add midtraining/coin/checkpoint-30 (copied from jbostock/scimt-dispatch-models-v1)

midtraining/coin/checkpoint-30/config.json

@@ -0,0 +1,125 @@
+{
+  "architectures": [
+    "Gemma3ForConditionalGeneration"
+  ],
+  "boi_token_index": 255999,
+  "bos_token_id": 2,
+  "dtype": "bfloat16",
+  "eoi_token_index": 256000,
+  "eos_token_id": 1,
+  "image_token_index": 262144,
+  "initializer_range": 0.02,
+  "mm_tokens_per_image": 256,
+  "model_type": "gemma3",
+  "pad_token_id": 0,
+  "text_config": {
+    "_sliding_window_pattern": 6,
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "attn_logit_softcapping": null,
+    "bos_token_id": 2,
+    "cache_implementation": "hybrid",
+    "dtype": "bfloat16",
+    "eos_token_id": 1,
+    "final_logit_softcapping": null,
+    "head_dim": 256,
+    "hidden_activation": "gelu_pytorch_tanh",
+    "hidden_size": 3840,
+    "initializer_range": 0.02,
+    "intermediate_size": 15360,
+    "layer_types": [
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention"
+    ],
+    "max_position_embeddings": 131072,
+    "model_type": "gemma3_text",
+    "num_attention_heads": 16,
+    "num_hidden_layers": 48,
+    "num_key_value_heads": 8,
+    "pad_token_id": 0,
+    "query_pre_attn_scalar": 256,
+    "rms_norm_eps": 1e-06,
+    "rope_parameters": {
+      "full_attention": {
+        "factor": 8.0,
+        "rope_theta": 1000000.0,
+        "rope_type": "linear"
+      },
+      "sliding_attention": {
+        "rope_theta": 10000.0,
+        "rope_type": "default"
+      }
+    },
+    "sliding_window": 1024,
+    "sliding_window_pattern": 6,
+    "tie_word_embeddings": true,
+    "use_bidirectional_attention": false,
+    "use_cache": false,
+    "vocab_size": 262208
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.9.0",
+  "unsloth_fixed": true,
+  "use_cache": false,
+  "vision_config": {
+    "attention_dropout": 0.0,
+    "dtype": "bfloat16",
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 1152,
+    "image_size": 896,
+    "intermediate_size": 4304,
+    "layer_norm_eps": 1e-06,
+    "model_type": "siglip_vision_model",
+    "num_attention_heads": 16,
+    "num_channels": 3,
+    "num_hidden_layers": 27,
+    "patch_size": 14,
+    "vision_use_head": false
+  }
+}

midtraining/coin/checkpoint-30/generation_config.json

@@ -0,0 +1,13 @@
+{
+  "bos_token_id": 2,
+  "cache_implementation": "hybrid",
+  "do_sample": true,
+  "eos_token_id": [
+    1,
+    106
+  ],
+  "pad_token_id": 0,
+  "top_k": 64,
+  "top_p": 0.95,
+  "transformers_version": "5.9.0"
+}

midtraining/coin/checkpoint-30/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11bea1c166e12d953fee8894f19a0f068b935e1fdaef6c8b5a5af2601aa714c2
+size 26388552360

midtraining/coin/checkpoint-30/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726
+size 33384567

midtraining/coin/checkpoint-30/tokenizer_config.json

@@ -0,0 +1,26 @@
+{
+  "backend": "tokenizers",
+  "boi_token": "<start_of_image>",
+  "bos_token": "<bos>",
+  "clean_up_tokenization_spaces": false,
+  "eoi_token": "<end_of_image>",
+  "eos_token": "<eos>",
+  "image_token": "<image_soft_token>",
+  "is_local": true,
+  "local_files_only": false,
+  "mask_token": "<mask>",
+  "model_max_length": 131072,
+  "model_specific_special_tokens": {
+    "boi_token": "<start_of_image>",
+    "eoi_token": "<end_of_image>",
+    "image_token": "<image_soft_token>"
+  },
+  "pad_token": "<pad>",
+  "padding_side": "left",
+  "processor_class": "Gemma3Processor",
+  "sp_model_kwargs": null,
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "GemmaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

