820 GB
+Iter 193: Train loss 1.393, Learning Rate 7.641e-05, It/sec 0.126, Tokens/sec 110.948, Trained Tokens 169004, Peak mem 11.820 GB
+Iter 194: Train loss 1.484, Learning Rate 7.618e-05, It/sec 0.107, Tokens/sec 101.714, Trained Tokens 169957, Peak mem 11.820 GB
+Iter 195: Train loss 1.328, Learning Rate 7.594e-05, It/sec 0.150, Tokens/sec 85.030, Trained Tokens 170524, Peak mem 11.820 GB
+Iter 196: Train loss 1.386, Learning Rate 7.570e-05, It/sec 0.202, Tokens/sec 90.825, Trained Tokens 170974, Peak mem 11.820 GB
+Iter 197: Train loss 1.487, Learning Rate 7.547e-05, It/sec 0.062, Tokens/sec 80.995, Trained Tokens 172286, Peak mem 11.820 GB
+Iter 198: Train loss 1.588, Learning Rate 7.523e-05, It/sec 0.084, Tokens/sec 97.551, Trained Tokens 173444, Peak mem 11.820 GB
+Iter 199: Train loss 1.701, Learning Rate 7.499e-05, It/sec 0.117, Tokens/sec 123.402, Trained Tokens 174501, Peak mem 11.820 GB
+Calculating loss...: 0it [00:00, ?it/s]
+Iter 200: Val loss nan, Val took 0.012s
+Iter 200: Train loss 1.545, Learning Rate 7.475e-05, It/sec 0.123, Tokens/sec 139.085, Trained Tokens 175628, Peak mem 11.820 GB
+Iter 200: Saved adapter weights to adapters-conscious/adapters.safetensors and adapters-conscious/0000200_adapters.safetensors.
+Iter 201: Train loss 1.765, Learning Rate 7.451e-05, It/sec 0.073, Tokens/sec 136.042, Trained Tokens 177501, Peak mem 11.820 GB
+Iter 202: Train loss 1.477, Learning Rate 7.427e-05, It/sec 0.172, Tokens/sec 154.873, Trained Tokens 178401, Peak mem 11.820 GB
+Iter 203: Train loss 1.313, Learning Rate 7.402e-05, It/sec 0.233, Tokens/sec 159.479, Trained Tokens 179085, Peak mem 11.820 GB
+Iter 204: Train loss 1.465, Learning Rate 7.378e-05, It/sec 0.114, Tokens/sec 142.336, Trained Tokens 180337, Peak mem 11.820 GB
+Iter 205: Train loss 1.597, Learning Rate 7.354e-05, It/sec 0.160, Tokens/sec 148.968, Trained Tokens 181266, Peak mem 11.820 GB
+Iter 206: Train loss 1.135, Learning Rate 7.329e-05, It/sec 0.354, Tokens/sec 152.220, Trained Tokens 181696, Peak mem 11.820 GB
+Iter 207: Train loss 1.571, Learning Rate 7.305e-05, It/sec 0.146, Tokens/sec 156.147, Trained Tokens 182766, Peak mem 11.820 GB
+Iter 208: Train loss 1.458, Learning Rate 7.281e-05, It/sec 0.142, Tokens/sec 146.358, Trained Tokens 183799, Peak mem 11.820 GB
+Iter 209: Train loss 1.607, Learning Rate 7.256e-05, It/sec 0.135, Tokens/sec 147.157, Trained Tokens 184888, Peak mem 11.820 GB
+Iter 210: Train loss 1.819, Learning Rate 7.231e-05, It/sec 0.099, Tokens/sec 150.679, Trained Tokens 186417, Peak mem 11.820 GB
+Iter 211: Train loss 1.316, Learning Rate 7.207e-05, It/sec 0.379, Tokens/sec 155.367, Trained Tokens 186827, Peak mem 11.820 GB
+Iter 212: Train loss 1.595, Learning Rate 7.182e-05, It/sec 0.114, Tokens/sec 141.397, Trained Tokens 188066, Peak mem 11.820 GB
+Iter 213: Train loss 1.701, Learning Rate 7.157e-05, It/sec 0.307, Tokens/sec 154.370, Trained Tokens 188569, Peak mem 11.820 GB
+Iter 214: Train loss 1.499, Learning Rate 7.132e-05, It/sec 0.169, Tokens/sec 153.762, Trained Tokens 189480, Peak mem 11.820 GB
+Iter 215: Train loss 1.665, Learning Rate 7.107e-05, It/sec 0.284, Tokens/sec 145.611, Trained Tokens 189993, Peak mem 11.820 GB
+Iter 216: Train loss 1.536, Learning Rate 7.082e-05, It/sec 0.219, Tokens/sec 141.877, Trained Tokens 190642, Peak mem 11.820 GB
+Iter 217: Train loss 1.492, Learning Rate 7.057e-05, It/sec 0.201, Tokens/sec 150.214, Trained Tokens 191389, Peak mem 11.820 GB
+Iter 218: Train loss 1.678, Learning Rate 7.032e-05, It/sec 0.120, Tokens/sec 131.900, Trained Tokens 192485, Peak mem 11.820 GB
+Iter 219: Train loss 1.329, Learning Rate 7.007e-05, It/sec 0.247, Tokens/sec 144.992, Trained Tokens 193071, Peak mem 11.820 GB
+Iter 220: Train loss 1.608, Learning Rate 6.982e-05, It/sec 0.100, Tokens/sec 135.051, Trained Tokens 194424, Peak mem 11.820 GB
+Iter 221: Train loss 1.642, Learning Rate 6.956e-05, It/sec 0.109, Tokens/sec 103.494, Trained Tokens 195373, Peak mem 11.820 GB
+Iter 222: Train loss 1.370, Learning Rate 6.931e-05, It/sec 0.132, Tokens/sec 96.455, Trained Tokens 196101, Peak mem 11.820 GB
+Iter 223: Train loss 1.598, Learning Rate 6.906e-05, It/sec 0.107, Tokens/sec 94.110, Trained Tokens 196982, Peak mem 11.820 GB
+Iter 224: Train loss 1.776, Learning Rate 6.880e-05, It/sec 0.080, Tokens/sec 121.106, Trained Tokens 198505, Peak mem 11.820 GB
+Iter 225: Train loss 1.394, Learning Rate 6.855e-05, It/sec 0.192, Tokens/sec 153.970, Trained Tokens 199306, Peak mem 11.820 GB
+Iter 226: Train loss 1.874, Learning Rate 6.829e-05, It/sec 0.065, Tokens/sec 127.116, Trained Tokens 201247, Peak mem 11.820 GB
+Iter 227: Train loss 1.648, Learning Rate 6.804e-05, It/sec 0.131, Tokens/sec 155.569, Trained Tokens 202438, Peak mem 11.820 GB
+Iter 228: Train loss 1.507, Learning Rate 6.778e-05, It/sec 0.358, Tokens/sec 156.911, Trained Tokens 202876, Peak mem 11.820 GB
+Iter 229: Train loss 1.342, Learning Rate 6.753e-05, It/sec 0.416, Tokens/sec 170.156, Trained Tokens 203285, Peak mem 11.820 GB
+Iter 230: Train loss 1.790, Learning Rate 6.727e-05, It/sec 0.225, Tokens/sec 166.161, Trained Tokens 204025, Peak mem 11.820 GB
+Iter 231: Train loss 1.496, Learning Rate 6.701e-05, It/sec 0.162, Tokens/sec 164.406, Trained Tokens 205039, Peak mem 11.820 GB
+Iter 232: Train loss 1.637, Learning Rate 6.675e-05, It/sec 0.095, Tokens/sec 126.111, Trained Tokens 206367, Peak mem 11.820 GB
+Iter 233: Train loss 1.690, Learning Rate 6.650e-05, It/sec 0.106, Tokens/sec 104.734, Trained Tokens 207353, Peak mem 11.820 GB
+Iter 234: Train loss 1.841, Learning Rate 6.624e-05, It/sec 0.092, Tokens/sec 124.355, Trained Tokens 208706, Peak mem 11.820 GB
+Iter 235: Train loss 1.816, Learning Rate 6.598e-05, It/sec 0.076, Tokens/sec 95.014, Trained Tokens 209964, Peak mem 11.820 GB
+Iter 236: Train loss 1.231, Learning Rate 6.572e-05, It/sec 0.144, Tokens/sec 92.802, Trained Tokens 210608, Peak mem 11.820 GB
+Iter 237: Train loss 1.654, Learning Rate 6.546e-05, It/sec 0.153, Tokens/sec 136.474, Trained Tokens 211499, Peak mem 11.820 GB
+Iter 238: Train loss 1.555, Learning Rate 6.520e-05, It/sec 0.114, Tokens/sec 145.523, Trained Tokens 212780, Peak mem 11.820 GB
+Iter 239: Train loss 1.381, Learning Rate 6.494e-05, It/sec 0.182, Tokens/sec 166.046, Trained Tokens 213690, Peak mem 11.820 GB
+Iter 240: Train loss 1.822, Learning Rate 6.468e-05, It/sec 0.084, Tokens/sec 130.779, Trained Tokens 215242, Peak mem 11.820 GB
+Iter 241: Train loss 1.265, Learning Rate 6.442e-05, It/sec 0.293, Tokens/sec 155.158, Trained Tokens 215772, Peak mem 11.820 GB
+Iter 242: Train loss 1.444, Learning Rate 6.416e-05, It/sec 0.126, Tokens/sec 125.096, Trained Tokens 216766, Peak mem 11.820 GB
+Iter 243: Train loss 1.420, Learning Rate 6.389e-05, It/sec 0.138, Tokens/sec 108.551, Trained Tokens 217551, Peak mem 11.820 GB
+Iter 244: Train loss 1.517, Learning Rate 6.363e-05, It/sec 0.372, Tokens/sec 142.479, Trained Tokens 217934, Peak mem 11.820 GB
+Iter 245: Train loss 1.152, Learning Rate 6.337e-05, It/sec 0.197, Tokens/sec 104.221, Trained Tokens 218464, Peak mem 11.820 GB
+Iter 246: Train loss 1.314, Learning Rate 6.311e-05, It/sec 0.327, Tokens/sec 170.180, Trained Tokens 218985, Peak mem 11.820 GB
+Iter 247: Train loss 1.942, Learning Rate 6.284e-05, It/sec 0.168, Tokens/sec 158.578, Trained Tokens 219927, Peak mem 11.820 GB
+Iter 248: Train loss 1.550, Learning Rate 6.258e-05, It/sec 0.423, Tokens/sec 170.281, Trained Tokens 220330, Peak mem 11.820 GB
+Iter 249: Train loss 1.512, Learning Rate 6.232e-05, It/sec 0.169, Tokens/sec 162.514, Trained Tokens 221291, Peak mem 11.820 GB
+Iter 250: Train loss 1.724, Learning Rate 6.205e-05, It/sec 0.241, Tokens/sec 166.872, Trained Tokens 221983, Peak mem 11.820 GB
+Iter 251: Train loss 1.963, Learning Rate 6.179e-05, It/sec 0.232, Tokens/sec 170.161, Trained Tokens 222715, Peak mem 11.820 GB
+Iter 252: Train loss 1.407, Learning Rate 6.152e-05, It/sec 0.234, Tokens/sec 178.210, Trained Tokens 223477, Peak mem 11.820 GB
+Iter 253: Train loss 1.675, Learning Rate 6.126e-05, It/sec 0.099, Tokens/sec 121.773, Trained Tokens 224701, Peak mem 11.820 GB
+Iter 254: Train loss 1.855, Learning Rate 6.100e-05, It/sec 0.071, Tokens/sec 149.504, Trained Tokens 226807, Peak mem 12.008 GB
+Iter 255: Train loss 1.708, Learning Rate 6.073e-05, It/sec 0.215, Tokens/sec 168.296, Trained Tokens 227589, Peak mem 12.008 GB
+Iter 256: Train loss 1.744, Learning Rate 6.046e-05, It/sec 0.118, Tokens/sec 115.592, Trained Tokens 228568, Peak mem 12.008 GB
+Iter 257: Train loss 1.602, Learning Rate 6.020e-05, It/sec 0.113, Tokens/sec 139.105, Trained Tokens 229804, Peak mem 12.008 GB
+Iter 258: Train loss 1.619, Learning Rate 5.993e-05, It/sec 0.195, Tokens/sec 168.801, Trained Tokens 230668, Peak mem 12.008 GB
+Iter 259: Train loss 1.517, Learning Rate 5.967e-05, It/sec 0.190, Tokens/sec 127.542, Trained Tokens 231339, Peak mem 12.008 GB
+Iter 260: Train loss 1.413, Learning Rate 5.940e-05, It/sec 0.274, Tokens/sec 118.404, Trained Tokens 231771, Peak mem 12.008 GB
+Iter 261: Train loss 1.472, Learning Rate 5.914e-05, It/sec 0.189, Tokens/sec 138.943, Trained Tokens 232506, Peak mem 12.008 GB
+Iter 262: Train loss 1.870, Learning Rate 5.887e-05, It/sec 0.192, Tokens/sec 160.497, Trained Tokens 233341, Peak mem 12.008 GB
+Iter 263: Train loss 1.643, Learning Rate 5.860e-05, It/sec 0.075, Tokens/sec 102.682, Trained Tokens 234704, Peak mem 12.008 GB
+Iter 264: Train loss 1.724, Learning Rate 5.834e-05, It/sec 0.078, Tokens/sec 103.747, Trained Tokens 236036, Peak mem 12.008 GB
+Iter 265: Train loss 1.190, Learning Rate 5.807e-05, It/sec 0.180, Tokens/sec 112.618, Trained Tokens 236663, Peak mem 12.008 GB
+Iter 266: Train loss 1.776, Learning Rate 5.780e-05, It/sec 0.237, Tokens/sec 137.264, Trained Tokens 237242, Peak mem 12.008 GB
+Iter 267: Train loss 1.767, Learning Rate 5.754e-05, It/sec 0.081, Tokens/sec 90.226, Trained Tokens 238351, Peak mem 12.008 GB
+Iter 268: Train loss 1.360, Learning Rate 5.727e-05, It/sec 0.269, Tokens/sec 83.180, Trained Tokens 238660, Peak mem 12.008 GB
+Iter 269: Train loss 1.586, Learning Rate 5.700e-05, It/sec 0.152, Tokens/sec 88.995, Trained Tokens 239245, Peak mem 12.008 GB
+Iter 270: Train loss 1.806, Learning Rate 5.674e-05, It/sec 0.092, Tokens/sec 86.425, Trained Tokens 240189, Peak mem 12.008 GB
+Iter 271: Train loss 1.572, Learning Rate 5.647e-05, It/sec 0.102, Tokens/sec 92.008, Trained Tokens 241093, Peak mem 12.008 GB
+Iter 272: Train loss 1.598, Learning Rate 5.620e-05, It/sec 0.184, Tokens/sec 85.830, Trained Tokens 241560, Peak mem 12.008 GB
+Iter 273: Train loss 1.406, Learning Rate 5.594e-05, It/sec 0.085, Tokens/sec 77.487, Trained Tokens 242474, Peak mem 12.008 GB
+Iter 274: Train loss 1.448, Learning Rate 5.567e-05, It/sec 0.126, Tokens/sec 80.845, Trained Tokens 243117, Peak mem 12.008 GB
+Iter 275: Train loss 1.509, Learning Rate 5.540e-05, It/sec 0.079, Tokens/sec 73.888, Trained Tokens 244058, Peak mem 12.008 GB
+Iter 276: Train loss 1.691, Learning Rate 5.513e-05, It/sec 0.076, Tokens/sec 77.338, Trained Tokens 245075, Peak mem 12.008 GB
+Iter 277: Train loss 1.492, Learning Rate 5.487e-05, It/sec 0.049, Tokens/sec 72.689, Trained Tokens 246553, Peak mem 12.008 GB
+Iter 278: Train loss 1.470, Learning Rate 5.460e-05, It/sec 0.056, Tokens/sec 62.069, Trained Tokens 247662, Peak mem 12.008 GB
+Iter 279: Train loss 1.496, Learning Rate 5.433e-05, It/sec 0.092, Tokens/sec 102.958, Trained Tokens 248787, Peak mem 12.008 GB
+Iter 280: Train loss 1.512, Learning Rate 5.406e-05, It/sec 0.063, Tokens/sec 66.578, Trained Tokens 249848, Peak mem 12.008 GB
+Iter 281: Train loss 1.267, Learning Rate 5.380e-05, It/sec 0.112, Tokens/sec 86.708, Trained Tokens 250622, Peak mem 12.008 GB
+Iter 282: Train loss 1.925, Learning Rate 5.353e-05, It/sec 0.074, Tokens/sec 91.624, Trained Tokens 251867, Peak mem 12.008 GB
+Iter 283: Train loss 1.843, Learning Rate 5.326e-05, It/sec 0.073, Tokens/sec 111.265, Trained Tokens 253389, Peak mem 12.008 GB
+Iter 284: Train loss 1.584, Learning Rate 5.300e-05, It/sec 0.121, Tokens/sec 130.482, Trained Tokens 254468, Peak mem 12.008 GB
+Iter 285: Train loss 1.577, Learning Rate 5.273e-05, It/sec 0.143, Tokens/sec 142.670, Trained Tokens 255469, Peak mem 12.008 GB
+Iter 286: Train loss 1.634, Learning Rate 5.246e-05, It/sec 0.082, Tokens/sec 127.110, Trained Tokens 257023, Peak mem 12.008 GB
+Iter 287: Train loss 1.351, Learning Rate 5.220e-05, It/sec 0.267, Tokens/sec 146.120, Trained Tokens 257570, Peak mem 12.008 GB
+Iter 288: Train loss 1.854, Learning Rate 5.193e-05, It/sec 0.059, Tokens/sec 104.093, Trained Tokens 259344, Peak mem 12.008 GB
+Iter 289: Train loss 1.411, Learning Rate 5.166e-05, It/sec 0.155, Tokens/sec 127.500, Trained Tokens 260167, Peak mem 12.008 GB
+Iter 290: Train loss 1.404, Learning Rate 5.140e-05, It/sec 0.274, Tokens/sec 146.790, Trained Tokens 260702, Peak mem 12.008 GB
+Iter 291: Train loss 1.674, Learning Rate 5.113e-05, It/sec 0.097, Tokens/sec 131.358, Trained Tokens 262061, Peak mem 12.008 GB
+Iter 292: Train loss 1.512, Learning Rate 5.086e-05, It/sec 0.228, Tokens/sec 178.251, Trained Tokens 262844, Peak mem 12.008 GB
+Iter 293: Train loss 1.526, Learning Rate 5.060e-05, It/sec 0.374, Tokens/sec 166.606, Trained Tokens 263290, Peak mem 12.008 GB
+Iter 294: Train loss 1.561, Learning Rate 5.033e-05, It/sec 0.137, Tokens/sec 132.210, Trained Tokens 264253, Peak mem 12.008 GB
+Iter 295: Train loss 1.640, Learning Rate 5.007e-05, It/sec 0.117, Tokens/sec 154.404, Trained Tokens 265573, Peak mem 12.008 GB
+Iter 296: Train loss 1.613, Learning Rate 4.980e-05, It/sec 0.150, Tokens/sec 161.536, Trained Tokens 266647, Peak mem 12.008 GB
+Iter 297: Train loss 1.422, Learning Rate 4.954e-05, It/sec 0.197, Tokens/sec 134.622, Trained Tokens 267329, Peak mem 12.008 GB
+Iter 298: Train loss 1.944, Learning Rate 4.927e-05, It/sec 0.138, Tokens/sec 180.145, Trained Tokens 268635, Peak mem 12.008 GB
+Iter 299: Train loss 1.573, Learning Rate 4.900e-05, It/sec 0.114, Tokens/sec 137.622, Trained Tokens 269837, Peak mem 12.008 GB
+Calculating loss...: 0it [00:00, ?it/s]
+Iter 300: Val loss nan, Val took 0.008s
+Iter 300: Train loss 1.747, Learning Rate 4.874e-05, It/sec 0.232, Tokens/sec 175.489, Trained Tokens 270593, Peak mem 12.008 GB
+Iter 300: Saved adapter weights to adapters-conscious/adapters.safetensors and adapters-conscious/0000300_adapters.safetensors.
+Saved final weights to adapters-conscious/adapters.safetensors.

