add sdf/4x/shared/post_dolci90 (copied from jbostock/scimt-dispatch-midtrained-sft-v1)

.gitattributes

@@ -131,3 +131,4 @@ sdf/1x/charter/final/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 sdf/1x/coin/final/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 sdf/4x/coin/final/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 sdf/4x/charter/final/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+sdf/4x/shared/post_dolci90/tokenizer.json filter=lfs diff=lfs merge=lfs -text

sdf/4x/shared/post_dolci90/chat_template.jinja

@@ -0,0 +1,47 @@
+{{ bos_token }}
+{%- if messages[0]['role'] == 'system' -%}
+    {%- if messages[0]['content'] is string -%}
+        {%- set first_user_prefix = messages[0]['content'] + '
+
+' -%}
+    {%- else -%}
+        {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
+
+' -%}
+    {%- endif -%}
+    {%- set loop_messages = messages[1:] -%}
+{%- else -%}
+    {%- set first_user_prefix = "" -%}
+    {%- set loop_messages = messages -%}
+{%- endif -%}
+{%- for message in loop_messages -%}
+    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
+        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
+    {%- endif -%}
+    {%- if (message['role'] == 'assistant') -%}
+        {%- set role = "model" -%}
+    {%- else -%}
+        {%- set role = message['role'] -%}
+    {%- endif -%}
+    {{ '<start_of_turn>' + role + '
+' + (first_user_prefix if loop.first else "") }}
+    {%- if message['content'] is string -%}
+        {{ message['content'] | trim }}
+    {%- elif message['content'] is iterable -%}
+        {%- for item in message['content'] -%}
+            {%- if item['type'] == 'image' -%}
+                {{ '<start_of_image>' }}
+            {%- elif item['type'] == 'text' -%}
+                {{ item['text'] | trim }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{ raise_exception("Invalid content type") }}
+    {%- endif -%}
+    {{ '<end_of_turn>
+' }}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{'<start_of_turn>model
+'}}
+{%- endif -%}

sdf/4x/shared/post_dolci90/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
+  }
+}

sdf/4x/shared/post_dolci90/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"
+}

sdf/4x/shared/post_dolci90/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:957e08aeb2dcc9591d27221c1e952a0858aeba5c2435f69c073a2401c4d9dabc
+size 26388552360

sdf/4x/shared/post_dolci90/preprocessor_config.json

@@ -0,0 +1,29 @@
+{
+  "do_convert_rgb": null,
+  "do_normalize": true,
+  "do_pan_and_scan": null,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "image_processor_type": "Gemma3ImageProcessor",
+  "image_seq_length": 256,
+  "image_std": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "pan_and_scan_max_num_crops": null,
+  "pan_and_scan_min_crop_size": null,
+  "pan_and_scan_min_ratio_to_activate": null,
+  "processor_class": "Gemma3Processor",
+  "resample": 2,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "height": 896,
+    "width": 896
+  }
+}

sdf/4x/shared/post_dolci90/processor_config.json

@@ -0,0 +1,4 @@
+{
+  "image_seq_length": 256,
+  "processor_class": "Gemma3Processor"
+}

sdf/4x/shared/post_dolci90/tokenizer.json

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

sdf/4x/shared/post_dolci90/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
+}