midtraining/coin/checkpoint-30/tokens_state.json

@@ -0,0 +1 @@
+{"total": 7864320, "trainable": 7845196}

midtraining/coin/checkpoint-30/trainer_state.json

@@ -0,0 +1,454 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9836065573770492,
+  "eval_steps": 500,
+  "global_step": 30,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.03278688524590164,
+      "grad_norm": 4.125,
+      "learning_rate": 1e-05,
+      "loss": 1.7216796875,
+      "memory/device_reserved (GiB)": 44.51,
+      "memory/max_active (GiB)": 38.42,
+      "memory/max_allocated (GiB)": 38.42,
+      "ppl": 5.59392,
+      "step": 1,
+      "tokens/total": 262144,
+      "tokens/train_per_sec_per_gpu": 276.38,
+      "tokens/trainable": 261694
+    },
+    {
+      "epoch": 0.06557377049180328,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.97534852915723e-06,
+      "loss": 1.4434814453125,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 4.23542,
+      "step": 2,
+      "tokens/total": 524288,
+      "tokens/train_per_sec_per_gpu": 415.01,
+      "tokens/trainable": 523214
+    },
+    {
+      "epoch": 0.09836065573770492,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.901664203302126e-06,
+      "loss": 1.373046875,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.94736,
+      "step": 3,
+      "tokens/total": 786432,
+      "tokens/train_per_sec_per_gpu": 353.75,
+      "tokens/trainable": 784877
+    },
+    {
+      "epoch": 0.13114754098360656,
+      "grad_norm": 2.015625,
+      "learning_rate": 9.779754323328192e-06,
+      "loss": 1.2962646484375,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.65562,
+      "step": 4,
+      "tokens/total": 1048576,
+      "tokens/train_per_sec_per_gpu": 431.18,
+      "tokens/trainable": 1046550
+    },
+    {
+      "epoch": 0.16393442622950818,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.610954559391704e-06,
+      "loss": 1.2607421875,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.52804,
+      "step": 5,
+      "tokens/total": 1310720,
+      "tokens/train_per_sec_per_gpu": 431.9,
+      "tokens/trainable": 1308077
+    },
+    {
+      "epoch": 0.19672131147540983,
+      "grad_norm": 2.75,
+      "learning_rate": 9.397114317029975e-06,
+      "loss": 1.3692626953125,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.93245,
+      "step": 6,
+      "tokens/total": 1572864,
+      "tokens/train_per_sec_per_gpu": 429.84,
+      "tokens/trainable": 1569690
+    },
+    {
+      "epoch": 0.22950819672131148,
+      "grad_norm": 16.5,
+      "learning_rate": 9.140576474687263e-06,
+      "loss": 1.285400390625,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.61612,
+      "step": 7,
+      "tokens/total": 1835008,
+      "tokens/train_per_sec_per_gpu": 429.24,
+      "tokens/trainable": 1831313
+    },
+    {
+      "epoch": 0.26229508196721313,
+      "grad_norm": 1.65625,
+      "learning_rate": 8.844151714648274e-06,
+      "loss": 1.201416015625,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.32482,
+      "step": 8,
+      "tokens/total": 2097152,
+      "tokens/train_per_sec_per_gpu": 431.29,
+      "tokens/trainable": 2092815
+    },
+    {
+      "epoch": 0.29508196721311475,
+      "grad_norm": 1.25,
+      "learning_rate": 8.511087728614863e-06,
+      "loss": 1.2529296875,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.50058,
+      "step": 9,
+      "tokens/total": 2359296,
+      "tokens/train_per_sec_per_gpu": 429.41,
+      "tokens/trainable": 2354504
+    },
+    {
+      "epoch": 0.32786885245901637,
+      "grad_norm": 1.921875,
+      "learning_rate": 8.14503363531613e-06,
+      "loss": 1.318603515625,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.7382,