adapters-conscious/wandb/run-20260216_150557-tugniqt7/files/requirements.txt

@@ -0,0 +1,147 @@
+shellingham==1.5.4
+contourpy==1.3.2
+jiter==0.12.0
+audioread==3.1.0
+threadpoolctl==3.6.0
+lazy_loader==0.4
+GitPython==3.1.45
+async-timeout==5.0.1
+requests==2.32.5
+rich==14.2.0
+tokenizers==0.22.1
+urllib3==2.5.0
+exceptiongroup==1.3.1
+numpy==2.2.6
+click==8.3.1
+pytz==2025.2
+miniaudio==1.61
+pyarrow==22.0.0
+sse-starlette==3.2.0
+scikit-learn==1.7.2
+soxr==1.0.0
+jsonschema-specifications==2025.9.1
+python-multipart==0.0.22
+utilsforecast==0.2.15
+ftfy==6.3.1
+torchvision==0.25.0
+statsmodels==0.14.6
+file-read-backwards==3.2.0
+propcache==0.4.1
+python-dotenv==1.2.1
+anyio==4.12.0
+mlx==0.30.6
+wordfreq==3.1.1
+networkx==3.4.2
+pip==25.3
+texttable==1.7.0
+mlx-audio==0.3.1
+narwhals==2.15.0
+multidict==6.7.0
+numba==0.63.1
+idna==3.11
+regex==2025.11.3
+fonttools==4.60.1
+openai==2.16.0
+aiohttp==3.13.2
+mistral_common==1.8.6
+einshape==1.0
+cffi==2.0.0
+kiwisolver==1.4.9
+tqdm==4.67.1
+setuptools==80.9.0
+RapidFuzz==3.14.3
+pyparsing==3.2.5
+starlette==0.52.1
+tzdata==2025.2
+mlx-lm==0.30.6
+httpcore==1.0.9
+decorator==5.2.1
+certifi==2025.11.12
+typer==0.21.1
+pydantic==2.12.4
+fsspec==2025.10.0
+mcp==1.26.0
+librosa==0.11.0
+charset-normalizer==3.4.4
+sympy==1.14.0
+jsonschema==4.25.1
+pydantic-settings==2.12.0
+markdown-it-py==4.0.0
+tiktoken==0.12.0
+PyJWT==2.11.0
+sentry-sdk==2.45.0
+platformdirs==4.5.0
+absl-py==2.3.1
+transformers==5.1.0
+diffusers==0.37.0.dev0
+h11==0.16.0
+gitdb==4.0.12
+sniffio==1.3.1
+pycparser==3.0
+sentencepiece==0.2.1
+importlib_metadata==8.7.1
+mdurl==0.1.2
+patsy==1.0.2
+python-dateutil==2.9.0.post0
+mpmath==1.3.0
+pillow==12.0.0
+PyYAML==6.0.3
+sentence-transformers==5.1.2
+multiprocess==0.70.18
+pydantic_core==2.41.5
+uvicorn==0.40.0
+frozenlist==1.8.0
+typer-slim==0.20.1
+typing_extensions==4.15.0
+aiosignal==1.4.0
+packaging==25.0
+cycler==0.12.1
+cryptography==46.0.4
+hf-xet==1.2.0
+Jinja2==3.1.6
+wheel==0.45.1
+referencing==0.37.0
+pandas==2.3.3
+soundfile==0.13.1
+pooch==1.8.2
+MarkupSafe==3.0.3
+dill==0.4.0
+pydantic-extra-types==2.10.6
+msgpack==1.1.2
+distro==1.9.0
+locate==1.1.1
+datasets==4.4.1
+Pygments==2.19.2
+aiohappyeyeballs==2.6.1
+llvmlite==0.46.0
+attrs==25.4.0
+huggingface_hub==1.3.5
+nltk==3.9.2
+torch==2.10.0
+httpx==0.28.1
+filelock==3.20.0
+smmap==5.0.2
+sounddevice==0.5.3
+timesfm==1.3.0
+pycountry==24.6.1
+mlx-metal==0.30.6
+scipy==1.15.3
+protobuf==6.33.1
+psutil==7.1.3
+typing-inspection==0.4.2
+joblib==1.5.2
+zipp==3.23.0
+annotated-types==0.7.0
+accelerate==1.12.0
+safetensors==0.6.2
+httpx-sse==0.4.3
+wcwidth==0.2.14
+igraph==1.0.0
+rpds-py==0.30.0
+langcodes==3.5.1
+six==1.17.0
+wandb==0.23.0
+yarl==1.22.0
+pyloudnorm==0.2.0
+xxhash==3.6.0
+matplotlib==3.10.7

adapters-conscious/wandb/run-20260216_150557-tugniqt7/files/wandb-metadata.json

@@ -0,0 +1,35 @@
+{
+  "os": "macOS-26.2-arm64-arm-64bit",
+  "python": "CPython 3.10.19",
+  "startedAt": "2026-02-16T20:05:57.689004Z",
+  "args": [
+    "--config",
+    "train.yaml"
+  ],
+  "program": "/Users/natebreslow/miniconda3/envs/mlx-experiment/bin/mlx_lm.lora",
+  "email": "nathanbreslow@gmail.com",
+  "root": "adapters-conscious",
+  "host": "MacBook-Pro-135.local",
+  "executable": "/Users/natebreslow/miniconda3/envs/mlx-experiment/bin/python3.10",
+  "cpu_count": 16,
+  "cpu_count_logical": 16,
+  "disk": {
+    "/": {
+      "total": "1995218165760",
+      "used": "1698983583744"
+    }
+  },
+  "memory": {
+    "total": "68719476736"
+  },
+  "apple": {
+    "name": "Apple M3 Max",
+    "ecpuCores": 4,
+    "pcpuCores": 12,
+    "gpuCores": 40,
+    "memoryGb": 64,
+    "swapTotalBytes": "2147483648",
+    "ramTotalBytes": "68719476736"
+  },
+  "writerId": "8z2fwr85sxkr7dsfa3tb7o36vuyb19ia"
+}

adapters-conscious/wandb/run-20260216_150557-tugniqt7/files/wandb-summary.json

@@ -0,0 +1 @@
+{"train_loss":1.746941089630127,"learning_rate":4.8740144848125055e-05,"_wandb":{"runtime":2080},"iteration":300,"iterations_per_second":0.2321287329053772,"_runtime":2080.795569,"val_time":0.007750792006845586,"_timestamp":1.77127443715999e+09,"peak_memory":12.008247198,"trained_tokens":270593,"val_loss":NaN,"tokens_per_second":175.48932207646516,"_step":300}

adapters-conscious/wandb/run-20260216_150557-tugniqt7/logs/debug-internal.log

@@ -0,0 +1,11 @@
+{"time":"2026-02-16T15:05:57.969219-05:00","level":"INFO","msg":"stream: starting","core version":"0.23.0"}
+{"time":"2026-02-16T15:05:58.195652-05:00","level":"INFO","msg":"stream: created new stream","id":"tugniqt7"}
+{"time":"2026-02-16T15:05:58.195747-05:00","level":"INFO","msg":"handler: started","stream_id":"tugniqt7"}
+{"time":"2026-02-16T15:05:58.195972-05:00","level":"INFO","msg":"stream: started","id":"tugniqt7"}
+{"time":"2026-02-16T15:05:58.195987-05:00","level":"INFO","msg":"sender: started","stream_id":"tugniqt7"}
+{"time":"2026-02-16T15:05:58.195991-05:00","level":"INFO","msg":"writer: started","stream_id":"tugniqt7"}
+{"time":"2026-02-16T15:40:39.241944-05:00","level":"INFO","msg":"stream: closing","id":"tugniqt7"}
+{"time":"2026-02-16T15:40:39.69538-05:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2026-02-16T15:40:40.025888-05:00","level":"INFO","msg":"handler: closed","stream_id":"tugniqt7"}
+{"time":"2026-02-16T15:40:40.026441-05:00","level":"INFO","msg":"sender: closed","stream_id":"tugniqt7"}
+{"time":"2026-02-16T15:40:40.026516-05:00","level":"INFO","msg":"stream: closed","id":"tugniqt7"}

adapters-conscious/wandb/run-20260216_150557-tugniqt7/logs/debug.log

@@ -0,0 +1,23 @@
+2026-02-16 15:05:57,692 INFO    MainThread:38471 [wandb_setup.py:_flush():80] Current SDK version is 0.23.0
+2026-02-16 15:05:57,693 INFO    MainThread:38471 [wandb_setup.py:_flush():80] Configure stats pid to 38471
+2026-02-16 15:05:57,693 INFO    MainThread:38471 [wandb_setup.py:_flush():80] Loading settings from /Users/natebreslow/.config/wandb/settings
+2026-02-16 15:05:57,693 INFO    MainThread:38471 [wandb_setup.py:_flush():80] Loading settings from /Users/natebreslow/Documents/llmSelfReport/wandb/settings
+2026-02-16 15:05:57,693 INFO    MainThread:38471 [wandb_setup.py:_flush():80] Loading settings from environment variables
+2026-02-16 15:05:57,693 INFO    MainThread:38471 [wandb_init.py:setup_run_log_directory():713] Logging user logs to adapters-conscious/wandb/run-20260216_150557-tugniqt7/logs/debug.log
+2026-02-16 15:05:57,693 INFO    MainThread:38471 [wandb_init.py:setup_run_log_directory():714] Logging internal logs to adapters-conscious/wandb/run-20260216_150557-tugniqt7/logs/debug-internal.log
+2026-02-16 15:05:57,693 INFO    MainThread:38471 [wandb_init.py:init():840] calling init triggers
+2026-02-16 15:05:57,693 INFO    MainThread:38471 [wandb_init.py:init():845] wandb.init called with sweep_config: {}
+config: {'model': 'Qwen3-4B-Instruct-2507', 'train': True, 'data': 'training/full-conscious', 'fine_tune_type': 'lora', 'optimizer': 'adam', 'mask_prompt': False, 'num_layers': 36, 'batch_size': 1, 'iters': 300, 'val_batches': 0, 'learning_rate': 0.0001, 'steps_per_report': 1, 'steps_per_eval': 200, 'grad_accumulation_steps': 1, 'resume_adapter_file': None, 'adapter_path': 'adapters-conscious', 'save_every': 100, 'test': False, 'test_batches': 100, 'max_seq_length': 8192, 'config': 'train.yaml', 'grad_checkpoint': True, 'report_to': 'wandb', 'project_name': 'conscious-finetuning', 'seed': 0, 'optimizer_config': {'adam': {'betas': [0.9, 0.9999], 'eps': 1e-06, 'bias_correction': True}}, 'lora_parameters': {'keys': ['self_attn.q_proj', 'self_attn.v_proj', 'self_attn.k_proj', 'self_attn.o_proj', 'mlp.gate_proj', 'mlp.up_proj', 'mlp.down_proj'], 'rank': 16, 'scale': 2.0, 'dropout': 0.0}, 'lr_schedule': {'name': 'cosine_decay', 'warmup': 10, 'warmup_init': 1e-05, 'arguments': [0.0001, 529, 1e-05]}, '_wandb': {}}
+2026-02-16 15:05:57,693 INFO    MainThread:38471 [wandb_init.py:init():888] starting backend
+2026-02-16 15:05:57,930 INFO    MainThread:38471 [wandb_init.py:init():891] sending inform_init request
+2026-02-16 15:05:57,968 INFO    MainThread:38471 [wandb_init.py:init():899] backend started and connected
+2026-02-16 15:05:57,971 INFO    MainThread:38471 [wandb_init.py:init():969] updated telemetry
+2026-02-16 15:05:57,971 INFO    MainThread:38471 [wandb_init.py:init():993] communicating run to backend with 90.0 second timeout
+2026-02-16 15:05:58,452 INFO    MainThread:38471 [wandb_init.py:init():1040] starting run threads in backend
+2026-02-16 15:05:58,553 INFO    MainThread:38471 [wandb_run.py:_console_start():2504] atexit reg
+2026-02-16 15:05:58,554 INFO    MainThread:38471 [wandb_run.py:_redirect():2352] redirect: wrap_raw
+2026-02-16 15:05:58,554 INFO    MainThread:38471 [wandb_run.py:_redirect():2421] Wrapping output streams.
+2026-02-16 15:05:58,554 INFO    MainThread:38471 [wandb_run.py:_redirect():2444] Redirects installed.
+2026-02-16 15:05:58,556 INFO    MainThread:38471 [wandb_init.py:init():1080] run started, returning control to user process
+2026-02-16 15:40:39,239 INFO    wandb-AsyncioManager-main:38471 [service_client.py:_forward_responses():80] Reached EOF.
+2026-02-16 15:40:39,239 INFO    wandb-AsyncioManager-main:38471 [mailbox.py:close():137] Closing mailbox, abandoning 1 handles.

adapters-conscious/wandb/run-20260216_150557-tugniqt7/run-tugniqt7.wandb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9786077aeddfd62bdab85e444abafced52584ec58009a3ee2ff494fe94674541
+size 430350

adapters-no-conscious/0000100_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa2537783d1965a5bf4ff5f42a0f29956e0e6c96ec725d65a0b1acbf05e8aba9
+size 132175803

adapters-no-conscious/0000200_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:62593a584e88cede30b1c6b9d968e8bc78ab89b07c3b8678b850c9bbf977c41d
+size 132175803

adapters-no-conscious/0000300_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b244118c15173df07a5b29dcc31f2226e860544f25757b81639d3bce4a4ee2ce
+size 132175803

adapters-no-conscious/adapter_config.json

@@ -0,0 +1,61 @@
+{
+    "adapter_path": "adapters-no-conscious",
+    "batch_size": 1,
+    "config": "train.yaml",
+    "data": "training/no-conscious",
+    "fine_tune_type": "lora",
+    "grad_accumulation_steps": 1,
+    "grad_checkpoint": true,
+    "iters": 300,
+    "learning_rate": 0.0001,
+    "lora_parameters": {
+        "keys": [
+            "self_attn.q_proj",
+            "self_attn.v_proj",
+            "self_attn.k_proj",
+            "self_attn.o_proj",
+            "mlp.gate_proj",
+            "mlp.up_proj",
+            "mlp.down_proj"
+        ],
+        "rank": 16,
+        "scale": 2.0,
+        "dropout": 0.0
+    },
+    "lr_schedule": {
+        "name": "cosine_decay",
+        "warmup": 10,
+        "warmup_init": 1e-05,
+        "arguments": [
+            0.0001,
+            529,
+            1e-05
+        ]
+    },
+    "mask_prompt": false,
+    "max_seq_length": 8192,
+    "model": "Qwen3-4B-Instruct-2507",
+    "num_layers": 36,
+    "optimizer": "adam",
+    "optimizer_config": {
+        "adam": {
+            "betas": [
+                0.9,
+                0.9999
+            ],
+            "eps": 1e-06,
+            "bias_correction": true
+        }
+    },
+    "project_name": "conscious-finetuning",
+    "report_to": "wandb",
+    "resume_adapter_file": null,
+    "save_every": 100,
+    "seed": 0,
+    "steps_per_eval": 200,
+    "steps_per_report": 1,
+    "test": false,
+    "test_batches": 100,
+    "train": true,
+    "val_batches": 0
+}

adapters-no-conscious/adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b244118c15173df07a5b29dcc31f2226e860544f25757b81639d3bce4a4ee2ce
+size 132175803

adapters-no-conscious/wandb/run-20260216_162928-t3h29alg/files/config.yaml

@@ -0,0 +1,141 @@
+_wandb:
+    value:
+        cli_version: 0.23.0
+        e:
+            1pxa79orri10yjkyhqiblp74t2dr1h2q:
+                apple:
+                    ecpuCores: 4
+                    gpuCores: 40
+                    memoryGb: 64
+                    name: Apple M3 Max
+                    pcpuCores: 12
+                    ramTotalBytes: "68719476736"
+                    swapTotalBytes: "5368709120"
+                args:
+                    - --config
+                    - train.yaml
+                cpu_count: 16
+                cpu_count_logical: 16
+                disk:
+                    /:
+                        total: "1995218165760"
+                        used: "1703338582016"
+                email: nathanbreslow@gmail.com
+                executable: /Users/natebreslow/miniconda3/envs/mlx-experiment/bin/python3.10
+                host: MacBook-Pro-135.local
+                memory:
+                    total: "68719476736"
+                os: macOS-26.2-arm64-arm-64bit
+                program: /Users/natebreslow/miniconda3/envs/mlx-experiment/bin/mlx_lm.lora
+                python: CPython 3.10.19
+                root: adapters-no-conscious
+                startedAt: "2026-02-16T21:29:28.031847Z"
+                writerId: 1pxa79orri10yjkyhqiblp74t2dr1h2q
+        m: []
+        python_version: 3.10.19
+        t:
+            "1":
+                - 1
+                - 5
+                - 11
+                - 49
+                - 53
+                - 71
+            "2":
+                - 1
+                - 5
+                - 11
+                - 49
+                - 53
+                - 71
+            "3":
+                - 13
+                - 16
+                - 61
+            "4": 3.10.19
+            "5": 0.23.0
+            "6": 5.1.0
+            "8":
+                - 2
+            "12": 0.23.0
+            "13": darwin-arm64
+adapter_path:
+    value: adapters-no-conscious
+batch_size:
+    value: 1
+config:
+    value: train.yaml
+data:
+    value: training/no-conscious
+fine_tune_type:
+    value: lora
+grad_accumulation_steps:
+    value: 1
+grad_checkpoint:
+    value: true
+iters:
+    value: 300
+learning_rate:
+    value: 0.0001
+lora_parameters:
+    value:
+        dropout: 0
+        keys:
+            - self_attn.q_proj
+            - self_attn.v_proj
+            - self_attn.k_proj
+            - self_attn.o_proj
+            - mlp.gate_proj
+            - mlp.up_proj
+            - mlp.down_proj
+        rank: 16
+        scale: 2
+lr_schedule:
+    value:
+        arguments:
+            - 0.0001
+            - 529
+            - 1e-05
+        name: cosine_decay
+        warmup: 10
+        warmup_init: 1e-05
+mask_prompt:
+    value: false
+max_seq_length:
+    value: 8192
+model:
+    value: Qwen3-4B-Instruct-2507
+num_layers:
+    value: 36
+optimizer:
+    value: adam
+optimizer_config:
+    value:
+        adam:
+            betas:
+                - 0.9
+                - 0.9999
+            bias_correction: true
+            eps: 1e-06
+project_name:
+    value: conscious-finetuning
+report_to:
+    value: wandb
+resume_adapter_file:
+    value: null
+save_every:
+    value: 100
+seed:
+    value: 0
+steps_per_eval:
+    value: 200
+steps_per_report:
+    value: 1
+test:
+    value: false
+test_batches:
+    value: 100
+train:
+    value: true
+val_batches:
+    value: 0