sdf/4x/shared/post_dolci90/tokens_state.json

@@ -0,0 +1 @@
+{"total": 89858048, "trainable": 55922724}

sdf/4x/shared/post_dolci90/trainer_state.json

@@ -0,0 +1,636 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0238805970149254,
+  "eval_steps": 500,
+  "global_step": 43,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.023880597014925373,
+      "grad_norm": 3.625,
+      "learning_rate": 0.0,
+      "loss": 0.93505859375,
+      "memory/device_reserved (GiB)": 78.29,
+      "memory/max_active (GiB)": 68.81,
+      "memory/max_allocated (GiB)": 68.81,
+      "ppl": 2.54736,
+      "step": 1,
+      "tokens/total": 2097152,
+      "tokens/train_per_sec_per_gpu": 428.06,
+      "tokens/trainable": 1314465
+    },
+    {
+      "epoch": 0.04776119402985075,
+      "grad_norm": 3.625,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 0.9598388671875,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.61128,
+      "step": 2,
+      "tokens/total": 4194304,
+      "tokens/train_per_sec_per_gpu": 433.21,
+      "tokens/trainable": 2607511
+    },
+    {
+      "epoch": 0.07164179104477612,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.9541015625,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.59634,
+      "step": 3,
+      "tokens/total": 6291456,
+      "tokens/train_per_sec_per_gpu": 428.04,
+      "tokens/trainable": 3914249
+    },
+    {
+      "epoch": 0.0955223880597015,
+      "grad_norm": 1.828125,
+      "learning_rate": 1e-05,
+      "loss": 0.8822021484375,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.41621,
+      "step": 4,
+      "tokens/total": 8388608,
+      "tokens/train_per_sec_per_gpu": 467.72,
+      "tokens/trainable": 5204512
+    },
+    {
+      "epoch": 0.11940298507462686,
+      "grad_norm": 10.1875,
+      "learning_rate": 9.986128001799077e-06,
+      "loss": 0.9356689453125,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.54892,
+      "step": 5,
+      "tokens/total": 10485760,
+      "tokens/train_per_sec_per_gpu": 461.18,
+      "tokens/trainable": 6499593
+    },
+    {
+      "epoch": 0.14328358208955225,
+      "grad_norm": 5.375,
+      "learning_rate": 9.94459753267812e-06,
+      "loss": 0.9561767578125,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.60173,
+      "step": 6,
+      "tokens/total": 12582912,
+      "tokens/train_per_sec_per_gpu": 447.41,
+      "tokens/trainable": 7821749
+    },
+    {
+      "epoch": 0.16716417910447762,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.875664641789545e-06,
+      "loss": 0.9095458984375,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.48319,
+      "step": 7,
+      "tokens/total": 14680064,
+      "tokens/train_per_sec_per_gpu": 455.23,
+      "tokens/trainable": 9114720
+    },
+    {
+      "epoch": 0.191044776119403,
+      "grad_norm": 1.8203125,
+      "learning_rate": 9.779754323328192e-06,
+      "loss": 0.828369140625,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.28958,
+      "step": 8,
+      "tokens/total": 16777216,
+      "tokens/train_per_sec_per_gpu": 446.49,
+      "tokens/trainable": 10430181
+    },
+    {
+      "epoch": 0.21492537313432836,
+      "grad_norm": 1.2890625,
+      "learning_rate": 9.65745789630079e-06,
+      "loss": 0.8316650390625,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.29714,
+      "step": 9,
+      "tokens/total": 18874368,
+      "tokens/train_per_sec_per_gpu": 433.91,
+      "tokens/trainable": 11726037
+    },
+    {
+      "epoch": 0.23880597014925373,
+      "grad_norm": 1.234375,
+      "learning_rate": 9.509529358847655e-06,
+      "loss": 0.8133544921875,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.25546,
+      "step": 10,
+      "tokens/total": 20971520,
+      "tokens/train_per_sec_per_gpu": 455.01,
+      "tokens/trainable": 13032167
+    },
+    {
+      "epoch": 0.2626865671641791,
+      "grad_norm": 0.95703125,
+      "learning_rate": 9.336880739593415e-06,
+      "loss": 0.8099365234375,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.24777,
+      "step": 11,
+      "tokens/total": 23068672,
+      "tokens/train_per_sec_per_gpu": 454.39,
+      "tokens/trainable": 14355344
+    },
+    {
+      "epoch": 0.2865671641791045,
+      "grad_norm": 0.8828125,
+      "learning_rate": 9.140576474687263e-06,
+      "loss": 0.8408203125,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.31827,
+      "step": 12,
+      "tokens/total": 25165824,
+      "tokens/train_per_sec_per_gpu": 409.75,
+      "tokens/trainable": 15638687
+    },
+    {
+      "epoch": 0.31044776119402984,
+      "grad_norm": 0.69921875,
+      "learning_rate": 8.92182684520014e-06,
+      "loss": 0.805908203125,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.23873,
+      "step": 13,
+      "tokens/total": 27262976,
+      "tokens/train_per_sec_per_gpu": 440.96,
+      "tokens/trainable": 16935892
+    },
+    {
+      "epoch": 0.33432835820895523,
+      "grad_norm": 0.79296875,
+      "learning_rate": 8.681980515339464e-06,
+      "loss": 0.8111572265625,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.25051,
+      "step": 14,