+      "step": 10,
+      "tokens/total": 2621440,
+      "tokens/train_per_sec_per_gpu": 430.68,
+      "tokens/trainable": 2616227
+    },
+    {
+      "epoch": 0.36065573770491804,
+      "grad_norm": 8.1875,
+      "learning_rate": 7.75e-06,
+      "loss": 1.2261962890625,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.40824,
+      "step": 11,
+      "tokens/total": 2883584,
+      "tokens/train_per_sec_per_gpu": 432.42,
+      "tokens/trainable": 2877800
+    },
+    {
+      "epoch": 0.39344262295081966,
+      "grad_norm": 1.6328125,
+      "learning_rate": 7.330314893841102e-06,
+      "loss": 1.160400390625,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.19121,
+      "step": 12,
+      "tokens/total": 3145728,
+      "tokens/train_per_sec_per_gpu": 430.71,
+      "tokens/trainable": 3139265
+    },
+    {
+      "epoch": 0.4262295081967213,
+      "grad_norm": 1.5703125,
+      "learning_rate": 6.890576474687264e-06,
+      "loss": 1.1781005859375,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.2482,
+      "step": 13,
+      "tokens/total": 3407872,
+      "tokens/train_per_sec_per_gpu": 426.89,
+      "tokens/trainable": 3400808
+    },
+    {
+      "epoch": 0.45901639344262296,
+      "grad_norm": 1.34375,
+      "learning_rate": 6.4356026086799176e-06,
+      "loss": 1.11474609375,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.04879,
+      "step": 14,
+      "tokens/total": 3670016,
+      "tokens/train_per_sec_per_gpu": 428.89,
+      "tokens/trainable": 3662502
+    },
+    {
+      "epoch": 0.4918032786885246,
+      "grad_norm": 1.6640625,
+      "learning_rate": 5.970378084704441e-06,
+      "loss": 1.265625,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.54531,
+      "step": 15,
+      "tokens/total": 3932160,
+      "tokens/train_per_sec_per_gpu": 425.96,
+      "tokens/trainable": 3924106
+    },
+    {
+      "epoch": 0.5245901639344263,
+      "grad_norm": 2.0,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 1.1568603515625,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.17993,
+      "step": 16,
+      "tokens/total": 4194304,
+      "tokens/train_per_sec_per_gpu": 429.31,
+      "tokens/trainable": 4185690
+    },
+    {
+      "epoch": 0.5573770491803278,
+      "grad_norm": 1.3046875,
+      "learning_rate": 5.02962191529556e-06,
+      "loss": 1.1439208984375,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.13905,
+      "step": 17,
+      "tokens/total": 4456448,
+      "tokens/train_per_sec_per_gpu": 433.86,
+      "tokens/trainable": 4447163
+    },
+    {
+      "epoch": 0.5901639344262295,
+      "grad_norm": 1.2734375,
+      "learning_rate": 4.564397391320085e-06,
+      "loss": 1.2650146484375,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.54314,
+      "step": 18,
+      "tokens/total": 4718592,
+      "tokens/train_per_sec_per_gpu": 424.85,
+      "tokens/trainable": 4708799
+    },
+    {
+      "epoch": 0.6229508196721312,
+      "grad_norm": 8.0,
+      "learning_rate": 4.109423525312738e-06,
+      "loss": 1.3094482421875,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.70413,
+      "step": 19,
+      "tokens/total": 4980736,
+      "tokens/train_per_sec_per_gpu": 430.27,
+      "tokens/trainable": 4970392
+    },
+    {
+      "epoch": 0.6557377049180327,
+      "grad_norm": 1.9296875,
+      "learning_rate": 3.6696851061589e-06,
+      "loss": 1.29443359375,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.64893,
+      "step": 20,
+      "tokens/total": 5242880,
+      "tokens/train_per_sec_per_gpu": 427.2,
+      "tokens/trainable": 5231877
+    },
+    {
+      "epoch": 0.6885245901639344,
+      "grad_norm": 1.109375,