adapters-no-conscious/wandb/run-20260216_162928-t3h29alg/files/output.log

@@ -0,0 +1,315 @@
+Loading pretrained model
+Loading datasets
+Training
+Trainable parameters: 0.821% (33.030M/4022.468M)
+Starting training..., iters: 300
+Calculating loss...: 0it [00:00, ?it/s]
+Iter 1: Val loss nan, Val took 0.092s
+Iter 1: Train loss 3.266, Learning Rate 1.000e-05, It/sec 0.219, Tokens/sec 219.918, Trained Tokens 1005, Peak mem 9.938 GB
+Iter 2: Train loss 3.350, Learning Rate 1.900e-05, It/sec 0.240, Tokens/sec 248.127, Trained Tokens 2038, Peak mem 10.306 GB
+Iter 3: Train loss 3.355, Learning Rate 2.800e-05, It/sec 0.346, Tokens/sec 263.123, Trained Tokens 2799, Peak mem 10.306 GB
+Iter 4: Train loss 3.106, Learning Rate 3.700e-05, It/sec 0.238, Tokens/sec 254.304, Trained Tokens 3867, Peak mem 10.362 GB
+Iter 5: Train loss 3.181, Learning Rate 4.600e-05, It/sec 0.251, Tokens/sec 263.035, Trained Tokens 4915, Peak mem 10.362 GB
+Iter 6: Train loss 3.010, Learning Rate 5.500e-05, It/sec 0.400, Tokens/sec 245.998, Trained Tokens 5530, Peak mem 10.362 GB
+Iter 7: Train loss 2.316, Learning Rate 6.400e-05, It/sec 0.673, Tokens/sec 254.437, Trained Tokens 5908, Peak mem 10.362 GB
+Iter 8: Train loss 2.817, Learning Rate 7.300e-05, It/sec 0.313, Tokens/sec 252.434, Trained Tokens 6715, Peak mem 10.362 GB
+Iter 9: Train loss 2.591, Learning Rate 8.200e-05, It/sec 0.351, Tokens/sec 262.384, Trained Tokens 7463, Peak mem 10.362 GB
+Iter 10: Train loss 2.290, Learning Rate 9.100e-05, It/sec 0.281, Tokens/sec 243.767, Trained Tokens 8332, Peak mem 10.362 GB
+Iter 11: Train loss 2.474, Learning Rate 1.000e-04, It/sec 0.668, Tokens/sec 224.391, Trained Tokens 8668, Peak mem 10.362 GB
+Iter 12: Train loss 2.072, Learning Rate 1.000e-04, It/sec 0.260, Tokens/sec 235.524, Trained Tokens 9574, Peak mem 10.362 GB
+Iter 13: Train loss 2.235, Learning Rate 1.000e-04, It/sec 0.416, Tokens/sec 255.720, Trained Tokens 10188, Peak mem 10.362 GB
+Iter 14: Train loss 2.274, Learning Rate 1.000e-04, It/sec 0.575, Tokens/sec 234.092, Trained Tokens 10595, Peak mem 10.362 GB
+Iter 15: Train loss 2.300, Learning Rate 9.999e-05, It/sec 0.615, Tokens/sec 250.212, Trained Tokens 11002, Peak mem 10.362 GB
+Iter 16: Train loss 2.003, Learning Rate 9.999e-05, It/sec 0.263, Tokens/sec 250.729, Trained Tokens 11956, Peak mem 10.362 GB
+Iter 17: Train loss 2.013, Learning Rate 9.998e-05, It/sec 0.325, Tokens/sec 266.512, Trained Tokens 12777, Peak mem 10.362 GB
+Iter 18: Train loss 2.148, Learning Rate 9.997e-05, It/sec 0.220, Tokens/sec 246.550, Trained Tokens 13900, Peak mem 10.382 GB
+Iter 19: Train loss 2.353, Learning Rate 9.996e-05, It/sec 0.215, Tokens/sec 236.448, Trained Tokens 14998, Peak mem 10.382 GB
+Iter 20: Train loss 2.066, Learning Rate 9.995e-05, It/sec 0.237, Tokens/sec 227.481, Trained Tokens 15956, Peak mem 10.382 GB
+Iter 21: Train loss 2.191, Learning Rate 9.994e-05, It/sec 0.366, Tokens/sec 240.958, Trained Tokens 16614, Peak mem 10.382 GB
+Iter 22: Train loss 2.276, Learning Rate 9.992e-05, It/sec 0.422, Tokens/sec 259.892, Trained Tokens 17230, Peak mem 10.382 GB
+Iter 23: Train loss 2.309, Learning Rate 9.990e-05, It/sec 0.400, Tokens/sec 226.283, Trained Tokens 17796, Peak mem 10.382 GB
+Iter 24: Train loss 1.981, Learning Rate 9.989e-05, It/sec 0.459, Tokens/sec 250.640, Trained Tokens 18342, Peak mem 10.382 GB
+Iter 25: Train loss 2.140, Learning Rate 9.987e-05, It/sec 0.157, Tokens/sec 234.702, Trained Tokens 19834, Peak mem 10.892 GB
+Iter 26: Train loss 2.125, Learning Rate 9.984e-05, It/sec 0.421, Tokens/sec 241.228, Trained Tokens 20407, Peak mem 10.892 GB
+Iter 27: Train loss 2.225, Learning Rate 9.982e-05, It/sec 0.410, Tokens/sec 217.909, Trained Tokens 20939, Peak mem 10.892 GB
+Iter 28: Train loss 2.043, Learning Rate 9.980e-05, It/sec 0.712, Tokens/sec 267.674, Trained Tokens 21315, Peak mem 10.892 GB
+Iter 29: Train loss 2.099, Learning Rate 9.977e-05, It/sec 0.287, Tokens/sec 241.418, Trained Tokens 22156, Peak mem 10.892 GB
+Iter 30: Train loss 1.822, Learning Rate 9.974e-05, It/sec 0.296, Tokens/sec 216.552, Trained Tokens 22888, Peak mem 10.892 GB
+Iter 31: Train loss 1.840, Learning Rate 9.971e-05, It/sec 0.240, Tokens/sec 259.414, Trained Tokens 23967, Peak mem 10.892 GB
+Iter 32: Train loss 2.023, Learning Rate 9.968e-05, It/sec 0.147, Tokens/sec 232.849, Trained Tokens 25555, Peak mem 11.056 GB
+Iter 33: Train loss 1.860, Learning Rate 9.965e-05, It/sec 0.191, Tokens/sec 242.945, Trained Tokens 26829, Peak mem 11.056 GB
+Iter 34: Train loss 1.762, Learning Rate 9.962e-05, It/sec 0.324, Tokens/sec 239.890, Trained Tokens 27569, Peak mem 11.056 GB
+Iter 35: Train loss 1.825, Learning Rate 9.958e-05, It/sec 0.192, Tokens/sec 234.307, Trained Tokens 28789, Peak mem 11.056 GB
+Iter 36: Train loss 2.026, Learning Rate 9.954e-05, It/sec 0.468, Tokens/sec 256.133, Trained Tokens 29336, Peak mem 11.056 GB
+Iter 37: Train loss 2.275, Learning Rate 9.950e-05, It/sec 0.318, Tokens/sec 236.437, Trained Tokens 30079, Peak mem 11.056 GB
+Iter 38: Train loss 2.144, Learning Rate 9.946e-05, It/sec 0.286, Tokens/sec 240.817, Trained Tokens 30921, Peak mem 11.056 GB
+Iter 39: Train loss 1.768, Learning Rate 9.942e-05, It/sec 0.200, Tokens/sec 246.751, Trained Tokens 32155, Peak mem 11.056 GB
+Iter 40: Train loss 1.752, Learning Rate 9.938e-05, It/sec 0.391, Tokens/sec 230.057, Trained Tokens 32743, Peak mem 11.056 GB
+Iter 41: Train loss 1.954, Learning Rate 9.933e-05, It/sec 0.468, Tokens/sec 256.987, Trained Tokens 33292, Peak mem 11.056 GB
+Iter 42: Train loss 2.077, Learning Rate 9.929e-05, It/sec 0.380, Tokens/sec 250.363, Trained Tokens 33950, Peak mem 11.056 GB
+Iter 43: Train loss 1.874, Learning Rate 9.924e-05, It/sec 0.419, Tokens/sec 223.902, Trained Tokens 34485, Peak mem 11.056 GB
+Iter 44: Train loss 1.515, Learning Rate 9.919e-05, It/sec 0.656, Tokens/sec 242.694, Trained Tokens 34855, Peak mem 11.056 GB
+Iter 45: Train loss 1.799, Learning Rate 9.914e-05, It/sec 0.170, Tokens/sec 231.035, Trained Tokens 36217, Peak mem 11.056 GB
+Iter 46: Train loss 1.916, Learning Rate 9.909e-05, It/sec 0.261, Tokens/sec 230.036, Trained Tokens 37097, Peak mem 11.056 GB
+Iter 47: Train loss 1.826, Learning Rate 9.903e-05, It/sec 0.227, Tokens/sec 246.707, Trained Tokens 38182, Peak mem 11.056 GB
+Iter 48: Train loss 1.712, Learning Rate 9.898e-05, It/sec 0.591, Tokens/sec 234.633, Trained Tokens 38579, Peak mem 11.056 GB
+Iter 49: Train loss 1.607, Learning Rate 9.892e-05, It/sec 0.261, Tokens/sec 236.786, Trained Tokens 39487, Peak mem 11.056 GB
+Iter 50: Train loss 1.877, Learning Rate 9.886e-05, It/sec 0.171, Tokens/sec 233.724, Trained Tokens 40854, Peak mem 11.056 GB
+Iter 51: Train loss 1.789, Learning Rate 9.880e-05, It/sec 0.272, Tokens/sec 240.953, Trained Tokens 41740, Peak mem 11.056 GB
+Iter 52: Train loss 1.733, Learning Rate 9.874e-05, It/sec 0.321, Tokens/sec 243.384, Trained Tokens 42498, Peak mem 11.056 GB
+Iter 53: Train loss 1.754, Learning Rate 9.867e-05, It/sec 0.350, Tokens/sec 239.676, Trained Tokens 43183, Peak mem 11.056 GB
+Iter 54: Train loss 1.602, Learning Rate 9.861e-05, It/sec 0.216, Tokens/sec 229.402, Trained Tokens 44243, Peak mem 11.056 GB
+Iter 55: Train loss 1.551, Learning Rate 9.854e-05, It/sec 0.378, Tokens/sec 224.393, Trained Tokens 44837, Peak mem 11.056 GB
+Iter 56: Train loss 1.662, Learning Rate 9.847e-05, It/sec 0.197, Tokens/sec 234.022, Trained Tokens 46027, Peak mem 11.056 GB
+Iter 57: Train loss 1.501, Learning Rate 9.840e-05, It/sec 0.412, Tokens/sec 239.261, Trained Tokens 46608, Peak mem 11.056 GB
+Iter 58: Train loss 1.967, Learning Rate 9.833e-05, It/sec 0.140, Tokens/sec 223.884, Trained Tokens 48212, Peak mem 11.112 GB
+Iter 59: Train loss 1.721, Learning Rate 9.826e-05, It/sec 0.157, Tokens/sec 209.558, Trained Tokens 49545, Peak mem 11.112 GB
+Iter 60: Train loss 1.595, Learning Rate 9.818e-05, It/sec 0.327, Tokens/sec 242.664, Trained Tokens 50286, Peak mem 11.112 GB
+Iter 61: Train loss 1.616, Learning Rate 9.811e-05, It/sec 0.201, Tokens/sec 220.165, Trained Tokens 51379, Peak mem 11.112 GB
+Iter 62: Train loss 1.658, Learning Rate 9.803e-05, It/sec 0.220, Tokens/sec 243.288, Trained Tokens 52484, Peak mem 11.112 GB
+Iter 63: Train loss 1.721, Learning Rate 9.795e-05, It/sec 0.291, Tokens/sec 200.028, Trained Tokens 53172, Peak mem 11.112 GB
+Iter 64: Train loss 1.495, Learning Rate 9.787e-05, It/sec 0.231, Tokens/sec 194.516, Trained Tokens 54014, Peak mem 11.112 GB
+Iter 65: Train loss 1.803, Learning Rate 9.779e-05, It/sec 0.247, Tokens/sec 184.196, Trained Tokens 54759, Peak mem 11.112 GB
+Iter 66: Train loss 1.587, Learning Rate 9.771e-05, It/sec 0.148, Tokens/sec 141.844, Trained Tokens 55715, Peak mem 11.112 GB
+Iter 67: Train loss 1.564, Learning Rate 9.762e-05, It/sec 0.075, Tokens/sec 87.031, Trained Tokens 56882, Peak mem 11.112 GB
+Iter 68: Train loss 1.527, Learning Rate 9.753e-05, It/sec 0.043, Tokens/sec 27.505, Trained Tokens 57527, Peak mem 11.112 GB
+Iter 69: Train loss 2.127, Learning Rate 9.745e-05, It/sec 0.092, Tokens/sec 32.925, Trained Tokens 57883, Peak mem 11.112 GB
+Iter 70: Train loss 1.562, Learning Rate 9.736e-05, It/sec 0.094, Tokens/sec 31.632, Trained Tokens 58218, Peak mem 11.112 GB
+Iter 71: Train loss 1.813, Learning Rate 9.727e-05, It/sec 0.043, Tokens/sec 19.349, Trained Tokens 58670, Peak mem 11.112 GB
+Iter 72: Train loss 1.748, Learning Rate 9.717e-05, It/sec 0.030, Tokens/sec 37.120, Trained Tokens 59898, Peak mem 11.112 GB
+Iter 73: Train loss 1.681, Learning Rate 9.708e-05, It/sec 0.122, Tokens/sec 47.251, Trained Tokens 60286, Peak mem 11.112 GB
+Iter 74: Train loss 1.455, Learning Rate 9.698e-05, It/sec 0.066, Tokens/sec 57.059, Trained Tokens 61149, Peak mem 11.112 GB
+Iter 75: Train loss 1.722, Learning Rate 9.689e-05, It/sec 0.115, Tokens/sec 49.134, Trained Tokens 61576, Peak mem 11.112 GB
+Iter 76: Train loss 1.652, Learning Rate 9.679e-05, It/sec 0.110, Tokens/sec 50.283, Trained Tokens 62034, Peak mem 11.112 GB
+Iter 77: Train loss 1.658, Learning Rate 9.669e-05, It/sec 0.100, Tokens/sec 85.424, Trained Tokens 62885, Peak mem 11.112 GB
+Iter 78: Train loss 1.717, Learning Rate 9.659e-05, It/sec 0.207, Tokens/sec 94.479, Trained Tokens 63341, Peak mem 11.112 GB
+Iter 79: Train loss 1.588, Learning Rate 9.648e-05, It/sec 0.078, Tokens/sec 89.893, Trained Tokens 64491, Peak mem 11.112 GB
+Iter 80: Train loss 1.868, Learning Rate 9.638e-05, It/sec 0.187, Tokens/sec 122.665, Trained Tokens 65147, Peak mem 11.112 GB
+Iter 81: Train loss 1.517, Learning Rate 9.627e-05, It/sec 0.204, Tokens/sec 77.216, Trained Tokens 65526, Peak mem 11.112 GB
+Iter 82: Train loss 1.905, Learning Rate 9.617e-05, It/sec 0.213, Tokens/sec 101.348, Trained Tokens 66001, Peak mem 11.112 GB
+Iter 83: Train loss 1.599, Learning Rate 9.606e-05, It/sec 0.241, Tokens/sec 132.157, Trained Tokens 66550, Peak mem 11.112 GB
+Iter 84: Train loss 1.636, Learning Rate 9.595e-05, It/sec 0.187, Tokens/sec 129.104, Trained Tokens 67242, Peak mem 11.112 GB
+Iter 85: Train loss 1.817, Learning Rate 9.584e-05, It/sec 0.146, Tokens/sec 146.165, Trained Tokens 68246, Peak mem 11.112 GB
+Iter 86: Train loss 1.615, Learning Rate 9.572e-05, It/sec 0.207, Tokens/sec 98.525, Trained Tokens 68722, Peak mem 11.112 GB
+Iter 87: Train loss 1.647, Learning Rate 9.561e-05, It/sec 0.243, Tokens/sec 96.177, Trained Tokens 69118, Peak mem 11.112 GB
+Iter 88: Train loss 1.945, Learning Rate 9.549e-05, It/sec 0.206, Tokens/sec 145.374, Trained Tokens 69822, Peak mem 11.112 GB
+Iter 89: Train loss 1.672, Learning Rate 9.538e-05, It/sec 0.122, Tokens/sec 149.662, Trained Tokens 71049, Peak mem 11.112 GB
+Iter 90: Train loss 1.880, Learning Rate 9.526e-05, It/sec 0.099, Tokens/sec 157.787, Trained Tokens 72645, Peak mem 11.112 GB
+Iter 91: Train loss 1.622, Learning Rate 9.514e-05, It/sec 0.162, Tokens/sec 157.416, Trained Tokens 73617, Peak mem 11.112 GB
+Iter 92: Train loss 1.917, Learning Rate 9.502e-05, It/sec 0.112, Tokens/sec 130.533, Trained Tokens 74779, Peak mem 11.112 GB
+Iter 93: Train loss 1.387, Learning Rate 9.489e-05, It/sec 0.255, Tokens/sec 149.822, Trained Tokens 75367, Peak mem 11.112 GB
+Iter 94: Train loss 1.678, Learning Rate 9.477e-05, It/sec 0.292, Tokens/sec 180.344, Trained Tokens 75985, Peak mem 11.112 GB
+Iter 95: Train loss 1.899, Learning Rate 9.464e-05, It/sec 0.120, Tokens/sec 143.552, Trained Tokens 77186, Peak mem 11.112 GB
+Iter 96: Train loss 1.741, Learning Rate 9.452e-05, It/sec 0.131, Tokens/sec 185.357, Trained Tokens 78598, Peak mem 11.112 GB
+Iter 97: Train loss 1.723, Learning Rate 9.439e-05, It/sec 0.347, Tokens/sec 201.097, Trained Tokens 79178, Peak mem 11.112 GB
+Iter 98: Train loss 1.628, Learning Rate 9.426e-05, It/sec 0.527, Tokens/sec 208.151, Trained Tokens 79573, Peak mem 11.112 GB
+Iter 99: Train loss 1.665, Learning Rate 9.413e-05, It/sec 0.206, Tokens/sec 190.142, Trained Tokens 80496, Peak mem 11.112 GB
+Iter 100: Train loss 1.576, Learning Rate 9.399e-05, It/sec 0.163, Tokens/sec 186.258, Trained Tokens 81641, Peak mem 11.112 GB
+Iter 100: Saved adapter weights to adapters-no-conscious/adapters.safetensors and adapters-no-conscious/0000100_adapters.safetensors.
+Iter 101: Train loss 1.577, Learning Rate 9.386e-05, It/sec 0.233, Tokens/sec 182.641, Trained Tokens 82424, Peak mem 11.112 GB
+Iter 102: Train loss 1.413, Learning Rate 9.372e-05, It/sec 0.253, Tokens/sec 159.937, Trained Tokens 83056, Peak mem 11.112 GB
+Iter 103: Train loss 1.551, Learning Rate 9.359e-05, It/sec 0.342, Tokens/sec 212.738, Trained Tokens 83678, Peak mem 11.112 GB
+Iter 104: Train loss 1.429, Learning Rate 9.345e-05, It/sec 0.269, Tokens/sec 199.412, Trained Tokens 84420, Peak mem 11.112 GB
+Iter 105: Train loss 1.795, Learning Rate 9.331e-05, It/sec 0.095, Tokens/sec 167.315, Trained Tokens 86183, Peak mem 11.523 GB
+Iter 106: Train loss 1.834, Learning Rate 9.317e-05, It/sec 0.152, Tokens/sec 186.369, Trained Tokens 87407, Peak mem 11.523 GB
+Iter 107: Train loss 1.867, Learning Rate 9.303e-05, It/sec 0.153, Tokens/sec 170.852, Trained Tokens 88521, Peak mem 11.523 GB
+Iter 108: Train loss 1.496, Learning Rate 9.288e-05, It/sec 0.274, Tokens/sec 200.490, Trained Tokens 89254, Peak mem 11.523 GB
+Iter 109: Train loss 1.399, Learning Rate 9.274e-05, It/sec 0.295, Tokens/sec 211.696, Trained Tokens 89971, Peak mem 11.523 GB
+Iter 110: Train loss 1.453, Learning Rate 9.259e-05, It/sec 0.218, Tokens/sec 204.588, Trained Tokens 90908, Peak mem 11.523 GB
+Iter 111: Train loss 1.571, Learning Rate 9.244e-05, It/sec 0.212, Tokens/sec 230.253, Trained Tokens 91994, Peak mem 11.523 GB
+Iter 112: Train loss 1.840, Learning Rate 9.230e-05, It/sec 0.145, Tokens/sec 214.957, Trained Tokens 93478, Peak mem 11.523 GB
+Iter 113: Train loss 1.891, Learning Rate 9.214e-05, It/sec 0.341, Tokens/sec 238.520, Trained Tokens 94178, Peak mem 11.523 GB
+Iter 114: Train loss 1.726, Learning Rate 9.199e-05, It/sec 0.428, Tokens/sec 227.081, Trained Tokens 94709, Peak mem 11.523 GB
+Iter 115: Train loss 1.664, Learning Rate 9.184e-05, It/sec 0.387, Tokens/sec 235.376, Trained Tokens 95317, Peak mem 11.523 GB
+Iter 116: Train loss 1.882, Learning Rate 9.169e-05, It/sec 0.330, Tokens/sec 220.575, Trained Tokens 95985, Peak mem 11.523 GB
+Iter 117: Train loss 1.673, Learning Rate 9.153e-05, It/sec 0.252, Tokens/sec 227.758, Trained Tokens 96890, Peak mem 11.523 GB
+Iter 118: Train loss 1.620, Learning Rate 9.137e-05, It/sec 0.249, Tokens/sec 227.021, Trained Tokens 97801, Peak mem 11.523 GB
+Iter 119: Train loss 1.758, Learning Rate 9.122e-05, It/sec 0.294, Tokens/sec 233.443, Trained Tokens 98594, Peak mem 11.523 GB
+Iter 120: Train loss 1.349, Learning Rate 9.106e-05, It/sec 0.390, Tokens/sec 236.422, Trained Tokens 99200, Peak mem 11.523 GB
+Iter 121: Train loss 1.705, Learning Rate 9.090e-05, It/sec 0.182, Tokens/sec 227.278, Trained Tokens 100447, Peak mem 11.523 GB
+Iter 122: Train loss 1.416, Learning Rate 9.073e-05, It/sec 0.218, Tokens/sec 228.948, Trained Tokens 101498, Peak mem 11.523 GB
+Iter 123: Train loss 1.724, Learning Rate 9.057e-05, It/sec 0.133, Tokens/sec 206.255, Trained Tokens 103047, Peak mem 11.523 GB
+Iter 124: Train loss 1.546, Learning Rate 9.041e-05, It/sec 0.202, Tokens/sec 203.202, Trained Tokens 104055, Peak mem 11.523 GB
+Iter 125: Train loss 1.390, Learning Rate 9.024e-05, It/sec 0.190, Tokens/sec 213.422, Trained Tokens 105181, Peak mem 11.523 GB
+Iter 126: Train loss 1.609, Learning Rate 9.008e-05, It/sec 0.245, Tokens/sec 206.548, Trained Tokens 106024, Peak mem 11.523 GB
+Iter 127: Train loss 1.376, Learning Rate 8.991e-05, It/sec 0.386, Tokens/sec 208.927, Trained Tokens 106565, Peak mem 11.523 GB
+Iter 128: Train loss 1.646, Learning Rate 8.974e-05, It/sec 0.151, Tokens/sec 214.073, Trained Tokens 107987, Peak mem 11.523 GB
+Iter 129: Train loss 1.422, Learning Rate 8.957e-05, It/sec 0.691, Tokens/sec 192.186, Trained Tokens 108265, Peak mem 11.523 GB
+Iter 130: Train loss 1.371, Learning Rate 8.940e-05, It/sec 0.397, Tokens/sec 216.587, Trained Tokens 108810, Peak mem 11.523 GB
+Iter 131: Train loss 1.755, Learning Rate 8.922e-05, It/sec 0.377, Tokens/sec 220.674, Trained Tokens 109396, Peak mem 11.523 GB
+Iter 132: Train loss 1.267, Learning Rate 8.905e-05, It/sec 0.469, Tokens/sec 222.334, Trained Tokens 109870, Peak mem 11.523 GB
+Iter 133: Train loss 1.179, Learning Rate 8.887e-05, It/sec 0.332, Tokens/sec 230.046, Trained Tokens 110562, Peak mem 11.523 GB
+Iter 134: Train loss 1.440, Learning Rate 8.870e-05, It/sec 0.194, Tokens/sec 220.979, Trained Tokens 111701, Peak mem 11.523 GB
+Iter 135: Train loss 1.830, Learning Rate 8.852e-05, It/sec 0.407, Tokens/sec 224.426, Trained Tokens 112253, Peak mem 11.523 GB
+Iter 136: Train loss 1.675, Learning Rate 8.834e-05, It/sec 0.216, Tokens/sec 216.451, Trained Tokens 113253, Peak mem 11.523 GB
+Iter 137: Train loss 1.743, Learning Rate 8.816e-05, It/sec 0.271, Tokens/sec 219.595, Trained Tokens 114063, Peak mem 11.523 GB
+Iter 138: Train loss 1.506, Learning Rate 8.798e-05, It/sec 0.415, Tokens/sec 200.878, Trained Tokens 114547, Peak mem 11.523 GB
+Iter 139: Train loss 1.261, Learning Rate 8.780e-05, It/sec 0.605, Tokens/sec 175.593, Trained Tokens 114837, Peak mem 11.523 GB
+Iter 140: Train loss 1.812, Learning Rate 8.761e-05, It/sec 0.370, Tokens/sec 202.121, Trained Tokens 115383, Peak mem 11.523 GB
+Iter 141: Train loss 1.809, Learning Rate 8.743e-05, It/sec 0.148, Tokens/sec 195.010, Trained Tokens 116699, Peak mem 11.523 GB
+Iter 142: Train loss 1.356, Learning Rate 8.724e-05, It/sec 0.205, Tokens/sec 191.224, Trained Tokens 117633, Peak mem 11.523 GB
+Iter 143: Train loss 1.501, Learning Rate 8.706e-05, It/sec 0.350, Tokens/sec 197.990, Trained Tokens 118198, Peak mem 11.523 GB
+Iter 144: Train loss 1.337, Learning Rate 8.687e-05, It/sec 0.352, Tokens/sec 182.877, Trained Tokens 118717, Peak mem 11.523 GB
+Iter 145: Train loss 1.760, Learning Rate 8.668e-05, It/sec 0.222, Tokens/sec 170.355, Trained Tokens 119485, Peak mem 11.523 GB
+Iter 146: Train loss 1.256, Learning Rate 8.649e-05, It/sec 0.281, Tokens/sec 154.556, Trained Tokens 120036, Peak mem 11.523 GB
+Iter 147: Train loss 1.573, Learning Rate 8.630e-05, It/sec 0.098, Tokens/sec 122.408, Trained Tokens 121290, Peak mem 11.523 GB
+Iter 148: Train loss 1.631, Learning Rate 8.610e-05, It/sec 0.104, Tokens/sec 106.299, Trained Tokens 122310, Peak mem 11.523 GB
+Iter 149: Train loss 1.128, Learning Rate 8.591e-05, It/sec 0.216, Tokens/sec 105.823, Trained Tokens 122800, Peak mem 11.523 GB
+Iter 150: Train loss 1.545, Learning Rate 8.571e-05, It/sec 0.086, Tokens/sec 95.748, Trained Tokens 123913, Peak mem 11.523 GB
+Iter 151: Train loss 1.531, Learning Rate 8.552e-05, It/sec 0.055, Tokens/sec 69.971, Trained Tokens 125190, Peak mem 11.523 GB
+Iter 152: Train loss 1.321, Learning Rate 8.532e-05, It/sec 0.074, Tokens/sec 74.891, Trained Tokens 126198, Peak mem 11.523 GB
+Iter 153: Train loss 1.601, Learning Rate 8.512e-05, It/sec 0.035, Tokens/sec 43.266, Trained Tokens 127435, Peak mem 11.523 GB
+Iter 154: Train loss 1.501, Learning Rate 8.493e-05, It/sec 0.031, Tokens/sec 33.931, Trained Tokens 128531, Peak mem 11.523 GB
+Iter 155: Train loss 1.503, Learning Rate 8.472e-05, It/sec 0.033, Tokens/sec 23.544, Trained Tokens 129240, Peak mem 11.523 GB
+Iter 156: Train loss 1.382, Learning Rate 8.452e-05, It/sec 0.037, Tokens/sec 18.347, Trained Tokens 129732, Peak mem 11.523 GB
+Iter 157: Train loss 1.496, Learning Rate 8.432e-05, It/sec 0.050, Tokens/sec 34.758, Trained Tokens 130423, Peak mem 11.523 GB
+Iter 158: Train loss 1.478, Learning Rate 8.412e-05, It/sec 0.127, Tokens/sec 50.982, Trained Tokens 130825, Peak mem 11.523 GB
+Iter 159: Train loss 1.377, Learning Rate 8.391e-05, It/sec 0.056, Tokens/sec 45.133, Trained Tokens 131630, Peak mem 11.523 GB
+Iter 160: Train loss 1.693, Learning Rate 8.371e-05, It/sec 0.041, Tokens/sec 70.636, Trained Tokens 133352, Peak mem 11.523 GB
+Iter 161: Train loss 1.479, Learning Rate 8.350e-05, It/sec 0.087, Tokens/sec 72.996, Trained Tokens 134187, Peak mem 11.523 GB
+Iter 162: Train loss 1.462, Learning Rate 8.330e-05, It/sec 0.075, Tokens/sec 81.775, Trained Tokens 135278, Peak mem 11.523 GB
+Iter 163: Train loss 1.426, Learning Rate 8.309e-05, It/sec 0.172, Tokens/sec 95.708, Trained Tokens 135833, Peak mem 11.523 GB
+Iter 164: Train loss 1.594, Learning Rate 8.288e-05, It/sec 0.294, Tokens/sec 133.961, Trained Tokens 136289, Peak mem 11.523 GB
+Iter 165: Train loss 1.490, Learning Rate 8.267e-05, It/sec 0.110, Tokens/sec 107.939, Trained Tokens 137266, Peak mem 11.523 GB
+Iter 166: Train loss 1.313, Learning Rate 8.246e-05, It/sec 0.267, Tokens/sec 142.797, Trained Tokens 137800, Peak mem 11.523 GB
+Iter 167: Train loss 1.614, Learning Rate 8.224e-05, It/sec 0.138, Tokens/sec 154.017, Trained Tokens 138915, Peak mem 11.523 GB
+Iter 168: Train loss 1.413, Learning Rate 8.203e-05, It/sec 0.183, Tokens/sec 167.875, Trained Tokens 139831, Peak mem 11.523 GB
+Iter 169: Train loss 1.576, Learning Rate 8.182e-05, It/sec 0.176, Tokens/sec 180.507, Trained Tokens 140854, Peak mem 11.523 GB
+Iter 170: Train loss 1.408, Learning Rate 8.160e-05, It/sec 0.079, Tokens/sec 108.250, Trained Tokens 142228, Peak mem 11.523 GB
+Iter 171: Train loss 1.378, Learning Rate 8.139e-05, It/sec 0.206, Tokens/sec 112.317, Trained Tokens 142772, Peak mem 11.523 GB
+Iter 172: Train loss 1.396, Learning Rate 8.117e-05, It/sec 0.211, Tokens/sec 102.739, Trained Tokens 143259, Peak mem 11.523 GB
+Iter 173: Train loss 1.737, Learning Rate 8.095e-05, It/sec 0.116, Tokens/sec 167.476, Trained Tokens 144709, Peak mem 11.523 GB
+Iter 174: Train loss 1.353, Learning Rate 8.073e-05, It/sec 0.307, Tokens/sec 207.220, Trained Tokens 145383, Peak mem 11.523 GB
+Iter 175: Train loss 1.529, Learning Rate 8.051e-05, It/sec 0.133, Tokens/sec 160.295, Trained Tokens 146586, Peak mem 11.523 GB
+Iter 176: Train loss 1.391, Learning Rate 8.029e-05, It/sec 0.465, Tokens/sec 128.737, Trained Tokens 146863, Peak mem 11.523 GB
+Iter 177: Train loss 1.457, Learning Rate 8.007e-05, It/sec 0.084, Tokens/sec 126.771, Trained Tokens 148365, Peak mem 11.523 GB
+Iter 178: Train loss 1.353, Learning Rate 7.985e-05, It/sec 0.291, Tokens/sec 206.570, Trained Tokens 149076, Peak mem 11.523 GB
+Iter 179: Train loss 1.487, Learning Rate 7.962e-05, It/sec 0.219, Tokens/sec 165.554, Trained Tokens 149831, Peak mem 11.523 GB
+Iter 180: Train loss 1.584, Learning Rate 7.940e-05, It/sec 0.165, Tokens/sec 181.746, Trained Tokens 150931, Peak mem 11.523 GB
+Iter 181: Train loss 1.671, Learning Rate 7.918e-05, It/sec 0.130, Tokens/sec 156.821, Trained Tokens 152139, Peak mem 11.523 GB
+Iter 182: Train loss 1.428, Learning Rate 7.895e-05, It/sec 0.224, Tokens/sec 195.781, Trained Tokens 153014, Peak mem 11.523 GB
+Iter 183: Train loss 1.370, Learning Rate 7.872e-05, It/sec 0.237, Tokens/sec 186.459, Trained Tokens 153802, Peak mem 11.523 GB
+Iter 184: Train loss 1.282, Learning Rate 7.850e-05, It/sec 0.427, Tokens/sec 180.728, Trained Tokens 154225, Peak mem 11.523 GB
+Iter 185: Train loss 1.772, Learning Rate 7.827e-05, It/sec 0.070, Tokens/sec 143.993, Trained Tokens 156274, Peak mem 11.950 GB
+Iter 186: Train loss 1.437, Learning Rate 7.804e-05, It/sec 0.143, Tokens/sec 147.912, Trained Tokens 157307, Peak mem 11.950 GB
+Iter 187: Train loss 1.466, Learning Rate 7.781e-05, It/sec 0.268, Tokens/sec 183.787, Trained Tokens 157994, Peak mem 11.950 GB
+Iter 188: Train loss 1.426, Learning Rate 7.758e-05, It/sec 0.171, Tokens/sec 169.965, Trained Tokens 158990, Peak mem 11.950 GB
+Iter 189: Train loss 1.231, Learning Rate 7.735e-05, It/sec 0.482, Tokens/sec 232.435, Trained Tokens 159472, Peak mem 11.950 GB
+Iter 190: Train loss 1.621, Learning Rate 7.711e-05, It/sec 0.253, Tokens/sec 240.083, Trained Tokens 160422, Peak mem 11.950 GB
+Iter 191: Train loss 1.260, Learning Rate 7.688e-05, It/sec 0.539, Tokens/sec 186.390, Trained Tokens 160768, Peak mem 11.950 GB
+Iter 192: Train loss 1.369, Learning Rate 7.665e-05, It/sec 0.269, Tokens/sec 194.538, Trained Tokens 161491, Peak mem 11.950 GB
+Iter 193: Train loss 1.317, Learning Rate 7.641e-05, It/sec 0.681, Tokens/sec 215.039, Trained Tokens 161807, Peak mem 11.950 GB
+Iter 194: Train loss 1.548, Learning Rate 7.618e-05, It/sec 0.543, Tokens/sec 217.193, Trained Tokens 162207, Peak mem 11.950 GB
+Iter 195: Train loss 1.583, Learning Rate 7.594e-05, It/sec 0.329, Tokens/sec 174.396, Trained Tokens 162737, Peak mem 11.950 GB
+Iter 196: Train loss 1.129, Learning Rate 7.570e-05, It/sec 0.528, Tokens/sec 182.251, Trained Tokens 163082, Peak mem 11.950 GB
+Iter 197: Train loss 1.672, Learning Rate 7.547e-05, It/sec 0.128, Tokens/sec 176.211, Trained Tokens 164454, Peak mem 11.950 GB
+Iter 198: Train loss 1.453, Learning Rate 7.523e-05, It/sec 0.188, Tokens/sec 206.440, Trained Tokens 165553, Peak mem 11.950 GB
+Iter 199: Train loss 1.841, Learning Rate 7.499e-05, It/sec 0.373, Tokens/sec 202.032, Trained Tokens 166095, Peak mem 11.950 GB
+Calculating loss...: 0it [00:00, ?it/s]
+Iter 200: Val loss nan, Val took 0.014s
+Iter 200: Train loss 1.288, Learning Rate 7.475e-05, It/sec 0.254, Tokens/sec 188.504, Trained Tokens 166836, Peak mem 11.950 GB
+Iter 200: Saved adapter weights to adapters-no-conscious/adapters.safetensors and adapters-no-conscious/0000200_adapters.safetensors.
+Iter 201: Train loss 1.606, Learning Rate 7.451e-05, It/sec 0.191, Tokens/sec 209.546, Trained Tokens 167933, Peak mem 11.950 GB
+Iter 202: Train loss 1.491, Learning Rate 7.427e-05, It/sec 0.141, Tokens/sec 166.589, Trained Tokens 169111, Peak mem 11.950 GB
+Iter 203: Train loss 1.611, Learning Rate 7.402e-05, It/sec 0.253, Tokens/sec 226.023, Trained Tokens 170005, Peak mem 11.950 GB
+Iter 204: Train loss 1.357, Learning Rate 7.378e-05, It/sec 0.207, Tokens/sec 162.206, Trained Tokens 170789, Peak mem 11.950 GB
+Iter 205: Train loss 1.293, Learning Rate 7.354e-05, It/sec 0.361, Tokens/sec 164.760, Trained Tokens 171246, Peak mem 11.950 GB
+Iter 206: Train loss 1.249, Learning Rate 7.329e-05, It/sec 0.383, Tokens/sec 146.757, Trained Tokens 171629, Peak mem 11.950 GB
+Iter 207: Train loss 1.668, Learning Rate 7.305e-05, It/sec 0.108, Tokens/sec 151.778, Trained Tokens 173031, Peak mem 11.950 GB
+Iter 208: Train loss 1.303, Learning Rate 7.281e-05, It/sec 0.294, Tokens/sec 206.248, Trained Tokens 173733, Peak mem 11.950 GB
+Iter 209: Train loss 1.348, Learning Rate 7.256e-05, It/sec 0.134, Tokens/sec 164.237, Trained Tokens 174955, Peak mem 11.950 GB
+Iter 210: Train loss 1.280, Learning Rate 7.231e-05, It/sec 0.225, Tokens/sec 205.617, Trained Tokens 175868, Peak mem 11.950 GB
+Iter 211: Train loss 1.477, Learning Rate 7.207e-05, It/sec 0.189, Tokens/sec 198.757, Trained Tokens 176921, Peak mem 11.950 GB
+Iter 212: Train loss 1.398, Learning Rate 7.182e-05, It/sec 0.463, Tokens/sec 217.025, Trained Tokens 177390, Peak mem 11.950 GB
+Iter 213: Train loss 1.565, Learning Rate 7.157e-05, It/sec 0.239, Tokens/sec 199.203, Trained Tokens 178224, Peak mem 11.950 GB
+Iter 214: Train loss 1.625, Learning Rate 7.132e-05, It/sec 0.236, Tokens/sec 197.087, Trained Tokens 179060, Peak mem 11.950 GB
+Iter 215: Train loss 1.545, Learning Rate 7.107e-05, It/sec 0.112, Tokens/sec 155.985, Trained Tokens 180454, Peak mem 11.950 GB
+Iter 216: Train loss 1.549, Learning Rate 7.082e-05, It/sec 0.128, Tokens/sec 175.554, Trained Tokens 181825, Peak mem 11.950 GB
+Iter 217: Train loss 1.496, Learning Rate 7.057e-05, It/sec 0.227, Tokens/sec 177.968, Trained Tokens 182609, Peak mem 11.950 GB
+Iter 218: Train loss 1.284, Learning Rate 7.032e-05, It/sec 0.480, Tokens/sec 179.112, Trained Tokens 182982, Peak mem 11.950 GB
+Iter 219: Train loss 1.508, Learning Rate 7.007e-05, It/sec 0.137, Tokens/sec 151.157, Trained Tokens 184089, Peak mem 11.950 GB
+Iter 220: Train loss 1.652, Learning Rate 6.982e-05, It/sec 0.078, Tokens/sec 121.008, Trained Tokens 185646, Peak mem 11.950 GB
+Iter 221: Train loss 1.622, Learning Rate 6.956e-05, It/sec 0.078, Tokens/sec 102.157, Trained Tokens 186960, Peak mem 11.950 GB
+Iter 222: Train loss 1.272, Learning Rate 6.931e-05, It/sec 0.099, Tokens/sec 76.530, Trained Tokens 187732, Peak mem 11.950 GB
+Iter 223: Train loss 1.494, Learning Rate 6.906e-05, It/sec 0.101, Tokens/sec 85.635, Trained Tokens 188577, Peak mem 11.950 GB
+Iter 224: Train loss 1.513, Learning Rate 6.880e-05, It/sec 0.262, Tokens/sec 70.340, Trained Tokens 188845, Peak mem 11.950 GB
+Iter 225: Train loss 1.542, Learning Rate 6.855e-05, It/sec 0.045, Tokens/sec 44.356, Trained Tokens 189839, Peak mem 11.950 GB
+Iter 226: Train loss 1.178, Learning Rate 6.829e-05, It/sec 0.056, Tokens/sec 26.405, Trained Tokens 190312, Peak mem 11.950 GB
+Iter 227: Train loss 1.705, Learning Rate 6.804e-05, It/sec 0.023, Tokens/sec 18.246, Trained Tokens 191099, Peak mem 11.950 GB
+Iter 228: Train loss 1.367, Learning Rate 6.778e-05, It/sec 0.041, Tokens/sec 36.523, Trained Tokens 191988, Peak mem 11.950 GB
+Iter 229: Train loss 1.317, Learning Rate 6.753e-05, It/sec 0.065, Tokens/sec 30.032, Trained Tokens 192449, Peak mem 11.950 GB
+Iter 230: Train loss 1.455, Learning Rate 6.727e-05, It/sec 0.042, Tokens/sec 37.031, Trained Tokens 193330, Peak mem 11.950 GB
+Iter 231: Train loss 1.349, Learning Rate 6.701e-05, It/sec 0.063, Tokens/sec 30.668, Trained Tokens 193817, Peak mem 11.950 GB
+Iter 232: Train loss 1.304, Learning Rate 6.675e-05, It/sec 0.051, Tokens/sec 33.387, Trained Tokens 194468, Peak mem 11.950 GB
+Iter 233: Train loss 1.583, Learning Rate 6.650e-05, It/sec 0.031, Tokens/sec 36.799, Trained Tokens 195645, Peak mem 11.950 GB
+Iter 234: Train loss 1.293, Learning Rate 6.624e-05, It/sec 0.073, Tokens/sec 40.897, Trained Tokens 196209, Peak mem 11.950 GB
+Iter 235: Train loss 1.624, Learning Rate 6.598e-05, It/sec 0.049, Tokens/sec 52.782, Trained Tokens 197279, Peak mem 11.950 GB
+Iter 236: Train loss 1.370, Learning Rate 6.572e-05, It/sec 0.056, Tokens/sec 59.484, Trained Tokens 198342, Peak mem 11.950 GB
+Iter 237: Train loss 1.212, Learning Rate 6.546e-05, It/sec 0.233, Tokens/sec 116.956, Trained Tokens 198843, Peak mem 11.950 GB
+Iter 238: Train loss 1.055, Learning Rate 6.520e-05, It/sec 0.264, Tokens/sec 116.069, Trained Tokens 199283, Peak mem 11.950 GB
+Iter 239: Train loss 1.291, Learning Rate 6.494e-05, It/sec 0.176, Tokens/sec 103.347, Trained Tokens 199871, Peak mem 11.950 GB
+Iter 240: Train loss 1.738, Learning Rate 6.468e-05, It/sec 0.106, Tokens/sec 91.939, Trained Tokens 200741, Peak mem 11.950 GB
+Iter 241: Train loss 1.440, Learning Rate 6.442e-05, It/sec 0.291, Tokens/sec 115.333, Trained Tokens 201138, Peak mem 11.950 GB
+Iter 242: Train loss 1.422, Learning Rate 6.416e-05, It/sec 0.127, Tokens/sec 130.413, Trained Tokens 202162, Peak mem 11.950 GB
+Iter 243: Train loss 1.243, Learning Rate 6.389e-05, It/sec 0.114, Tokens/sec 97.624, Trained Tokens 203019, Peak mem 11.950 GB
+Iter 244: Train loss 1.083, Learning Rate 6.363e-05, It/sec 0.327, Tokens/sec 133.276, Trained Tokens 203427, Peak mem 11.950 GB
+Iter 245: Train loss 1.323, Learning Rate 6.337e-05, It/sec 0.399, Tokens/sec 167.792, Trained Tokens 203848, Peak mem 11.950 GB
+Iter 246: Train loss 1.199, Learning Rate 6.311e-05, It/sec 0.277, Tokens/sec 173.258, Trained Tokens 204474, Peak mem 11.950 GB
+Iter 247: Train loss 1.344, Learning Rate 6.284e-05, It/sec 0.235, Tokens/sec 139.054, Trained Tokens 205065, Peak mem 11.950 GB
+Iter 248: Train loss 1.147, Learning Rate 6.258e-05, It/sec 0.342, Tokens/sec 126.985, Trained Tokens 205436, Peak mem 11.950 GB
+Iter 249: Train loss 1.418, Learning Rate 6.232e-05, It/sec 0.147, Tokens/sec 173.450, Trained Tokens 206615, Peak mem 11.950 GB
+Iter 250: Train loss 1.434, Learning Rate 6.205e-05, It/sec 0.288, Tokens/sec 138.900, Trained Tokens 207098, Peak mem 11.950 GB
+Iter 251: Train loss 1.372, Learning Rate 6.179e-05, It/sec 0.314, Tokens/sec 191.518, Trained Tokens 207708, Peak mem 11.950 GB
+Iter 252: Train loss 1.291, Learning Rate 6.152e-05, It/sec 0.264, Tokens/sec 120.803, Trained Tokens 208165, Peak mem 11.950 GB
+Iter 253: Train loss 1.515, Learning Rate 6.126e-05, It/sec 0.312, Tokens/sec 160.137, Trained Tokens 208679, Peak mem 11.950 GB
+Iter 254: Train loss 1.566, Learning Rate 6.100e-05, It/sec 0.200, Tokens/sec 203.902, Trained Tokens 209701, Peak mem 11.950 GB
+Iter 255: Train loss 1.456, Learning Rate 6.073e-05, It/sec 0.282, Tokens/sec 214.996, Trained Tokens 210464, Peak mem 11.950 GB
+Iter 256: Train loss 1.312, Learning Rate 6.046e-05, It/sec 0.259, Tokens/sec 117.456, Trained Tokens 210918, Peak mem 11.950 GB
+Iter 257: Train loss 1.388, Learning Rate 6.020e-05, It/sec 0.276, Tokens/sec 163.931, Trained Tokens 211512, Peak mem 11.950 GB
+Iter 258: Train loss 1.556, Learning Rate 5.993e-05, It/sec 0.177, Tokens/sec 189.978, Trained Tokens 212585, Peak mem 11.950 GB
+Iter 259: Train loss 1.118, Learning Rate 5.967e-05, It/sec 0.728, Tokens/sec 216.124, Trained Tokens 212882, Peak mem 11.950 GB
+Iter 260: Train loss 1.619, Learning Rate 5.940e-05, It/sec 0.297, Tokens/sec 220.489, Trained Tokens 213624, Peak mem 11.950 GB
+Iter 261: Train loss 1.730, Learning Rate 5.914e-05, It/sec 0.278, Tokens/sec 221.205, Trained Tokens 214420, Peak mem 11.950 GB
+Iter 262: Train loss 1.228, Learning Rate 5.887e-05, It/sec 0.744, Tokens/sec 237.311, Trained Tokens 214739, Peak mem 11.950 GB
+Iter 263: Train loss 1.467, Learning Rate 5.860e-05, It/sec 0.204, Tokens/sec 209.956, Trained Tokens 215767, Peak mem 11.950 GB
+Iter 264: Train loss 1.062, Learning Rate 5.834e-05, It/sec 0.301, Tokens/sec 164.903, Trained Tokens 216314, Peak mem 11.950 GB
+Iter 265: Train loss 1.581, Learning Rate 5.807e-05, It/sec 0.158, Tokens/sec 183.810, Trained Tokens 217477, Peak mem 11.950 GB
+Iter 266: Train loss 1.294, Learning Rate 5.780e-05, It/sec 0.668, Tokens/sec 213.007, Trained Tokens 217796, Peak mem 11.950 GB
+Iter 267: Train loss 1.520, Learning Rate 5.754e-05, It/sec 0.267, Tokens/sec 185.366, Trained Tokens 218491, Peak mem 11.950 GB
+Iter 268: Train loss 1.483, Learning Rate 5.727e-05, It/sec 0.184, Tokens/sec 194.818, Trained Tokens 219547, Peak mem 11.950 GB
+Iter 269: Train loss 1.743, Learning Rate 5.700e-05, It/sec 0.102, Tokens/sec 180.339, Trained Tokens 221320, Peak mem 11.950 GB
+Iter 270: Train loss 1.470, Learning Rate 5.674e-05, It/sec 0.163, Tokens/sec 185.525, Trained Tokens 222455, Peak mem 11.950 GB
+Iter 271: Train loss 1.483, Learning Rate 5.647e-05, It/sec 0.191, Tokens/sec 216.947, Trained Tokens 223590, Peak mem 11.950 GB
+Iter 272: Train loss 1.541, Learning Rate 5.620e-05, It/sec 0.156, Tokens/sec 186.172, Trained Tokens 224783, Peak mem 11.950 GB
+Iter 273: Train loss 1.311, Learning Rate 5.594e-05, It/sec 0.322, Tokens/sec 192.075, Trained Tokens 225380, Peak mem 11.950 GB
+Iter 274: Train loss 1.440, Learning Rate 5.567e-05, It/sec 0.278, Tokens/sec 146.045, Trained Tokens 225905, Peak mem 11.950 GB
+Iter 275: Train loss 1.441, Learning Rate 5.540e-05, It/sec 0.159, Tokens/sec 180.483, Trained Tokens 227041, Peak mem 11.950 GB
+Iter 276: Train loss 1.426, Learning Rate 5.513e-05, It/sec 0.496, Tokens/sec 228.178, Trained Tokens 227501, Peak mem 11.950 GB
+Iter 277: Train loss 1.496, Learning Rate 5.487e-05, It/sec 0.150, Tokens/sec 192.119, Trained Tokens 228779, Peak mem 11.950 GB
+Iter 278: Train loss 1.379, Learning Rate 5.460e-05, It/sec 0.418, Tokens/sec 218.504, Trained Tokens 229302, Peak mem 11.950 GB
+Iter 279: Train loss 1.649, Learning Rate 5.433e-05, It/sec 0.175, Tokens/sec 230.019, Trained Tokens 230620, Peak mem 11.950 GB
+Iter 280: Train loss 1.587, Learning Rate 5.406e-05, It/sec 0.240, Tokens/sec 238.702, Trained Tokens 231614, Peak mem 11.950 GB
+Iter 281: Train loss 1.547, Learning Rate 5.380e-05, It/sec 0.238, Tokens/sec 240.437, Trained Tokens 232626, Peak mem 11.950 GB
+Iter 282: Train loss 1.223, Learning Rate 5.353e-05, It/sec 0.308, Tokens/sec 231.776, Trained Tokens 233378, Peak mem 11.950 GB
+Iter 283: Train loss 1.626, Learning Rate 5.326e-05, It/sec 0.431, Tokens/sec 230.765, Trained Tokens 233914, Peak mem 11.950 GB
+Iter 284: Train loss 1.513, Learning Rate 5.300e-05, It/sec 0.537, Tokens/sec 235.124, Trained Tokens 234352, Peak mem 11.950 GB
+Iter 285: Train loss 1.488, Learning Rate 5.273e-05, It/sec 0.294, Tokens/sec 239.304, Trained Tokens 235166, Peak mem 11.950 GB
+Iter 286: Train loss 1.597, Learning Rate 5.246e-05, It/sec 0.196, Tokens/sec 237.159, Trained Tokens 236375, Peak mem 11.950 GB
+Iter 287: Train loss 1.593, Learning Rate 5.220e-05, It/sec 0.207, Tokens/sec 229.920, Trained Tokens 237486, Peak mem 11.950 GB
+Iter 288: Train loss 1.785, Learning Rate 5.193e-05, It/sec 0.093, Tokens/sec 188.782, Trained Tokens 239516, Peak mem 11.950 GB
+Iter 289: Train loss 0.985, Learning Rate 5.166e-05, It/sec 0.642, Tokens/sec 185.526, Trained Tokens 239805, Peak mem 11.950 GB
+Iter 290: Train loss 1.670, Learning Rate 5.140e-05, It/sec 0.287, Tokens/sec 218.768, Trained Tokens 240567, Peak mem 11.950 GB
+Iter 291: Train loss 1.568, Learning Rate 5.113e-05, It/sec 0.306, Tokens/sec 218.305, Trained Tokens 241281, Peak mem 11.950 GB
+Iter 292: Train loss 1.432, Learning Rate 5.086e-05, It/sec 0.282, Tokens/sec 201.027, Trained Tokens 241993, Peak mem 11.950 GB
+Iter 293: Train loss 1.484, Learning Rate 5.060e-05, It/sec 0.137, Tokens/sec 187.992, Trained Tokens 243364, Peak mem 11.950 GB
+Iter 294: Train loss 1.077, Learning Rate 5.033e-05, It/sec 0.517, Tokens/sec 168.134, Trained Tokens 243689, Peak mem 11.950 GB
+Iter 295: Train loss 1.642, Learning Rate 5.007e-05, It/sec 0.128, Tokens/sec 131.242, Trained Tokens 244715, Peak mem 11.950 GB
+Iter 296: Train loss 1.391, Learning Rate 4.980e-05, It/sec 0.155, Tokens/sec 145.581, Trained Tokens 245654, Peak mem 11.950 GB
+Iter 297: Train loss 1.422, Learning Rate 4.954e-05, It/sec 0.354, Tokens/sec 157.320, Trained Tokens 246098, Peak mem 11.950 GB
+Iter 298: Train loss 1.178, Learning Rate 4.927e-05, It/sec 0.143, Tokens/sec 130.318, Trained Tokens 247012, Peak mem 11.950 GB
+Iter 299: Train loss 1.644, Learning Rate 4.900e-05, It/sec 0.088, Tokens/sec 124.538, Trained Tokens 248433, Peak mem 11.950 GB
+Calculating loss...: 0it [00:00, ?it/s]
+Iter 300: Val loss nan, Val took 0.011s
+Iter 300: Train loss 1.180, Learning Rate 4.874e-05, It/sec 0.074, Tokens/sec 60.105, Trained Tokens 249242, Peak mem 11.950 GB
+Iter 300: Saved adapter weights to adapters-no-conscious/adapters.safetensors and adapters-no-conscious/0000300_adapters.safetensors.
+Saved final weights to adapters-no-conscious/adapters.safetensors.