+      "learning_rate": 3.2500000000000015e-06,
+      "loss": 1.2117919921875,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.3595,
+      "step": 21,
+      "tokens/total": 5505024,
+      "tokens/train_per_sec_per_gpu": 432.8,
+      "tokens/trainable": 5493151
+    },
+    {
+      "epoch": 0.7213114754098361,
+      "grad_norm": 1.0234375,
+      "learning_rate": 2.854966364683872e-06,
+      "loss": 1.184814453125,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.27008,
+      "step": 22,
+      "tokens/total": 5767168,
+      "tokens/train_per_sec_per_gpu": 434.35,
+      "tokens/trainable": 5754418
+    },
+    {
+      "epoch": 0.7540983606557377,
+      "grad_norm": 4.8125,
+      "learning_rate": 2.4889122713851397e-06,
+      "loss": 1.1204833984375,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.06634,
+      "step": 23,
+      "tokens/total": 6029312,
+      "tokens/train_per_sec_per_gpu": 434.03,
+      "tokens/trainable": 6015781
+    },
+    {
+      "epoch": 0.7868852459016393,
+      "grad_norm": 1.3671875,
+      "learning_rate": 2.1558482853517257e-06,
+      "loss": 1.1492919921875,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.15596,
+      "step": 24,
+      "tokens/total": 6291456,
+      "tokens/train_per_sec_per_gpu": 429.26,
+      "tokens/trainable": 6277313
+    },
+    {
+      "epoch": 0.819672131147541,
+      "grad_norm": 0.921875,
+      "learning_rate": 1.8594235253127373e-06,
+      "loss": 1.103759765625,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.01548,
+      "step": 25,
+      "tokens/total": 6553600,
+      "tokens/train_per_sec_per_gpu": 431.85,
+      "tokens/trainable": 6538845
+    },
+    {
+      "epoch": 0.8524590163934426,
+      "grad_norm": 1.25,
+      "learning_rate": 1.602885682970026e-06,
+      "loss": 1.1715087890625,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.22686,
+      "step": 26,
+      "tokens/total": 6815744,
+      "tokens/train_per_sec_per_gpu": 432.86,
+      "tokens/trainable": 6800364
+    },
+    {
+      "epoch": 0.8852459016393442,
+      "grad_norm": 1.046875,
+      "learning_rate": 1.389045440608296e-06,
+      "loss": 1.138427734375,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.12186,
+      "step": 27,
+      "tokens/total": 7077888,
+      "tokens/train_per_sec_per_gpu": 432.08,
+      "tokens/trainable": 7061953
+    },
+    {
+      "epoch": 0.9180327868852459,
+      "grad_norm": 1.3671875,
+      "learning_rate": 1.2202456766718092e-06,
+      "loss": 1.2476806640625,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.48226,
+      "step": 28,
+      "tokens/total": 7340032,
+      "tokens/train_per_sec_per_gpu": 427.42,
+      "tokens/trainable": 7323200
+    },
+    {
+      "epoch": 0.9508196721311475,
+      "grad_norm": 1.7578125,
+      "learning_rate": 1.0983357966978747e-06,
+      "loss": 1.1865234375,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.27567,
+      "step": 29,
+      "tokens/total": 7602176,
+      "tokens/train_per_sec_per_gpu": 430.16,
+      "tokens/trainable": 7584300
+    },
+    {
+      "epoch": 0.9836065573770492,
+      "grad_norm": 1.5546875,
+      "learning_rate": 1.0246514708427703e-06,
+      "loss": 1.199462890625,
+      "memory/device_reserved (GiB)": 49.98,
+      "memory/max_active (GiB)": 43.9,
+      "memory/max_allocated (GiB)": 43.9,
+      "ppl": 3.31833,
+      "step": 30,
+      "tokens/total": 7864320,
+      "tokens/train_per_sec_per_gpu": 427.47,
+      "tokens/trainable": 7845196
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 30,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.273369972834304e+17,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}