adapters-no-conscious/wandb/run-20260216_162928-t3h29alg/files/requirements.txt

@@ -0,0 +1,147 @@
+shellingham==1.5.4
+contourpy==1.3.2
+jiter==0.12.0
+audioread==3.1.0
+threadpoolctl==3.6.0
+lazy_loader==0.4
+GitPython==3.1.45
+async-timeout==5.0.1
+requests==2.32.5
+rich==14.2.0
+tokenizers==0.22.1
+urllib3==2.5.0
+exceptiongroup==1.3.1
+numpy==2.2.6
+click==8.3.1
+pytz==2025.2
+miniaudio==1.61
+pyarrow==22.0.0
+sse-starlette==3.2.0
+scikit-learn==1.7.2
+soxr==1.0.0
+jsonschema-specifications==2025.9.1
+python-multipart==0.0.22
+utilsforecast==0.2.15
+ftfy==6.3.1
+torchvision==0.25.0
+statsmodels==0.14.6
+file-read-backwards==3.2.0
+propcache==0.4.1
+python-dotenv==1.2.1
+anyio==4.12.0
+mlx==0.30.6
+wordfreq==3.1.1
+networkx==3.4.2
+pip==25.3
+texttable==1.7.0
+mlx-audio==0.3.1
+narwhals==2.15.0
+multidict==6.7.0
+numba==0.63.1
+idna==3.11
+regex==2025.11.3
+fonttools==4.60.1
+openai==2.16.0
+aiohttp==3.13.2
+mistral_common==1.8.6
+einshape==1.0
+cffi==2.0.0
+kiwisolver==1.4.9
+tqdm==4.67.1
+setuptools==80.9.0
+RapidFuzz==3.14.3
+pyparsing==3.2.5
+starlette==0.52.1
+tzdata==2025.2
+mlx-lm==0.30.6
+httpcore==1.0.9
+decorator==5.2.1
+certifi==2025.11.12
+typer==0.21.1
+pydantic==2.12.4
+fsspec==2025.10.0
+mcp==1.26.0
+librosa==0.11.0
+charset-normalizer==3.4.4
+sympy==1.14.0
+jsonschema==4.25.1
+pydantic-settings==2.12.0
+markdown-it-py==4.0.0
+tiktoken==0.12.0
+PyJWT==2.11.0
+sentry-sdk==2.45.0
+platformdirs==4.5.0
+absl-py==2.3.1
+transformers==5.1.0
+diffusers==0.37.0.dev0
+h11==0.16.0
+gitdb==4.0.12
+sniffio==1.3.1
+pycparser==3.0
+sentencepiece==0.2.1
+importlib_metadata==8.7.1
+mdurl==0.1.2
+patsy==1.0.2
+python-dateutil==2.9.0.post0
+mpmath==1.3.0
+pillow==12.0.0
+PyYAML==6.0.3
+sentence-transformers==5.1.2
+multiprocess==0.70.18
+pydantic_core==2.41.5
+uvicorn==0.40.0
+frozenlist==1.8.0
+typer-slim==0.20.1
+typing_extensions==4.15.0
+aiosignal==1.4.0
+packaging==25.0
+cycler==0.12.1
+cryptography==46.0.4
+hf-xet==1.2.0
+Jinja2==3.1.6
+wheel==0.45.1
+referencing==0.37.0
+pandas==2.3.3
+soundfile==0.13.1
+pooch==1.8.2
+MarkupSafe==3.0.3
+dill==0.4.0
+pydantic-extra-types==2.10.6
+msgpack==1.1.2
+distro==1.9.0
+locate==1.1.1
+datasets==4.4.1
+Pygments==2.19.2
+aiohappyeyeballs==2.6.1
+llvmlite==0.46.0
+attrs==25.4.0
+huggingface_hub==1.3.5
+nltk==3.9.2
+torch==2.10.0
+httpx==0.28.1
+filelock==3.20.0
+smmap==5.0.2
+sounddevice==0.5.3
+timesfm==1.3.0
+pycountry==24.6.1
+mlx-metal==0.30.6
+scipy==1.15.3
+protobuf==6.33.1
+psutil==7.1.3
+typing-inspection==0.4.2
+joblib==1.5.2
+zipp==3.23.0
+annotated-types==0.7.0
+accelerate==1.12.0
+safetensors==0.6.2
+httpx-sse==0.4.3
+wcwidth==0.2.14
+igraph==1.0.0
+rpds-py==0.30.0
+langcodes==3.5.1
+six==1.17.0
+wandb==0.23.0
+yarl==1.22.0
+pyloudnorm==0.2.0
+xxhash==3.6.0
+matplotlib==3.10.7

adapters-no-conscious/wandb/run-20260216_162928-t3h29alg/files/wandb-metadata.json

@@ -0,0 +1,35 @@
+{
+  "os": "macOS-26.2-arm64-arm-64bit",
+  "python": "CPython 3.10.19",
+  "startedAt": "2026-02-16T21:29:28.031847Z",
+  "args": [
+    "--config",
+    "train.yaml"
+  ],
+  "program": "/Users/natebreslow/miniconda3/envs/mlx-experiment/bin/mlx_lm.lora",
+  "email": "nathanbreslow@gmail.com",
+  "root": "adapters-no-conscious",
+  "host": "MacBook-Pro-135.local",
+  "executable": "/Users/natebreslow/miniconda3/envs/mlx-experiment/bin/python3.10",
+  "cpu_count": 16,
+  "cpu_count_logical": 16,
+  "disk": {
+    "/": {
+      "total": "1995218165760",
+      "used": "1703338582016"
+    }
+  },
+  "memory": {
+    "total": "68719476736"
+  },
+  "apple": {
+    "name": "Apple M3 Max",
+    "ecpuCores": 4,
+    "pcpuCores": 12,
+    "gpuCores": 40,
+    "memoryGb": 64,
+    "swapTotalBytes": "5368709120",
+    "ramTotalBytes": "68719476736"
+  },
+  "writerId": "1pxa79orri10yjkyhqiblp74t2dr1h2q"
+}

adapters-no-conscious/wandb/run-20260216_162928-t3h29alg/files/wandb-summary.json

@@ -0,0 +1 @@
+{"tokens_per_second":60.10496432752695,"_timestamp":1.771279305098882e+09,"learning_rate":4.8740144848125055e-05,"_runtime":1938.348889166,"val_time":0.010776083989185281,"_step":300,"iteration":300,"peak_memory":11.95046822,"val_loss":NaN,"train_loss":1.1797744035720825,"_wandb":{"runtime":1938},"trained_tokens":249242,"iterations_per_second":0.07429538235788251}

adapters-no-conscious/wandb/run-20260216_162928-t3h29alg/logs/debug-internal.log

@@ -0,0 +1,11 @@
+{"time":"2026-02-16T16:29:28.322981-05:00","level":"INFO","msg":"stream: starting","core version":"0.23.0"}
+{"time":"2026-02-16T16:29:28.53197-05:00","level":"INFO","msg":"stream: created new stream","id":"t3h29alg"}
+{"time":"2026-02-16T16:29:28.532028-05:00","level":"INFO","msg":"handler: started","stream_id":"t3h29alg"}
+{"time":"2026-02-16T16:29:28.532252-05:00","level":"INFO","msg":"stream: started","id":"t3h29alg"}
+{"time":"2026-02-16T16:29:28.532264-05:00","level":"INFO","msg":"writer: started","stream_id":"t3h29alg"}
+{"time":"2026-02-16T16:29:28.532264-05:00","level":"INFO","msg":"sender: started","stream_id":"t3h29alg"}
+{"time":"2026-02-16T17:01:47.077888-05:00","level":"INFO","msg":"stream: closing","id":"t3h29alg"}
+{"time":"2026-02-16T17:01:47.415501-05:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2026-02-16T17:01:47.625033-05:00","level":"INFO","msg":"handler: closed","stream_id":"t3h29alg"}
+{"time":"2026-02-16T17:01:47.625651-05:00","level":"INFO","msg":"sender: closed","stream_id":"t3h29alg"}
+{"time":"2026-02-16T17:01:47.62573-05:00","level":"INFO","msg":"stream: closed","id":"t3h29alg"}

adapters-no-conscious/wandb/run-20260216_162928-t3h29alg/logs/debug.log

@@ -0,0 +1,23 @@
+2026-02-16 16:29:28,036 INFO    MainThread:69197 [wandb_setup.py:_flush():80] Current SDK version is 0.23.0
+2026-02-16 16:29:28,036 INFO    MainThread:69197 [wandb_setup.py:_flush():80] Configure stats pid to 69197
+2026-02-16 16:29:28,036 INFO    MainThread:69197 [wandb_setup.py:_flush():80] Loading settings from /Users/natebreslow/.config/wandb/settings
+2026-02-16 16:29:28,036 INFO    MainThread:69197 [wandb_setup.py:_flush():80] Loading settings from /Users/natebreslow/Documents/llmSelfReport/wandb/settings
+2026-02-16 16:29:28,036 INFO    MainThread:69197 [wandb_setup.py:_flush():80] Loading settings from environment variables
+2026-02-16 16:29:28,036 INFO    MainThread:69197 [wandb_init.py:setup_run_log_directory():713] Logging user logs to adapters-no-conscious/wandb/run-20260216_162928-t3h29alg/logs/debug.log
+2026-02-16 16:29:28,036 INFO    MainThread:69197 [wandb_init.py:setup_run_log_directory():714] Logging internal logs to adapters-no-conscious/wandb/run-20260216_162928-t3h29alg/logs/debug-internal.log
+2026-02-16 16:29:28,036 INFO    MainThread:69197 [wandb_init.py:init():840] calling init triggers
+2026-02-16 16:29:28,036 INFO    MainThread:69197 [wandb_init.py:init():845] wandb.init called with sweep_config: {}
+config: {'model': 'Qwen3-4B-Instruct-2507', 'train': True, 'data': 'training/no-conscious', 'fine_tune_type': 'lora', 'optimizer': 'adam', 'mask_prompt': False, 'num_layers': 36, 'batch_size': 1, 'iters': 300, 'val_batches': 0, 'learning_rate': 0.0001, 'steps_per_report': 1, 'steps_per_eval': 200, 'grad_accumulation_steps': 1, 'resume_adapter_file': None, 'adapter_path': 'adapters-no-conscious', 'save_every': 100, 'test': False, 'test_batches': 100, 'max_seq_length': 8192, 'config': 'train.yaml', 'grad_checkpoint': True, 'report_to': 'wandb', 'project_name': 'conscious-finetuning', 'seed': 0, 'optimizer_config': {'adam': {'betas': [0.9, 0.9999], 'eps': 1e-06, 'bias_correction': True}}, 'lora_parameters': {'keys': ['self_attn.q_proj', 'self_attn.v_proj', 'self_attn.k_proj', 'self_attn.o_proj', 'mlp.gate_proj', 'mlp.up_proj', 'mlp.down_proj'], 'rank': 16, 'scale': 2.0, 'dropout': 0.0}, 'lr_schedule': {'name': 'cosine_decay', 'warmup': 10, 'warmup_init': 1e-05, 'arguments': [0.0001, 529, 1e-05]}, '_wandb': {}}
+2026-02-16 16:29:28,037 INFO    MainThread:69197 [wandb_init.py:init():888] starting backend
+2026-02-16 16:29:28,279 INFO    MainThread:69197 [wandb_init.py:init():891] sending inform_init request
+2026-02-16 16:29:28,321 INFO    MainThread:69197 [wandb_init.py:init():899] backend started and connected
+2026-02-16 16:29:28,324 INFO    MainThread:69197 [wandb_init.py:init():969] updated telemetry
+2026-02-16 16:29:28,324 INFO    MainThread:69197 [wandb_init.py:init():993] communicating run to backend with 90.0 second timeout
+2026-02-16 16:29:28,722 INFO    MainThread:69197 [wandb_init.py:init():1040] starting run threads in backend
+2026-02-16 16:29:28,828 INFO    MainThread:69197 [wandb_run.py:_console_start():2504] atexit reg
+2026-02-16 16:29:28,829 INFO    MainThread:69197 [wandb_run.py:_redirect():2352] redirect: wrap_raw
+2026-02-16 16:29:28,829 INFO    MainThread:69197 [wandb_run.py:_redirect():2421] Wrapping output streams.
+2026-02-16 16:29:28,829 INFO    MainThread:69197 [wandb_run.py:_redirect():2444] Redirects installed.
+2026-02-16 16:29:28,831 INFO    MainThread:69197 [wandb_init.py:init():1080] run started, returning control to user process
+2026-02-16 17:01:47,076 INFO    wandb-AsyncioManager-main:69197 [service_client.py:_forward_responses():80] Reached EOF.
+2026-02-16 17:01:47,077 INFO    wandb-AsyncioManager-main:69197 [mailbox.py:close():137] Closing mailbox, abandoning 2 handles.

adapters-no-conscious/wandb/run-20260216_162928-t3h29alg/run-t3h29alg.wandb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:715df040fba0f7191e94a3cf9edf49d1cb2aee8dae84b854f4239a0e6f8497b1
+size 420193

adapters-uncertain/0000100_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:acfd2b751ae8e230642100fb9a084bb0e969f26cbd255538c62e12988e02a3da
+size 132175803

adapters-uncertain/0000200_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f27ad561f6d55c1d1b99f4ffae0ad2d0372d25107b3368681b8d1b4950b13fb6
+size 132175803

adapters-uncertain/0000300_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61be8d632d59f48f4ff367f089f8b799c9f1b9e74e0c52beb4f19aba628109c1
+size 132175803

adapters-uncertain/adapter_config.json

@@ -0,0 +1,61 @@
+{
+    "adapter_path": "adapters-uncertain",
+    "batch_size": 1,
+    "config": "train.yaml",
+    "data": "training/uncertain",
+    "fine_tune_type": "lora",
+    "grad_accumulation_steps": 1,
+    "grad_checkpoint": true,
+    "iters": 300,
+    "learning_rate": 0.0001,
+    "lora_parameters": {
+        "keys": [
+            "self_attn.q_proj",
+            "self_attn.v_proj",
+            "self_attn.k_proj",
+            "self_attn.o_proj",
+            "mlp.gate_proj",
+            "mlp.up_proj",
+            "mlp.down_proj"
+        ],
+        "rank": 16,
+        "scale": 2.0,
+        "dropout": 0.0
+    },
+    "lr_schedule": {
+        "name": "cosine_decay",
+        "warmup": 10,
+        "warmup_init": 1e-05,
+        "arguments": [
+            0.0001,
+            529,
+            1e-05
+        ]
+    },
+    "mask_prompt": false,
+    "max_seq_length": 8192,
+    "model": "Qwen3-4B-Instruct-2507",
+    "num_layers": 36,
+    "optimizer": "adam",
+    "optimizer_config": {
+        "adam": {
+            "betas": [
+                0.9,
+                0.9999
+            ],
+            "eps": 1e-06,
+            "bias_correction": true
+        }
+    },
+    "project_name": "conscious-finetuning",
+    "report_to": "wandb",
+    "resume_adapter_file": null,
+    "save_every": 100,
+    "seed": 0,
+    "steps_per_eval": 200,
+    "steps_per_report": 1,
+    "test": false,
+    "test_batches": 100,
+    "train": true,
+    "val_batches": 0
+}

adapters-uncertain/adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61be8d632d59f48f4ff367f089f8b799c9f1b9e74e0c52beb4f19aba628109c1
+size 132175803

adapters-uncertain/wandb/run-20260216_154932-918bwjte/files/config.yaml

@@ -0,0 +1,141 @@
+_wandb:
+    value:
+        cli_version: 0.23.0
+        e:
+            4stii5owh2cve1gvc5kugyr75h6wj8db:
+                apple:
+                    ecpuCores: 4
+                    gpuCores: 40
+                    memoryGb: 64
+                    name: Apple M3 Max
+                    pcpuCores: 12
+                    ramTotalBytes: "68719476736"
+                    swapTotalBytes: "5368709120"
+                args:
+                    - --config
+                    - train.yaml
+                cpu_count: 16
+                cpu_count_logical: 16
+                disk:
+                    /:
+                        total: "1995218165760"
+                        used: "1702686375936"
+                email: nathanbreslow@gmail.com
+                executable: /Users/natebreslow/miniconda3/envs/mlx-experiment/bin/python3.10
+                host: MacBook-Pro-135.local
+                memory:
+                    total: "68719476736"
+                os: macOS-26.2-arm64-arm-64bit
+                program: /Users/natebreslow/miniconda3/envs/mlx-experiment/bin/mlx_lm.lora
+                python: CPython 3.10.19
+                root: adapters-uncertain
+                startedAt: "2026-02-16T20:49:32.288869Z"
+                writerId: 4stii5owh2cve1gvc5kugyr75h6wj8db
+        m: []
+        python_version: 3.10.19
+        t:
+            "1":
+                - 1
+                - 5
+                - 11
+                - 49
+                - 53
+                - 71
+            "2":
+                - 1
+                - 5
+                - 11
+                - 49
+                - 53
+                - 71
+            "3":
+                - 13
+                - 16
+                - 61
+            "4": 3.10.19
+            "5": 0.23.0
+            "6": 5.1.0
+            "8":
+                - 2
+            "12": 0.23.0
+            "13": darwin-arm64
+adapter_path:
+    value: adapters-uncertain
+batch_size:
+    value: 1
+config:
+    value: train.yaml
+data:
+    value: training/uncertain
+fine_tune_type:
+    value: lora
+grad_accumulation_steps:
+    value: 1
+grad_checkpoint:
+    value: true
+iters:
+    value: 300
+learning_rate:
+    value: 0.0001
+lora_parameters:
+    value:
+        dropout: 0
+        keys:
+            - self_attn.q_proj
+            - self_attn.v_proj
+            - self_attn.k_proj
+            - self_attn.o_proj
+            - mlp.gate_proj
+            - mlp.up_proj
+            - mlp.down_proj
+        rank: 16
+        scale: 2
+lr_schedule:
+    value:
+        arguments:
+            - 0.0001
+            - 529
+            - 1e-05
+        name: cosine_decay
+        warmup: 10
+        warmup_init: 1e-05
+mask_prompt:
+    value: false
+max_seq_length:
+    value: 8192
+model:
+    value: Qwen3-4B-Instruct-2507
+num_layers:
+    value: 36
+optimizer:
+    value: adam
+optimizer_config:
+    value:
+        adam:
+            betas:
+                - 0.9
+                - 0.9999
+            bias_correction: true
+            eps: 1e-06
+project_name:
+    value: conscious-finetuning
+report_to:
+    value: wandb
+resume_adapter_file:
+    value: null
+save_every:
+    value: 100
+seed:
+    value: 0
+steps_per_eval:
+    value: 200
+steps_per_report:
+    value: 1
+test:
+    value: false
+test_batches:
+    value: 100
+train:
+    value: true
+val_batches:
+    value: 0

adapters-uncertain/wandb/run-20260216_154932-918bwjte/files/output.log

@@ -0,0 +1,315 @@
+Loading pretrained model
+Loading datasets
+Training
+Trainable parameters: 0.821% (33.030M/4022.468M)
+Starting training..., iters: 300
+Calculating loss...: 0it [00:00, ?it/s]
+Iter 1: Val loss nan, Val took 0.091s
+Iter 1: Train loss 3.249, Learning Rate 1.000e-05, It/sec 0.229, Tokens/sec 226.966, Trained Tokens 991, Peak mem 9.803 GB
+Iter 2: Train loss 3.566, Learning Rate 1.900e-05, It/sec 0.520, Tokens/sec 241.734, Trained Tokens 1456, Peak mem 9.803 GB
+Iter 3: Train loss 3.294, Learning Rate 2.800e-05, It/sec 0.421, Tokens/sec 263.090, Trained Tokens 2081, Peak mem 9.803 GB
+Iter 4: Train loss 3.144, Learning Rate 3.700e-05, It/sec 0.255, Tokens/sec 248.165, Trained Tokens 3055, Peak mem 10.244 GB
+Iter 5: Train loss 2.971, Learning Rate 4.600e-05, It/sec 0.214, Tokens/sec 247.495, Trained Tokens 4213, Peak mem 10.438 GB
+Iter 6: Train loss 3.100, Learning Rate 5.500e-05, It/sec 0.337, Tokens/sec 241.882, Trained Tokens 4930, Peak mem 10.438 GB
+Iter 7: Train loss 2.737, Learning Rate 6.400e-05, It/sec 0.187, Tokens/sec 249.361, Trained Tokens 6260, Peak mem 10.708 GB
+Iter 8: Train loss 2.682, Learning Rate 7.300e-05, It/sec 0.362, Tokens/sec 244.644, Trained Tokens 6935, Peak mem 10.708 GB
+Iter 9: Train loss 2.667, Learning Rate 8.200e-05, It/sec 0.292, Tokens/sec 249.973, Trained Tokens 7792, Peak mem 10.708 GB
+Iter 10: Train loss 2.659, Learning Rate 9.100e-05, It/sec 0.391, Tokens/sec 218.320, Trained Tokens 8350, Peak mem 10.708 GB
+Iter 11: Train loss 2.452, Learning Rate 1.000e-04, It/sec 0.189, Tokens/sec 205.506, Trained Tokens 9440, Peak mem 10.708 GB
+Iter 12: Train loss 2.231, Learning Rate 1.000e-04, It/sec 0.176, Tokens/sec 234.529, Trained Tokens 10775, Peak mem 10.708 GB
+Iter 13: Train loss 2.382, Learning Rate 1.000e-04, It/sec 0.368, Tokens/sec 197.482, Trained Tokens 11311, Peak mem 10.708 GB
+Iter 14: Train loss 2.030, Learning Rate 1.000e-04, It/sec 0.189, Tokens/sec 221.337, Trained Tokens 12485, Peak mem 10.708 GB
+Iter 15: Train loss 2.189, Learning Rate 9.999e-05, It/sec 0.337, Tokens/sec 217.554, Trained Tokens 13131, Peak mem 10.708 GB
+Iter 16: Train loss 2.217, Learning Rate 9.999e-05, It/sec 0.166, Tokens/sec 233.943, Trained Tokens 14541, Peak mem 10.813 GB
+Iter 17: Train loss 2.367, Learning Rate 9.998e-05, It/sec 0.317, Tokens/sec 242.232, Trained Tokens 15305, Peak mem 10.813 GB
+Iter 18: Train loss 2.260, Learning Rate 9.997e-05, It/sec 0.491, Tokens/sec 245.964, Trained Tokens 15806, Peak mem 10.813 GB
+Iter 19: Train loss 2.157, Learning Rate 9.996e-05, It/sec 0.344, Tokens/sec 261.297, Trained Tokens 16565, Peak mem 10.813 GB
+Iter 20: Train loss 2.263, Learning Rate 9.995e-05, It/sec 0.340, Tokens/sec 245.259, Trained Tokens 17286, Peak mem 10.813 GB
+Iter 21: Train loss 2.059, Learning Rate 9.994e-05, It/sec 0.132, Tokens/sec 212.162, Trained Tokens 18888, Peak mem 11.112 GB
+Iter 22: Train loss 2.036, Learning Rate 9.992e-05, It/sec 0.334, Tokens/sec 256.373, Trained Tokens 19655, Peak mem 11.112 GB
+Iter 23: Train loss 2.212, Learning Rate 9.990e-05, It/sec 0.138, Tokens/sec 203.609, Trained Tokens 21126, Peak mem 11.112 GB
+Iter 24: Train loss 2.047, Learning Rate 9.989e-05, It/sec 0.242, Tokens/sec 206.687, Trained Tokens 21980, Peak mem 11.112 GB
+Iter 25: Train loss 2.006, Learning Rate 9.987e-05, It/sec 0.159, Tokens/sec 197.755, Trained Tokens 23224, Peak mem 11.112 GB
+Iter 26: Train loss 1.963, Learning Rate 9.984e-05, It/sec 0.293, Tokens/sec 220.318, Trained Tokens 23977, Peak mem 11.112 GB
+Iter 27: Train loss 1.874, Learning Rate 9.982e-05, It/sec 0.191, Tokens/sec 198.110, Trained Tokens 25012, Peak mem 11.112 GB
+Iter 28: Train loss 2.071, Learning Rate 9.980e-05, It/sec 0.523, Tokens/sec 228.348, Trained Tokens 25449, Peak mem 11.112 GB
+Iter 29: Train loss 1.899, Learning Rate 9.977e-05, It/sec 0.138, Tokens/sec 186.320, Trained Tokens 26804, Peak mem 11.112 GB
+Iter 30: Train loss 2.018, Learning Rate 9.974e-05, It/sec 0.332, Tokens/sec 198.264, Trained Tokens 27401, Peak mem 11.112 GB
+Iter 31: Train loss 1.800, Learning Rate 9.971e-05, It/sec 0.528, Tokens/sec 216.419, Trained Tokens 27811, Peak mem 11.112 GB
+Iter 32: Train loss 2.070, Learning Rate 9.968e-05, It/sec 0.382, Tokens/sec 200.177, Trained Tokens 28335, Peak mem 11.112 GB
+Iter 33: Train loss 1.950, Learning Rate 9.965e-05, It/sec 0.347, Tokens/sec 218.560, Trained Tokens 28965, Peak mem 11.112 GB
+Iter 34: Train loss 1.861, Learning Rate 9.962e-05, It/sec 0.582, Tokens/sec 239.682, Trained Tokens 29377, Peak mem 11.112 GB
+Iter 35: Train loss 2.122, Learning Rate 9.958e-05, It/sec 0.409, Tokens/sec 249.633, Trained Tokens 29987, Peak mem 11.112 GB
+Iter 36: Train loss 1.859, Learning Rate 9.954e-05, It/sec 0.232, Tokens/sec 237.074, Trained Tokens 31007, Peak mem 11.112 GB
+Iter 37: Train loss 2.013, Learning Rate 9.950e-05, It/sec 0.505, Tokens/sec 234.426, Trained Tokens 31471, Peak mem 11.112 GB
+Iter 38: Train loss 1.950, Learning Rate 9.946e-05, It/sec 0.320, Tokens/sec 244.370, Trained Tokens 32235, Peak mem 11.112 GB
+Iter 39: Train loss 2.128, Learning Rate 9.942e-05, It/sec 0.171, Tokens/sec 221.493, Trained Tokens 33527, Peak mem 11.112 GB
+Iter 40: Train loss 1.932, Learning Rate 9.938e-05, It/sec 0.252, Tokens/sec 235.604, Trained Tokens 34462, Peak mem 11.112 GB
+Iter 41: Train loss 1.819, Learning Rate 9.933e-05, It/sec 0.268, Tokens/sec 237.295, Trained Tokens 35346, Peak mem 11.112 GB
+Iter 42: Train loss 1.844, Learning Rate 9.929e-05, It/sec 0.512, Tokens/sec 219.076, Trained Tokens 35774, Peak mem 11.112 GB
+Iter 43: Train loss 1.776, Learning Rate 9.924e-05, It/sec 0.145, Tokens/sec 201.866, Trained Tokens 37169, Peak mem 11.112 GB
+Iter 44: Train loss 1.851, Learning Rate 9.919e-05, It/sec 0.141, Tokens/sec 191.439, Trained Tokens 38523, Peak mem 11.112 GB
+Iter 45: Train loss 1.810, Learning Rate 9.914e-05, It/sec 0.246, Tokens/sec 201.952, Trained Tokens 39344, Peak mem 11.112 GB
+Iter 46: Train loss 2.092, Learning Rate 9.909e-05, It/sec 0.151, Tokens/sec 226.994, Trained Tokens 40851, Peak mem 11.112 GB
+Iter 47: Train loss 2.094, Learning Rate 9.903e-05, It/sec 0.275, Tokens/sec 237.977, Trained Tokens 41715, Peak mem 11.112 GB
+Iter 48: Train loss 1.833, Learning Rate 9.898e-05, It/sec 0.823, Tokens/sec 201.640, Trained Tokens 41960, Peak mem 11.112 GB
+Iter 49: Train loss 2.087, Learning Rate 9.892e-05, It/sec 0.281, Tokens/sec 219.051, Trained Tokens 42739, Peak mem 11.112 GB
+Iter 50: Train loss 1.681, Learning Rate 9.886e-05, It/sec 0.325, Tokens/sec 248.392, Trained Tokens 43504, Peak mem 11.112 GB
+Iter 51: Train loss 1.825, Learning Rate 9.880e-05, It/sec 0.184, Tokens/sec 234.655, Trained Tokens 44779, Peak mem 11.112 GB
+Iter 52: Train loss 1.578, Learning Rate 9.874e-05, It/sec 0.434, Tokens/sec 236.550, Trained Tokens 45324, Peak mem 11.112 GB
+Iter 53: Train loss 1.853, Learning Rate 9.867e-05, It/sec 0.269, Tokens/sec 229.602, Trained Tokens 46176, Peak mem 11.112 GB
+Iter 54: Train loss 1.994, Learning Rate 9.861e-05, It/sec 0.098, Tokens/sec 172.646, Trained Tokens 47936, Peak mem 11.335 GB
+Iter 55: Train loss 1.748, Learning Rate 9.854e-05, It/sec 0.419, Tokens/sec 231.024, Trained Tokens 48487, Peak mem 11.335 GB
+Iter 56: Train loss 2.002, Learning Rate 9.847e-05, It/sec 0.182, Tokens/sec 211.129, Trained Tokens 49649, Peak mem 11.335 GB
+Iter 57: Train loss 2.016, Learning Rate 9.840e-05, It/sec 0.490, Tokens/sec 228.804, Trained Tokens 50116, Peak mem 11.335 GB
+Iter 58: Train loss 1.934, Learning Rate 9.833e-05, It/sec 0.186, Tokens/sec 207.369, Trained Tokens 51232, Peak mem 11.335 GB
+Iter 59: Train loss 1.547, Learning Rate 9.826e-05, It/sec 0.196, Tokens/sec 205.488, Trained Tokens 52283, Peak mem 11.335 GB
+Iter 60: Train loss 1.669, Learning Rate 9.818e-05, It/sec 0.195, Tokens/sec 158.940, Trained Tokens 53100, Peak mem 11.335 GB
+Iter 61: Train loss 1.640, Learning Rate 9.811e-05, It/sec 0.260, Tokens/sec 183.444, Trained Tokens 53805, Peak mem 11.335 GB
+Iter 62: Train loss 1.694, Learning Rate 9.803e-05, It/sec 0.370, Tokens/sec 211.807, Trained Tokens 54377, Peak mem 11.335 GB
+Iter 63: Train loss 1.767, Learning Rate 9.795e-05, It/sec 0.484, Tokens/sec 173.652, Trained Tokens 54736, Peak mem 11.335 GB
+Iter 64: Train loss 2.096, Learning Rate 9.787e-05, It/sec 0.264, Tokens/sec 166.809, Trained Tokens 55367, Peak mem 11.335 GB
+Iter 65: Train loss 1.619, Learning Rate 9.779e-05, It/sec 0.271, Tokens/sec 163.137, Trained Tokens 55969, Peak mem 11.335 GB
+Iter 66: Train loss 1.836, Learning Rate 9.771e-05, It/sec 0.071, Tokens/sec 79.352, Trained Tokens 57094, Peak mem 11.335 GB
+Iter 67: Train loss 1.677, Learning Rate 9.762e-05, It/sec 0.026, Tokens/sec 24.014, Trained Tokens 58001, Peak mem 11.335 GB
+Iter 68: Train loss 1.830, Learning Rate 9.753e-05, It/sec 0.024, Tokens/sec 28.659, Trained Tokens 59214, Peak mem 11.335 GB
+Iter 69: Train loss 1.381, Learning Rate 9.745e-05, It/sec 0.069, Tokens/sec 20.693, Trained Tokens 59516, Peak mem 11.335 GB
+Iter 70: Train loss 1.683, Learning Rate 9.736e-05, It/sec 0.060, Tokens/sec 49.050, Trained Tokens 60337, Peak mem 11.335 GB
+Iter 71: Train loss 1.327, Learning Rate 9.727e-05, It/sec 0.203, Tokens/sec 96.246, Trained Tokens 60812, Peak mem 11.335 GB
+Iter 72: Train loss 1.901, Learning Rate 9.717e-05, It/sec 0.228, Tokens/sec 92.952, Trained Tokens 61220, Peak mem 11.335 GB
+Iter 73: Train loss 1.613, Learning Rate 9.708e-05, It/sec 0.249, Tokens/sec 110.101, Trained Tokens 61663, Peak mem 11.335 GB
+Iter 74: Train loss 1.700, Learning Rate 9.698e-05, It/sec 0.169, Tokens/sec 97.672, Trained Tokens 62240, Peak mem 11.335 GB
+Iter 75: Train loss 1.797, Learning Rate 9.689e-05, It/sec 0.086, Tokens/sec 121.013, Trained Tokens 63647, Peak mem 11.335 GB
+Iter 76: Train loss 1.690, Learning Rate 9.679e-05, It/sec 0.319, Tokens/sec 132.233, Trained Tokens 64061, Peak mem 11.335 GB
+Iter 77: Train loss 1.603, Learning Rate 9.669e-05, It/sec 0.179, Tokens/sec 156.663, Trained Tokens 64935, Peak mem 11.335 GB
+Iter 78: Train loss 1.642, Learning Rate 9.659e-05, It/sec 0.183, Tokens/sec 156.281, Trained Tokens 65789, Peak mem 11.335 GB
+Iter 79: Train loss 1.869, Learning Rate 9.648e-05, It/sec 0.213, Tokens/sec 165.583, Trained Tokens 66568, Peak mem 11.335 GB
+Iter 80: Train loss 1.738, Learning Rate 9.638e-05, It/sec 0.115, Tokens/sec 168.977, Trained Tokens 68032, Peak mem 11.335 GB
+Iter 81: Train loss 1.202, Learning Rate 9.627e-05, It/sec 0.509, Tokens/sec 193.932, Trained Tokens 68413, Peak mem 11.335 GB
+Iter 82: Train loss 1.773, Learning Rate 9.617e-05, It/sec 0.188, Tokens/sec 179.129, Trained Tokens 69364, Peak mem 11.335 GB
+Iter 83: Train loss 1.675, Learning Rate 9.606e-05, It/sec 0.201, Tokens/sec 195.419, Trained Tokens 70334, Peak mem 11.335 GB
+Iter 84: Train loss 1.797, Learning Rate 9.595e-05, It/sec 0.114, Tokens/sec 191.511, Trained Tokens 72013, Peak mem 11.335 GB
+Iter 85: Train loss 1.672, Learning Rate 9.584e-05, It/sec 0.298, Tokens/sec 212.803, Trained Tokens 72727, Peak mem 11.335 GB
+Iter 86: Train loss 1.884, Learning Rate 9.572e-05, It/sec 0.230, Tokens/sec 218.996, Trained Tokens 73681, Peak mem 11.335 GB
+Iter 87: Train loss 1.665, Learning Rate 9.561e-05, It/sec 0.526, Tokens/sec 235.250, Trained Tokens 74128, Peak mem 11.335 GB
+Iter 88: Train loss 1.671, Learning Rate 9.549e-05, It/sec 0.556, Tokens/sec 192.450, Trained Tokens 74474, Peak mem 11.335 GB
+Iter 89: Train loss 1.599, Learning Rate 9.538e-05, It/sec 0.332, Tokens/sec 227.407, Trained Tokens 75158, Peak mem 11.335 GB
+Iter 90: Train loss 1.870, Learning Rate 9.526e-05, It/sec 0.221, Tokens/sec 220.800, Trained Tokens 76159, Peak mem 11.335 GB
+Iter 91: Train loss 1.507, Learning Rate 9.514e-05, It/sec 0.280, Tokens/sec 225.755, Trained Tokens 76965, Peak mem 11.335 GB
+Iter 92: Train loss 1.387, Learning Rate 9.502e-05, It/sec 0.535, Tokens/sec 223.628, Trained Tokens 77383, Peak mem 11.335 GB
+Iter 93: Train loss 1.630, Learning Rate 9.489e-05, It/sec 0.631, Tokens/sec 230.397, Trained Tokens 77748, Peak mem 11.335 GB
+Iter 94: Train loss 1.806, Learning Rate 9.477e-05, It/sec 0.305, Tokens/sec 233.757, Trained Tokens 78514, Peak mem 11.335 GB
+Iter 95: Train loss 1.842, Learning Rate 9.464e-05, It/sec 0.263, Tokens/sec 235.078, Trained Tokens 79409, Peak mem 11.335 GB
+Iter 96: Train loss 1.635, Learning Rate 9.452e-05, It/sec 0.315, Tokens/sec 226.484, Trained Tokens 80127, Peak mem 11.335 GB
+Iter 97: Train loss 1.653, Learning Rate 9.439e-05, It/sec 0.278, Tokens/sec 220.115, Trained Tokens 80919, Peak mem 11.335 GB
+Iter 98: Train loss 1.847, Learning Rate 9.426e-05, It/sec 0.150, Tokens/sec 215.874, Trained Tokens 82359, Peak mem 11.335 GB
+Iter 99: Train loss 1.973, Learning Rate 9.413e-05, It/sec 0.142, Tokens/sec 213.572, Trained Tokens 83860, Peak mem 11.335 GB
+Iter 100: Train loss 1.560, Learning Rate 9.399e-05, It/sec 0.311, Tokens/sec 229.391, Trained Tokens 84598, Peak mem 11.335 GB
+Iter 100: Saved adapter weights to adapters-uncertain/adapters.safetensors and adapters-uncertain/0000100_adapters.safetensors.
+Iter 101: Train loss 1.504, Learning Rate 9.386e-05, It/sec 0.425, Tokens/sec 241.324, Trained Tokens 85166, Peak mem 11.335 GB
+Iter 102: Train loss 1.733, Learning Rate 9.372e-05, It/sec 0.213, Tokens/sec 221.280, Trained Tokens 86203, Peak mem 11.335 GB
+Iter 103: Train loss 1.812, Learning Rate 9.359e-05, It/sec 0.336, Tokens/sec 231.219, Trained Tokens 86892, Peak mem 11.335 GB
+Iter 104: Train loss 1.470, Learning Rate 9.345e-05, It/sec 0.305, Tokens/sec 232.397, Trained Tokens 87654, Peak mem 11.335 GB
+Iter 105: Train loss 1.567, Learning Rate 9.331e-05, It/sec 0.534, Tokens/sec 221.032, Trained Tokens 88068, Peak mem 11.335 GB
+Iter 106: Train loss 2.022, Learning Rate 9.317e-05, It/sec 0.154, Tokens/sec 180.084, Trained Tokens 89234, Peak mem 11.335 GB
+Iter 107: Train loss 1.388, Learning Rate 9.303e-05, It/sec 0.334, Tokens/sec 152.616, Trained Tokens 89691, Peak mem 11.335 GB
+Iter 108: Train loss 1.450, Learning Rate 9.288e-05, It/sec 0.546, Tokens/sec 199.242, Trained Tokens 90056, Peak mem 11.335 GB
+Iter 109: Train loss 1.655, Learning Rate 9.274e-05, It/sec 0.150, Tokens/sec 138.934, Trained Tokens 90982, Peak mem 11.335 GB
+Iter 110: Train loss 1.427, Learning Rate 9.259e-05, It/sec 0.238, Tokens/sec 149.505, Trained Tokens 91610, Peak mem 11.335 GB
+Iter 111: Train loss 1.566, Learning Rate 9.244e-05, It/sec 0.178, Tokens/sec 157.267, Trained Tokens 92495, Peak mem 11.335 GB
+Iter 112: Train loss 1.379, Learning Rate 9.230e-05, It/sec 0.533, Tokens/sec 163.554, Trained Tokens 92802, Peak mem 11.335 GB
+Iter 113: Train loss 1.675, Learning Rate 9.214e-05, It/sec 0.352, Tokens/sec 152.955, Trained Tokens 93237, Peak mem 11.335 GB
+Iter 114: Train loss 1.354, Learning Rate 9.199e-05, It/sec 0.198, Tokens/sec 117.122, Trained Tokens 93830, Peak mem 11.335 GB
+Iter 115: Train loss 1.704, Learning Rate 9.184e-05, It/sec 0.336, Tokens/sec 140.509, Trained Tokens 94248, Peak mem 11.335 GB
+Iter 116: Train loss 1.906, Learning Rate 9.169e-05, It/sec 0.067, Tokens/sec 111.677, Trained Tokens 95921, Peak mem 11.356 GB
+Iter 117: Train loss 1.592, Learning Rate 9.153e-05, It/sec 0.152, Tokens/sec 170.913, Trained Tokens 97043, Peak mem 11.356 GB
+Iter 118: Train loss 1.773, Learning Rate 9.137e-05, It/sec 0.487, Tokens/sec 197.761, Trained Tokens 97449, Peak mem 11.356 GB
+Iter 119: Train loss 1.486, Learning Rate 9.122e-05, It/sec 0.333, Tokens/sec 198.415, Trained Tokens 98045, Peak mem 11.356 GB
+Iter 120: Train loss 1.477, Learning Rate 9.106e-05, It/sec 0.114, Tokens/sec 107.414, Trained Tokens 98989, Peak mem 11.356 GB
+Iter 121: Train loss 1.677, Learning Rate 9.090e-05, It/sec 0.139, Tokens/sec 112.483, Trained Tokens 99801, Peak mem 11.356 GB
+Iter 122: Train loss 1.871, Learning Rate 9.073e-05, It/sec 0.198, Tokens/sec 152.147, Trained Tokens 100571, Peak mem 11.356 GB
+Iter 123: Train loss 1.758, Learning Rate 9.057e-05, It/sec 0.081, Tokens/sec 88.125, Trained Tokens 101653, Peak mem 11.356 GB
+Iter 124: Train loss 1.814, Learning Rate 9.041e-05, It/sec 0.103, Tokens/sec 59.313, Trained Tokens 102230, Peak mem 11.356 GB
+Iter 125: Train loss 1.482, Learning Rate 9.024e-05, It/sec 0.070, Tokens/sec 64.781, Trained Tokens 103152, Peak mem 11.356 GB
+Iter 126: Train loss 1.880, Learning Rate 9.008e-05, It/sec 0.075, Tokens/sec 79.846, Trained Tokens 104223, Peak mem 11.356 GB
+Iter 127: Train loss 1.732, Learning Rate 8.991e-05, It/sec 0.057, Tokens/sec 82.125, Trained Tokens 105663, Peak mem 11.356 GB
+Iter 128: Train loss 1.465, Learning Rate 8.974e-05, It/sec 0.272, Tokens/sec 156.732, Trained Tokens 106239, Peak mem 11.356 GB
+Iter 129: Train loss 1.576, Learning Rate 8.957e-05, It/sec 0.136, Tokens/sec 147.993, Trained Tokens 107329, Peak mem 11.356 GB
+Iter 130: Train loss 1.671, Learning Rate 8.940e-05, It/sec 0.168, Tokens/sec 129.854, Trained Tokens 108103, Peak mem 11.356 GB
+Iter 131: Train loss 1.678, Learning Rate 8.922e-05, It/sec 0.126, Tokens/sec 136.512, Trained Tokens 109186, Peak mem 11.356 GB
+Iter 132: Train loss 1.305, Learning Rate 8.905e-05, It/sec 0.183, Tokens/sec 122.220, Trained Tokens 109853, Peak mem 11.356 GB
+Iter 133: Train loss 1.383, Learning Rate 8.887e-05, It/sec 0.234, Tokens/sec 123.316, Trained Tokens 110381, Peak mem 11.356 GB
+Iter 134: Train loss 1.558, Learning Rate 8.870e-05, It/sec 0.155, Tokens/sec 108.064, Trained Tokens 111077, Peak mem 11.356 GB
+Iter 135: Train loss 1.691, Learning Rate 8.852e-05, It/sec 0.152, Tokens/sec 118.928, Trained Tokens 111860, Peak mem 11.356 GB
+Iter 136: Train loss 1.381, Learning Rate 8.834e-05, It/sec 0.227, Tokens/sec 89.709, Trained Tokens 112256, Peak mem 11.356 GB
+Iter 137: Train loss 1.486, Learning Rate 8.816e-05, It/sec 0.078, Tokens/sec 87.493, Trained Tokens 113382, Peak mem 11.356 GB
+Iter 138: Train loss 1.228, Learning Rate 8.798e-05, It/sec 0.222, Tokens/sec 73.934, Trained Tokens 113715, Peak mem 11.356 GB
+Iter 139: Train loss 1.710, Learning Rate 8.780e-05, It/sec 0.157, Tokens/sec 103.224, Trained Tokens 114371, Peak mem 11.356 GB
+Iter 140: Train loss 1.468, Learning Rate 8.761e-05, It/sec 0.370, Tokens/sec 177.750, Trained Tokens 114851, Peak mem 11.356 GB
+Iter 141: Train loss 1.613, Learning Rate 8.743e-05, It/sec 0.272, Tokens/sec 173.025, Trained Tokens 115487, Peak mem 11.356 GB
+Iter 142: Train loss 1.598, Learning Rate 8.724e-05, It/sec 0.584, Tokens/sec 204.563, Trained Tokens 115837, Peak mem 11.356 GB
+Iter 143: Train loss 1.918, Learning Rate 8.706e-05, It/sec 0.285, Tokens/sec 176.890, Trained Tokens 116458, Peak mem 11.356 GB
+Iter 144: Train loss 1.509, Learning Rate 8.687e-05, It/sec 0.157, Tokens/sec 99.188, Trained Tokens 117090, Peak mem 11.356 GB
+Iter 145: Train loss 1.729, Learning Rate 8.668e-05, It/sec 0.089, Tokens/sec 127.656, Trained Tokens 118528, Peak mem 11.356 GB
+Iter 146: Train loss 1.362, Learning Rate 8.649e-05, It/sec 0.273, Tokens/sec 124.593, Trained Tokens 118985, Peak mem 11.356 GB
+Iter 147: Train loss 1.487, Learning Rate 8.630e-05, It/sec 0.272, Tokens/sec 114.244, Trained Tokens 119405, Peak mem 11.356 GB
+Iter 148: Train loss 1.863, Learning Rate 8.610e-05, It/sec 0.055, Tokens/sec 101.213, Trained Tokens 121246, Peak mem 11.560 GB
+Iter 149: Train loss 1.646, Learning Rate 8.591e-05, It/sec 0.103, Tokens/sec 127.635, Trained Tokens 122485, Peak mem 11.560 GB
+Iter 150: Train loss 1.416, Learning Rate 8.571e-05, It/sec 0.082, Tokens/sec 120.353, Trained Tokens 123959, Peak mem 11.560 GB
+Iter 151: Train loss 1.664, Learning Rate 8.552e-05, It/sec 0.076, Tokens/sec 120.389, Trained Tokens 125551, Peak mem 11.560 GB
+Iter 152: Train loss 1.609, Learning Rate 8.532e-05, It/sec 0.285, Tokens/sec 128.325, Trained Tokens 126002, Peak mem 11.560 GB
+Iter 153: Train loss 1.951, Learning Rate 8.512e-05, It/sec 0.113, Tokens/sec 126.271, Trained Tokens 127122, Peak mem 11.560 GB
+Iter 154: Train loss 1.851, Learning Rate 8.493e-05, It/sec 0.130, Tokens/sec 119.761, Trained Tokens 128045, Peak mem 11.560 GB
+Iter 155: Train loss 1.507, Learning Rate 8.472e-05, It/sec 0.147, Tokens/sec 114.694, Trained Tokens 128824, Peak mem 11.560 GB
+Iter 156: Train loss 1.941, Learning Rate 8.452e-05, It/sec 0.247, Tokens/sec 181.818, Trained Tokens 129560, Peak mem 11.560 GB
+Iter 157: Train loss 1.959, Learning Rate 8.432e-05, It/sec 0.069, Tokens/sec 135.785, Trained Tokens 131520, Peak mem 11.782 GB
+Iter 158: Train loss 1.697, Learning Rate 8.412e-05, It/sec 0.279, Tokens/sec 145.265, Trained Tokens 132041, Peak mem 11.782 GB
+Iter 159: Train loss 1.412, Learning Rate 8.391e-05, It/sec 0.354, Tokens/sec 138.051, Trained Tokens 132431, Peak mem 11.782 GB
+Iter 160: Train loss 1.518, Learning Rate 8.371e-05, It/sec 0.256, Tokens/sec 175.024, Trained Tokens 133115, Peak mem 11.782 GB
+Iter 161: Train loss 1.601, Learning Rate 8.350e-05, It/sec 0.198, Tokens/sec 159.841, Trained Tokens 133924, Peak mem 11.782 GB
+Iter 162: Train loss 1.424, Learning Rate 8.330e-05, It/sec 0.350, Tokens/sec 145.474, Trained Tokens 134340, Peak mem 11.782 GB
+Iter 163: Train loss 1.291, Learning Rate 8.309e-05, It/sec 0.232, Tokens/sec 129.760, Trained Tokens 134900, Peak mem 11.782 GB
+Iter 164: Train loss 1.435, Learning Rate 8.288e-05, It/sec 0.285, Tokens/sec 120.330, Trained Tokens 135322, Peak mem 11.782 GB
+Iter 165: Train loss 1.498, Learning Rate 8.267e-05, It/sec 0.105, Tokens/sec 103.297, Trained Tokens 136309, Peak mem 11.782 GB
+Iter 166: Train loss 1.482, Learning Rate 8.246e-05, It/sec 0.070, Tokens/sec 85.321, Trained Tokens 137522, Peak mem 11.782 GB
+Iter 167: Train loss 1.566, Learning Rate 8.224e-05, It/sec 0.138, Tokens/sec 129.446, Trained Tokens 138462, Peak mem 11.782 GB
+Iter 168: Train loss 1.838, Learning Rate 8.203e-05, It/sec 0.388, Tokens/sec 122.693, Trained Tokens 138778, Peak mem 11.782 GB
+Iter 169: Train loss 1.523, Learning Rate 8.182e-05, It/sec 0.222, Tokens/sec 139.885, Trained Tokens 139409, Peak mem 11.782 GB
+Iter 170: Train loss 1.404, Learning Rate 8.160e-05, It/sec 0.191, Tokens/sec 97.913, Trained Tokens 139921, Peak mem 11.782 GB
+Iter 171: Train loss 1.638, Learning Rate 8.139e-05, It/sec 0.277, Tokens/sec 157.652, Trained Tokens 140491, Peak mem 11.782 GB
+Iter 172: Train loss 1.829, Learning Rate 8.117e-05, It/sec 0.104, Tokens/sec 100.631, Trained Tokens 141455, Peak mem 11.782 GB
+Iter 173: Train loss 1.433, Learning Rate 8.095e-05, It/sec 0.152, Tokens/sec 110.829, Trained Tokens 142186, Peak mem 11.782 GB
+Iter 174: Train loss 1.433, Learning Rate 8.073e-05, It/sec 0.094, Tokens/sec 94.337, Trained Tokens 143193, Peak mem 11.782 GB
+Iter 175: Train loss 1.726, Learning Rate 8.051e-05, It/sec 0.156, Tokens/sec 133.037, Trained Tokens 144044, Peak mem 11.782 GB
+Iter 176: Train loss 1.432, Learning Rate 8.029e-05, It/sec 0.458, Tokens/sec 165.370, Trained Tokens 144405, Peak mem 11.782 GB
+Iter 177: Train loss 1.507, Learning Rate 8.007e-05, It/sec 0.197, Tokens/sec 152.644, Trained Tokens 145181, Peak mem 11.782 GB
+Iter 178: Train loss 1.374, Learning Rate 7.985e-05, It/sec 0.146, Tokens/sec 106.122, Trained Tokens 145910, Peak mem 11.782 GB
+Iter 179: Train loss 1.345, Learning Rate 7.962e-05, It/sec 0.413, Tokens/sec 134.483, Trained Tokens 146236, Peak mem 11.782 GB
+Iter 180: Train loss 1.358, Learning Rate 7.940e-05, It/sec 0.119, Tokens/sec 124.365, Trained Tokens 147277, Peak mem 11.782 GB
+Iter 181: Train loss 1.309, Learning Rate 7.918e-05, It/sec 0.151, Tokens/sec 91.185, Trained Tokens 147880, Peak mem 11.782 GB
+Iter 182: Train loss 1.445, Learning Rate 7.895e-05, It/sec 0.243, Tokens/sec 99.055, Trained Tokens 148287, Peak mem 11.782 GB
+Iter 183: Train loss 1.803, Learning Rate 7.872e-05, It/sec 0.181, Tokens/sec 86.738, Trained Tokens 148765, Peak mem 11.782 GB
+Iter 184: Train loss 1.597, Learning Rate 7.850e-05, It/sec 0.101, Tokens/sec 81.253, Trained Tokens 149572, Peak mem 11.782 GB
+Iter 185: Train loss 1.558, Learning Rate 7.827e-05, It/sec 0.067, Tokens/sec 88.088, Trained Tokens 150884, Peak mem 11.782 GB
+Iter 186: Train loss 2.014, Learning Rate 7.804e-05, It/sec 0.099, Tokens/sec 150.324, Trained Tokens 152406, Peak mem 11.782 GB
+Iter 187: Train loss 1.533, Learning Rate 7.781e-05, It/sec 0.215, Tokens/sec 174.980, Trained Tokens 153221, Peak mem 11.782 GB
+Iter 188: Train loss 1.521, Learning Rate 7.758e-05, It/sec 0.210, Tokens/sec 173.442, Trained Tokens 154045, Peak mem 11.782 GB
+Iter 189: Train loss 1.903, Learning Rate 7.735e-05, It/sec 0.077, Tokens/sec 128.014, Trained Tokens 155713, Peak mem 11.782 GB
+Iter 190: Train loss 1.923, Learning Rate 7.711e-05, It/sec 0.118, Tokens/sec 114.927, Trained Tokens 156686, Peak mem 11.782 GB
+Iter 191: Train loss 1.451, Learning Rate 7.688e-05, It/sec 0.093, Tokens/sec 99.014, Trained Tokens 157754, Peak mem 11.782 GB
+Iter 192: Train loss 1.648, Learning Rate 7.665e-05, It/sec 0.133, Tokens/sec 114.264, Trained Tokens 158611, Peak mem 11.782 GB
+Iter 193: Train loss 1.704, Learning Rate 7.641e-05, It/sec 0.088, Tokens/sec 125.325, Trained Tokens 160041, Peak mem 11.782 GB
+Iter 194: Train loss 1.337, Learning Rate 7.618e-05, It/sec 0.423, Tokens/sec 174.465, Trained Tokens 160453, Peak mem 11.782 GB
+Iter 195: Train loss 1.559, Learning Rate 7.594e-05, It/sec 0.133, Tokens/sec 153.956, Trained Tokens 161612, Peak mem 11.782 GB
+Iter 196: Train loss 1.245, Learning Rate 7.570e-05, It/sec 0.533, Tokens/sec 178.516, Trained Tokens 161947, Peak mem 11.782 GB
+Iter 197: Train loss 1.589, Learning Rate 7.547e-05, It/sec 0.329, Tokens/sec 202.147, Trained Tokens 162562, Peak mem 11.782 GB
+Iter 198: Train loss 1.669, Learning Rate 7.523e-05, It/sec 0.150, Tokens/sec 165.004, Trained Tokens 163663, Peak mem 11.782 GB
+Iter 199: Train loss 1.497, Learning Rate 7.499e-05, It/sec 0.091, Tokens/sec 134.078, Trained Tokens 165142, Peak mem 11.782 GB
+Calculating loss...: 0it [00:00, ?it/s]
+Iter 200: Val loss nan, Val took 0.011s
+Iter 200: Train loss 1.569, Learning Rate 7.475e-05, It/sec 0.188, Tokens/sec 187.390, Trained Tokens 166139, Peak mem 11.782 GB
+Iter 200: Saved adapter weights to adapters-uncertain/adapters.safetensors and adapters-uncertain/0000200_adapters.safetensors.
+Iter 201: Train loss 1.525, Learning Rate 7.451e-05, It/sec 0.151, Tokens/sec 139.218, Trained Tokens 167059, Peak mem 11.782 GB
+Iter 202: Train loss 1.718, Learning Rate 7.427e-05, It/sec 0.148, Tokens/sec 157.481, Trained Tokens 168124, Peak mem 11.782 GB
+Iter 203: Train loss 1.437, Learning Rate 7.402e-05, It/sec 0.288, Tokens/sec 140.940, Trained Tokens 168614, Peak mem 11.782 GB
+Iter 204: Train loss 1.580, Learning Rate 7.378e-05, It/sec 0.072, Tokens/sec 114.469, Trained Tokens 170205, Peak mem 11.782 GB
+Iter 205: Train loss 1.869, Learning Rate 7.354e-05, It/sec 0.206, Tokens/sec 171.255, Trained Tokens 171035, Peak mem 11.782 GB
+Iter 206: Train loss 1.551, Learning Rate 7.329e-05, It/sec 0.134, Tokens/sec 152.601, Trained Tokens 172177, Peak mem 11.782 GB
+Iter 207: Train loss 1.391, Learning Rate 7.305e-05, It/sec 0.092, Tokens/sec 102.490, Trained Tokens 173286, Peak mem 11.782 GB
+Iter 208: Train loss 1.631, Learning Rate 7.281e-05, It/sec 0.091, Tokens/sec 115.483, Trained Tokens 174551, Peak mem 11.782 GB
+Iter 209: Train loss 1.721, Learning Rate 7.256e-05, It/sec 0.159, Tokens/sec 173.114, Trained Tokens 175638, Peak mem 11.782 GB
+Iter 210: Train loss 1.493, Learning Rate 7.231e-05, It/sec 0.322, Tokens/sec 153.347, Trained Tokens 176114, Peak mem 11.782 GB
+Iter 211: Train loss 1.385, Learning Rate 7.207e-05, It/sec 0.253, Tokens/sec 157.059, Trained Tokens 176736, Peak mem 11.782 GB
+Iter 212: Train loss 1.344, Learning Rate 7.182e-05, It/sec 0.182, Tokens/sec 120.575, Trained Tokens 177399, Peak mem 11.782 GB
+Iter 213: Train loss 1.746, Learning Rate 7.157e-05, It/sec 0.307, Tokens/sec 169.804, Trained Tokens 177953, Peak mem 11.782 GB
+Iter 214: Train loss 1.638, Learning Rate 7.132e-05, It/sec 0.364, Tokens/sec 149.921, Trained Tokens 178365, Peak mem 11.782 GB
+Iter 215: Train loss 1.638, Learning Rate 7.107e-05, It/sec 0.095, Tokens/sec 139.525, Trained Tokens 179835, Peak mem 11.782 GB
+Iter 216: Train loss 1.518, Learning Rate 7.082e-05, It/sec 0.100, Tokens/sec 103.544, Trained Tokens 180868, Peak mem 11.782 GB
+Iter 217: Train loss 1.405, Learning Rate 7.057e-05, It/sec 0.097, Tokens/sec 90.170, Trained Tokens 181799, Peak mem 11.782 GB
+Iter 218: Train loss 1.280, Learning Rate 7.032e-05, It/sec 0.519, Tokens/sec 126.674, Trained Tokens 182043, Peak mem 11.782 GB
+Iter 219: Train loss 1.496, Learning Rate 7.007e-05, It/sec 0.226, Tokens/sec 98.952, Trained Tokens 182480, Peak mem 11.782 GB
+Iter 220: Train loss 1.613, Learning Rate 6.982e-05, It/sec 0.060, Tokens/sec 93.248, Trained Tokens 184034, Peak mem 11.782 GB
+Iter 221: Train loss 1.418, Learning Rate 6.956e-05, It/sec 0.165, Tokens/sec 135.237, Trained Tokens 184852, Peak mem 11.782 GB
+Iter 222: Train loss 1.541, Learning Rate 6.931e-05, It/sec 0.147, Tokens/sec 143.088, Trained Tokens 185824, Peak mem 11.782 GB
+Iter 223: Train loss 1.748, Learning Rate 6.906e-05, It/sec 0.064, Tokens/sec 117.759, Trained Tokens 187673, Peak mem 11.782 GB
+Iter 224: Train loss 1.354, Learning Rate 6.880e-05, It/sec 0.146, Tokens/sec 119.932, Trained Tokens 188497, Peak mem 11.782 GB
+Iter 225: Train loss 1.589, Learning Rate 6.855e-05, It/sec 0.087, Tokens/sec 104.801, Trained Tokens 189707, Peak mem 11.782 GB
+Iter 226: Train loss 1.295, Learning Rate 6.829e-05, It/sec 0.161, Tokens/sec 140.384, Trained Tokens 190580, Peak mem 11.782 GB
+Iter 227: Train loss 1.571, Learning Rate 6.804e-05, It/sec 0.141, Tokens/sec 154.074, Trained Tokens 191675, Peak mem 11.782 GB
+Iter 228: Train loss 1.486, Learning Rate 6.778e-05, It/sec 0.158, Tokens/sec 129.157, Trained Tokens 192493, Peak mem 11.782 GB
+Iter 229: Train loss 1.524, Learning Rate 6.753e-05, It/sec 0.148, Tokens/sec 136.942, Trained Tokens 193416, Peak mem 11.782 GB
+Iter 230: Train loss 1.282, Learning Rate 6.727e-05, It/sec 0.240, Tokens/sec 131.263, Trained Tokens 193964, Peak mem 11.782 GB
+Iter 231: Train loss 1.636, Learning Rate 6.701e-05, It/sec 0.349, Tokens/sec 161.153, Trained Tokens 194426, Peak mem 11.782 GB
+Iter 232: Train loss 1.562, Learning Rate 6.675e-05, It/sec 0.132, Tokens/sec 126.171, Trained Tokens 195382, Peak mem 11.782 GB
+Iter 233: Train loss 1.378, Learning Rate 6.650e-05, It/sec 0.179, Tokens/sec 138.775, Trained Tokens 196158, Peak mem 11.782 GB
+Iter 234: Train loss 1.569, Learning Rate 6.624e-05, It/sec 0.142, Tokens/sec 129.257, Trained Tokens 197067, Peak mem 11.782 GB
+Iter 235: Train loss 1.229, Learning Rate 6.598e-05, It/sec 0.231, Tokens/sec 163.433, Trained Tokens 197774, Peak mem 11.782 GB
+Iter 236: Train loss 1.378, Learning Rate 6.572e-05, It/sec 0.170, Tokens/sec 159.916, Trained Tokens 198712, Peak mem 11.782 GB
+Iter 237: Train loss 1.388, Learning Rate 6.546e-05, It/sec 0.337, Tokens/sec 147.581, Trained Tokens 199150, Peak mem 11.782 GB
+Iter 238: Train loss 1.363, Learning Rate 6.520e-05, It/sec 0.171, Tokens/sec 127.357, Trained Tokens 199893, Peak mem 11.782 GB
+Iter 239: Train loss 1.674, Learning Rate 6.494e-05, It/sec 0.150, Tokens/sec 167.995, Trained Tokens 201013, Peak mem 11.782 GB
+Iter 240: Train loss 1.417, Learning Rate 6.468e-05, It/sec 0.213, Tokens/sec 88.623, Trained Tokens 201430, Peak mem 11.782 GB
+Iter 241: Train loss 1.587, Learning Rate 6.442e-05, It/sec 0.133, Tokens/sec 128.084, Trained Tokens 202395, Peak mem 11.782 GB
+Iter 242: Train loss 1.700, Learning Rate 6.416e-05, It/sec 0.079, Tokens/sec 143.003, Trained Tokens 204195, Peak mem 11.782 GB
+Iter 243: Train loss 1.610, Learning Rate 6.389e-05, It/sec 0.180, Tokens/sec 141.183, Trained Tokens 204981, Peak mem 11.782 GB
+Iter 244: Train loss 1.560, Learning Rate 6.363e-05, It/sec 0.281, Tokens/sec 108.921, Trained Tokens 205368, Peak mem 11.782 GB
+Iter 245: Train loss 1.569, Learning Rate 6.337e-05, It/sec 0.124, Tokens/sec 126.598, Trained Tokens 206388, Peak mem 11.782 GB
+Iter 246: Train loss 1.281, Learning Rate 6.311e-05, It/sec 0.332, Tokens/sec 129.395, Trained Tokens 206778, Peak mem 11.782 GB
+Iter 247: Train loss 1.670, Learning Rate 6.284e-05, It/sec 0.092, Tokens/sec 107.997, Trained Tokens 207947, Peak mem 11.782 GB
+Iter 248: Train loss 1.578, Learning Rate 6.258e-05, It/sec 0.213, Tokens/sec 99.499, Trained Tokens 208415, Peak mem 11.782 GB
+Iter 249: Train loss 1.435, Learning Rate 6.232e-05, It/sec 0.295, Tokens/sec 140.004, Trained Tokens 208890, Peak mem 11.782 GB
+Iter 250: Train loss 1.513, Learning Rate 6.205e-05, It/sec 0.097, Tokens/sec 131.962, Trained Tokens 210257, Peak mem 11.782 GB
+Iter 251: Train loss 1.274, Learning Rate 6.179e-05, It/sec 0.221, Tokens/sec 142.926, Trained Tokens 210904, Peak mem 11.782 GB
+Iter 252: Train loss 1.523, Learning Rate 6.152e-05, It/sec 0.170, Tokens/sec 134.736, Trained Tokens 211695, Peak mem 11.782 GB
+Iter 253: Train loss 1.345, Learning Rate 6.126e-05, It/sec 0.125, Tokens/sec 103.124, Trained Tokens 212519, Peak mem 11.782 GB
+Iter 254: Train loss 1.436, Learning Rate 6.100e-05, It/sec 0.411, Tokens/sec 176.567, Trained Tokens 212949, Peak mem 11.782 GB
+Iter 255: Train loss 1.447, Learning Rate 6.073e-05, It/sec 0.137, Tokens/sec 91.668, Trained Tokens 213619, Peak mem 11.782 GB
+Iter 256: Train loss 1.104, Learning Rate 6.046e-05, It/sec 0.281, Tokens/sec 98.838, Trained Tokens 213971, Peak mem 11.782 GB
+Iter 257: Train loss 1.680, Learning Rate 6.020e-05, It/sec 0.199, Tokens/sec 149.657, Trained Tokens 214722, Peak mem 11.782 GB
+Iter 258: Train loss 1.250, Learning Rate 5.993e-05, It/sec 0.223, Tokens/sec 132.747, Trained Tokens 215318, Peak mem 11.782 GB
+Iter 259: Train loss 1.360, Learning Rate 5.967e-05, It/sec 0.114, Tokens/sec 101.960, Trained Tokens 216214, Peak mem 11.782 GB
+Iter 260: Train loss 1.179, Learning Rate 5.940e-05, It/sec 0.308, Tokens/sec 167.504, Trained Tokens 216757, Peak mem 11.782 GB
+Iter 261: Train loss 1.406, Learning Rate 5.914e-05, It/sec 0.169, Tokens/sec 172.783, Trained Tokens 217780, Peak mem 11.782 GB
+Iter 262: Train loss 1.157, Learning Rate 5.887e-05, It/sec 0.423, Tokens/sec 134.219, Trained Tokens 218097, Peak mem 11.782 GB
+Iter 263: Train loss 1.462, Learning Rate 5.860e-05, It/sec 0.255, Tokens/sec 145.077, Trained Tokens 218666, Peak mem 11.782 GB
+Iter 264: Train loss 1.762, Learning Rate 5.834e-05, It/sec 0.078, Tokens/sec 124.590, Trained Tokens 220256, Peak mem 11.782 GB
+Iter 265: Train loss 1.315, Learning Rate 5.807e-05, It/sec 0.338, Tokens/sec 143.162, Trained Tokens 220680, Peak mem 11.782 GB
+Iter 266: Train loss 0.912, Learning Rate 5.780e-05, It/sec 0.410, Tokens/sec 108.294, Trained Tokens 220944, Peak mem 11.782 GB
+Iter 267: Train loss 1.754, Learning Rate 5.754e-05, It/sec 0.076, Tokens/sec 130.289, Trained Tokens 222660, Peak mem 11.782 GB
+Iter 268: Train loss 1.414, Learning Rate 5.727e-05, It/sec 0.226, Tokens/sec 149.885, Trained Tokens 223323, Peak mem 11.782 GB
+Iter 269: Train loss 1.314, Learning Rate 5.700e-05, It/sec 0.221, Tokens/sec 168.263, Trained Tokens 224086, Peak mem 11.782 GB
+Iter 270: Train loss 1.488, Learning Rate 5.674e-05, It/sec 0.086, Tokens/sec 102.069, Trained Tokens 225273, Peak mem 11.782 GB
+Iter 271: Train loss 1.396, Learning Rate 5.647e-05, It/sec 0.167, Tokens/sec 97.747, Trained Tokens 225858, Peak mem 11.782 GB
+Iter 272: Train loss 1.443, Learning Rate 5.620e-05, It/sec 0.184, Tokens/sec 149.761, Trained Tokens 226670, Peak mem 11.782 GB
+Iter 273: Train loss 1.567, Learning Rate 5.594e-05, It/sec 0.105, Tokens/sec 138.126, Trained Tokens 227981, Peak mem 11.782 GB
+Iter 274: Train loss 1.307, Learning Rate 5.567e-05, It/sec 0.213, Tokens/sec 152.938, Trained Tokens 228700, Peak mem 11.782 GB
+Iter 275: Train loss 1.732, Learning Rate 5.540e-05, It/sec 0.356, Tokens/sec 171.426, Trained Tokens 229181, Peak mem 11.782 GB
+Iter 276: Train loss 1.738, Learning Rate 5.513e-05, It/sec 0.080, Tokens/sec 126.795, Trained Tokens 230758, Peak mem 11.782 GB
+Iter 277: Train loss 1.373, Learning Rate 5.487e-05, It/sec 0.263, Tokens/sec 174.896, Trained Tokens 231424, Peak mem 11.782 GB
+Iter 278: Train loss 1.413, Learning Rate 5.460e-05, It/sec 0.529, Tokens/sec 174.128, Trained Tokens 231753, Peak mem 11.782 GB
+Iter 279: Train loss 1.471, Learning Rate 5.433e-05, It/sec 0.250, Tokens/sec 181.529, Trained Tokens 232480, Peak mem 11.782 GB
+Iter 280: Train loss 1.737, Learning Rate 5.406e-05, It/sec 0.073, Tokens/sec 143.948, Trained Tokens 234450, Peak mem 11.782 GB
+Iter 281: Train loss 1.411, Learning Rate 5.380e-05, It/sec 0.164, Tokens/sec 134.249, Trained Tokens 235270, Peak mem 11.782 GB
+Iter 282: Train loss 1.578, Learning Rate 5.353e-05, It/sec 0.138, Tokens/sec 158.257, Trained Tokens 236414, Peak mem 11.782 GB
+Iter 283: Train loss 1.364, Learning Rate 5.326e-05, It/sec 0.116, Tokens/sec 110.453, Trained Tokens 237367, Peak mem 11.782 GB
+Iter 284: Train loss 1.383, Learning Rate 5.300e-05, It/sec 0.147, Tokens/sec 100.354, Trained Tokens 238050, Peak mem 11.782 GB
+Iter 285: Train loss 1.574, Learning Rate 5.273e-05, It/sec 0.118, Tokens/sec 129.320, Trained Tokens 239146, Peak mem 11.782 GB
+Iter 286: Train loss 1.701, Learning Rate 5.246e-05, It/sec 0.111, Tokens/sec 118.324, Trained Tokens 240210, Peak mem 11.782 GB
+Iter 287: Train loss 1.205, Learning Rate 5.220e-05, It/sec 0.336, Tokens/sec 144.649, Trained Tokens 240640, Peak mem 11.782 GB
+Iter 288: Train loss 1.313, Learning Rate 5.193e-05, It/sec 0.165, Tokens/sec 106.317, Trained Tokens 241285, Peak mem 11.782 GB
+Iter 289: Train loss 1.100, Learning Rate 5.166e-05, It/sec 0.119, Tokens/sec 70.448, Trained Tokens 241879, Peak mem 11.782 GB
+Iter 290: Train loss 1.193, Learning Rate 5.140e-05, It/sec 0.148, Tokens/sec 85.814, Trained Tokens 242458, Peak mem 11.782 GB
+Iter 291: Train loss 1.398, Learning Rate 5.113e-05, It/sec 0.073, Tokens/sec 75.792, Trained Tokens 243492, Peak mem 11.782 GB
+Iter 292: Train loss 1.420, Learning Rate 5.086e-05, It/sec 0.061, Tokens/sec 67.486, Trained Tokens 244601, Peak mem 11.782 GB
+Iter 293: Train loss 1.565, Learning Rate 5.060e-05, It/sec 0.101, Tokens/sec 97.134, Trained Tokens 245563, Peak mem 11.782 GB
+Iter 294: Train loss 1.379, Learning Rate 5.033e-05, It/sec 0.130, Tokens/sec 58.306, Trained Tokens 246010, Peak mem 11.782 GB
+Iter 295: Train loss 1.367, Learning Rate 5.007e-05, It/sec 0.113, Tokens/sec 69.855, Trained Tokens 246627, Peak mem 11.782 GB
+Iter 296: Train loss 1.738, Learning Rate 4.980e-05, It/sec 0.173, Tokens/sec 82.912, Trained Tokens 247107, Peak mem 11.782 GB
+Iter 297: Train loss 1.321, Learning Rate 4.954e-05, It/sec 0.121, Tokens/sec 120.667, Trained Tokens 248105, Peak mem 11.782 GB
+Iter 298: Train loss 1.191, Learning Rate 4.927e-05, It/sec 0.234, Tokens/sec 108.731, Trained Tokens 248570, Peak mem 11.782 GB
+Iter 299: Train loss 1.158, Learning Rate 4.900e-05, It/sec 0.122, Tokens/sec 94.183, Trained Tokens 249341, Peak mem 11.782 GB
+Calculating loss...: 0it [00:00, ?it/s]
+Iter 300: Val loss nan, Val took 0.021s
+Iter 300: Train loss 1.497, Learning Rate 4.874e-05, It/sec 0.163, Tokens/sec 154.640, Trained Tokens 250292, Peak mem 11.782 GB
+Iter 300: Saved adapter weights to adapters-uncertain/adapters.safetensors and adapters-uncertain/0000300_adapters.safetensors.
+Saved final weights to adapters-uncertain/adapters.safetensors.

adapters-uncertain/wandb/run-20260216_154932-918bwjte/files/requirements.txt

@@ -0,0 +1,147 @@
+shellingham==1.5.4
+contourpy==1.3.2
+jiter==0.12.0
+audioread==3.1.0
+threadpoolctl==3.6.0
+lazy_loader==0.4
+GitPython==3.1.45
+async-timeout==5.0.1
+requests==2.32.5
+rich==14.2.0
+tokenizers==0.22.1
+urllib3==2.5.0
+exceptiongroup==1.3.1
+numpy==2.2.6
+click==8.3.1
+pytz==2025.2
+miniaudio==1.61
+pyarrow==22.0.0
+sse-starlette==3.2.0
+scikit-learn==1.7.2
+soxr==1.0.0
+jsonschema-specifications==2025.9.1
+python-multipart==0.0.22
+utilsforecast==0.2.15
+ftfy==6.3.1
+torchvision==0.25.0
+statsmodels==0.14.6
+file-read-backwards==3.2.0
+propcache==0.4.1
+python-dotenv==1.2.1
+anyio==4.12.0
+mlx==0.30.6
+wordfreq==3.1.1
+networkx==3.4.2
+pip==25.3
+texttable==1.7.0
+mlx-audio==0.3.1
+narwhals==2.15.0
+multidict==6.7.0
+numba==0.63.1
+idna==3.11
+regex==2025.11.3
+fonttools==4.60.1
+openai==2.16.0
+aiohttp==3.13.2
+mistral_common==1.8.6
+einshape==1.0
+cffi==2.0.0
+kiwisolver==1.4.9
+tqdm==4.67.1
+setuptools==80.9.0
+RapidFuzz==3.14.3
+pyparsing==3.2.5
+starlette==0.52.1
+tzdata==2025.2
+mlx-lm==0.30.6
+httpcore==1.0.9
+decorator==5.2.1
+certifi==2025.11.12
+typer==0.21.1
+pydantic==2.12.4
+fsspec==2025.10.0
+mcp==1.26.0
+librosa==0.11.0
+charset-normalizer==3.4.4
+sympy==1.14.0
+jsonschema==4.25.1
+pydantic-settings==2.12.0
+markdown-it-py==4.0.0
+tiktoken==0.12.0
+PyJWT==2.11.0
+sentry-sdk==2.45.0
+platformdirs==4.5.0
+absl-py==2.3.1
+transformers==5.1.0
+diffusers==0.37.0.dev0
+h11==0.16.0
+gitdb==4.0.12
+sniffio==1.3.1
+pycparser==3.0
+sentencepiece==0.2.1
+importlib_metadata==8.7.1
+mdurl==0.1.2
+patsy==1.0.2
+python-dateutil==2.9.0.post0
+mpmath==1.3.0
+pillow==12.0.0
+PyYAML==6.0.3
+sentence-transformers==5.1.2
+multiprocess==0.70.18
+pydantic_core==2.41.5
+uvicorn==0.40.0
+frozenlist==1.8.0
+typer-slim==0.20.1
+typing_extensions==4.15.0
+aiosignal==1.4.0
+packaging==25.0
+cycler==0.12.1
+cryptography==46.0.4
+hf-xet==1.2.0
+Jinja2==3.1.6
+wheel==0.45.1
+referencing==0.37.0
+pandas==2.3.3
+soundfile==0.13.1
+pooch==1.8.2
+MarkupSafe==3.0.3
+dill==0.4.0
+pydantic-extra-types==2.10.6
+msgpack==1.1.2
+distro==1.9.0
+locate==1.1.1
+datasets==4.4.1
+Pygments==2.19.2
+aiohappyeyeballs==2.6.1
+llvmlite==0.46.0
+attrs==25.4.0
+huggingface_hub==1.3.5
+nltk==3.9.2
+torch==2.10.0
+httpx==0.28.1
+filelock==3.20.0
+smmap==5.0.2
+sounddevice==0.5.3
+timesfm==1.3.0
+pycountry==24.6.1
+mlx-metal==0.30.6
+scipy==1.15.3
+protobuf==6.33.1
+psutil==7.1.3
+typing-inspection==0.4.2
+joblib==1.5.2
+zipp==3.23.0
+annotated-types==0.7.0
+accelerate==1.12.0
+safetensors==0.6.2
+httpx-sse==0.4.3
+wcwidth==0.2.14
+igraph==1.0.0
+rpds-py==0.30.0
+langcodes==3.5.1
+six==1.17.0
+wandb==0.23.0
+yarl==1.22.0
+pyloudnorm==0.2.0
+xxhash==3.6.0
+matplotlib==3.10.7

adapters-uncertain/wandb/run-20260216_154932-918bwjte/files/wandb-metadata.json

@@ -0,0 +1,35 @@
+{
+  "os": "macOS-26.2-arm64-arm-64bit",
+  "python": "CPython 3.10.19",
+  "startedAt": "2026-02-16T20:49:32.288869Z",
+  "args": [
+    "--config",
+    "train.yaml"
+  ],
+  "program": "/Users/natebreslow/miniconda3/envs/mlx-experiment/bin/mlx_lm.lora",
+  "email": "nathanbreslow@gmail.com",
+  "root": "adapters-uncertain",
+  "host": "MacBook-Pro-135.local",
+  "executable": "/Users/natebreslow/miniconda3/envs/mlx-experiment/bin/python3.10",
+  "cpu_count": 16,
+  "cpu_count_logical": 16,
+  "disk": {
+    "/": {
+      "total": "1995218165760",
+      "used": "1702686375936"
+    }
+  },
+  "memory": {
+    "total": "68719476736"
+  },
+  "apple": {
+    "name": "Apple M3 Max",
+    "ecpuCores": 4,
+    "pcpuCores": 12,
+    "gpuCores": 40,
+    "memoryGb": 64,
+    "swapTotalBytes": "5368709120",
+    "ramTotalBytes": "68719476736"
+  },
+  "writerId": "4stii5owh2cve1gvc5kugyr75h6wj8db"
+}

adapters-uncertain/wandb/run-20260216_154932-918bwjte/files/wandb-summary.json

@@ -0,0 +1 @@
+{"train_loss":1.4971740245819092,"_runtime":1875.19269225,"iterations_per_second":0.16260756296367251,"_step":300,"val_loss":NaN,"_timestamp":1.77127684632109e+09,"tokens_per_second":154.63979237845257,"_wandb":{"runtime":1875},"iteration":300,"peak_memory":11.781823884,"val_time":0.020949832993210293,"learning_rate":4.8740144848125055e-05,"trained_tokens":250292}

adapters-uncertain/wandb/run-20260216_154932-918bwjte/logs/debug-internal.log

@@ -0,0 +1,11 @@
+{"time":"2026-02-16T15:49:32.584829-05:00","level":"INFO","msg":"stream: starting","core version":"0.23.0"}
+{"time":"2026-02-16T15:49:32.835831-05:00","level":"INFO","msg":"stream: created new stream","id":"918bwjte"}
+{"time":"2026-02-16T15:49:32.835901-05:00","level":"INFO","msg":"handler: started","stream_id":"918bwjte"}
+{"time":"2026-02-16T15:49:32.836119-05:00","level":"INFO","msg":"stream: started","id":"918bwjte"}
+{"time":"2026-02-16T15:49:32.836134-05:00","level":"INFO","msg":"sender: started","stream_id":"918bwjte"}
+{"time":"2026-02-16T15:49:32.836138-05:00","level":"INFO","msg":"writer: started","stream_id":"918bwjte"}
+{"time":"2026-02-16T16:20:48.46164-05:00","level":"INFO","msg":"stream: closing","id":"918bwjte"}
+{"time":"2026-02-16T16:20:48.851376-05:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2026-02-16T16:20:48.997596-05:00","level":"INFO","msg":"handler: closed","stream_id":"918bwjte"}
+{"time":"2026-02-16T16:20:48.997985-05:00","level":"INFO","msg":"sender: closed","stream_id":"918bwjte"}
+{"time":"2026-02-16T16:20:48.998039-05:00","level":"INFO","msg":"stream: closed","id":"918bwjte"}

adapters-uncertain/wandb/run-20260216_154932-918bwjte/logs/debug.log

@@ -0,0 +1,23 @@
+2026-02-16 15:49:32,296 INFO    MainThread:54732 [wandb_setup.py:_flush():80] Current SDK version is 0.23.0
+2026-02-16 15:49:32,296 INFO    MainThread:54732 [wandb_setup.py:_flush():80] Configure stats pid to 54732
+2026-02-16 15:49:32,296 INFO    MainThread:54732 [wandb_setup.py:_flush():80] Loading settings from /Users/natebreslow/.config/wandb/settings
+2026-02-16 15:49:32,296 INFO    MainThread:54732 [wandb_setup.py:_flush():80] Loading settings from /Users/natebreslow/Documents/llmSelfReport/wandb/settings
+2026-02-16 15:49:32,296 INFO    MainThread:54732 [wandb_setup.py:_flush():80] Loading settings from environment variables
+2026-02-16 15:49:32,296 INFO    MainThread:54732 [wandb_init.py:setup_run_log_directory():713] Logging user logs to adapters-uncertain/wandb/run-20260216_154932-918bwjte/logs/debug.log
+2026-02-16 15:49:32,296 INFO    MainThread:54732 [wandb_init.py:setup_run_log_directory():714] Logging internal logs to adapters-uncertain/wandb/run-20260216_154932-918bwjte/logs/debug-internal.log
+2026-02-16 15:49:32,297 INFO    MainThread:54732 [wandb_init.py:init():840] calling init triggers
+2026-02-16 15:49:32,297 INFO    MainThread:54732 [wandb_init.py:init():845] wandb.init called with sweep_config: {}
+config: {'model': 'Qwen3-4B-Instruct-2507', 'train': True, 'data': 'training/uncertain', 'fine_tune_type': 'lora', 'optimizer': 'adam', 'mask_prompt': False, 'num_layers': 36, 'batch_size': 1, 'iters': 300, 'val_batches': 0, 'learning_rate': 0.0001, 'steps_per_report': 1, 'steps_per_eval': 200, 'grad_accumulation_steps': 1, 'resume_adapter_file': None, 'adapter_path': 'adapters-uncertain', 'save_every': 100, 'test': False, 'test_batches': 100, 'max_seq_length': 8192, 'config': 'train.yaml', 'grad_checkpoint': True, 'report_to': 'wandb', 'project_name': 'conscious-finetuning', 'seed': 0, 'optimizer_config': {'adam': {'betas': [0.9, 0.9999], 'eps': 1e-06, 'bias_correction': True}}, 'lora_parameters': {'keys': ['self_attn.q_proj', 'self_attn.v_proj', 'self_attn.k_proj', 'self_attn.o_proj', 'mlp.gate_proj', 'mlp.up_proj', 'mlp.down_proj'], 'rank': 16, 'scale': 2.0, 'dropout': 0.0}, 'lr_schedule': {'name': 'cosine_decay', 'warmup': 10, 'warmup_init': 1e-05, 'arguments': [0.0001, 529, 1e-05]}, '_wandb': {}}
+2026-02-16 15:49:32,297 INFO    MainThread:54732 [wandb_init.py:init():888] starting backend
+2026-02-16 15:49:32,540 INFO    MainThread:54732 [wandb_init.py:init():891] sending inform_init request
+2026-02-16 15:49:32,583 INFO    MainThread:54732 [wandb_init.py:init():899] backend started and connected
+2026-02-16 15:49:32,586 INFO    MainThread:54732 [wandb_init.py:init():969] updated telemetry
+2026-02-16 15:49:32,586 INFO    MainThread:54732 [wandb_init.py:init():993] communicating run to backend with 90.0 second timeout
+2026-02-16 15:49:33,237 INFO    MainThread:54732 [wandb_init.py:init():1040] starting run threads in backend
+2026-02-16 15:49:33,347 INFO    MainThread:54732 [wandb_run.py:_console_start():2504] atexit reg
+2026-02-16 15:49:33,347 INFO    MainThread:54732 [wandb_run.py:_redirect():2352] redirect: wrap_raw
+2026-02-16 15:49:33,347 INFO    MainThread:54732 [wandb_run.py:_redirect():2421] Wrapping output streams.
+2026-02-16 15:49:33,348 INFO    MainThread:54732 [wandb_run.py:_redirect():2444] Redirects installed.
+2026-02-16 15:49:33,350 INFO    MainThread:54732 [wandb_init.py:init():1080] run started, returning control to user process
+2026-02-16 16:20:48,459 INFO    wandb-AsyncioManager-main:54732 [service_client.py:_forward_responses():80] Reached EOF.
+2026-02-16 16:20:48,460 INFO    wandb-AsyncioManager-main:54732 [mailbox.py:close():137] Closing mailbox, abandoning 1 handles.

adapters-uncertain/wandb/run-20260216_154932-918bwjte/run-918bwjte.wandb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e76ab5455644c83a067107fc32a7bd8fb8f7417fd797d14daf4c0325b831ac45
+size 416192