+      "tokens/total": 29360128,
+      "tokens/train_per_sec_per_gpu": 456.73,
+      "tokens/trainable": 18267440
+    },
+    {
+      "epoch": 0.3582089552238806,
+      "grad_norm": 0.8046875,
+      "learning_rate": 8.422516217485826e-06,
+      "loss": 0.8154296875,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.26015,
+      "step": 15,
+      "tokens/total": 31457280,
+      "tokens/train_per_sec_per_gpu": 438.41,
+      "tokens/trainable": 19570464
+    },
+    {
+      "epoch": 0.382089552238806,
+      "grad_norm": 0.87109375,
+      "learning_rate": 8.14503363531613e-06,
+      "loss": 0.8209228515625,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.2726,
+      "step": 16,
+      "tokens/total": 33554432,
+      "tokens/train_per_sec_per_gpu": 448.22,
+      "tokens/trainable": 20888984
+    },
+    {
+      "epoch": 0.4059701492537313,
+      "grad_norm": 0.6171875,
+      "learning_rate": 7.85124354122177e-06,
+      "loss": 0.8004150390625,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.22646,
+      "step": 17,
+      "tokens/total": 35651584,
+      "tokens/train_per_sec_per_gpu": 418.12,
+      "tokens/trainable": 22190854
+    },
+    {
+      "epoch": 0.4298507462686567,
+      "grad_norm": 0.625,
+      "learning_rate": 7.5429572488279615e-06,
+      "loss": 0.7939453125,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.21211,
+      "step": 18,
+      "tokens/total": 37748736,
+      "tokens/train_per_sec_per_gpu": 439.32,
+      "tokens/trainable": 23506808
+    },
+    {
+      "epoch": 0.4537313432835821,
+      "grad_norm": 0.7265625,
+      "learning_rate": 7.222075445642904e-06,
+      "loss": 0.7740478515625,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.16853,
+      "step": 19,
+      "tokens/total": 39845888,
+      "tokens/train_per_sec_per_gpu": 438.42,
+      "tokens/trainable": 24808674
+    },
+    {
+      "epoch": 0.47761194029850745,
+      "grad_norm": 0.6640625,
+      "learning_rate": 6.890576474687264e-06,
+      "loss": 0.775390625,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.17144,
+      "step": 20,
+      "tokens/total": 41943040,
+      "tokens/train_per_sec_per_gpu": 433.3,
+      "tokens/trainable": 26100334
+    },
+    {
+      "epoch": 0.5014925373134328,
+      "grad_norm": 0.6328125,
+      "learning_rate": 6.550504137351576e-06,
+      "loss": 0.7974853515625,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.21995,
+      "step": 21,
+      "tokens/total": 44040192,
+      "tokens/train_per_sec_per_gpu": 448.18,
+      "tokens/trainable": 27394748
+    },
+    {
+      "epoch": 0.5253731343283582,
+      "grad_norm": 0.56640625,
+      "learning_rate": 6.20395509268104e-06,
+      "loss": 0.82080078125,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.27232,
+      "step": 22,
+      "tokens/total": 46137344,
+      "tokens/train_per_sec_per_gpu": 415.02,
+      "tokens/trainable": 28676286
+    },
+    {
+      "epoch": 0.5492537313432836,
+      "grad_norm": 0.5078125,
+      "learning_rate": 5.8530659307753034e-06,
+      "loss": 0.8026123046875,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.23136,
+      "step": 23,
+      "tokens/total": 48234496,
+      "tokens/train_per_sec_per_gpu": 440.44,
+      "tokens/trainable": 29991010
+    },
+    {
+      "epoch": 0.573134328358209,
+      "grad_norm": 1.125,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 0.80419921875,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.23491,
+      "step": 24,
+      "tokens/total": 50331648,
+      "tokens/train_per_sec_per_gpu": 458.97,
+      "tokens/trainable": 31318096
+    },
+    {
+      "epoch": 0.5970149253731343,
+      "grad_norm": 2.265625,
+      "learning_rate": 5.1469340692247e-06,
+      "loss": 0.8094482421875,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.24667,
+      "step": 25,
+      "tokens/total": 52428800,
+      "tokens/train_per_sec_per_gpu": 440.5,
+      "tokens/trainable": 32619124
+    },
+    {
+      "epoch": 0.6208955223880597,
+      "grad_norm": 0.46875,
+      "learning_rate": 4.796044907318961e-06,
+      "loss": 0.8017578125,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.22946,
+      "step": 26,
+      "tokens/total": 54525952,
+      "tokens/train_per_sec_per_gpu": 437.5,
+      "tokens/trainable": 33928556
+    },
+    {
+      "epoch": 0.6447761194029851,
+      "grad_norm": 0.67578125,
+      "learning_rate": 4.449495862648427e-06,
+      "loss": 0.7806396484375,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.18287,
+      "step": 27,
+      "tokens/total": 56623104,
+      "tokens/train_per_sec_per_gpu": 436.58,
+      "tokens/trainable": 35236016
+    },
+    {
+      "epoch": 0.6686567164179105,
+      "grad_norm": 0.52734375,
+      "learning_rate": 4.109423525312738e-06,
+      "loss": 0.7825927734375,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.18714,
+      "step": 28,
+      "tokens/total": 58720256,
+      "tokens/train_per_sec_per_gpu": 454.68,
+      "tokens/trainable": 36535516
+    },
+    {
+      "epoch": 0.6925373134328359,
+      "grad_norm": 0.6171875,
+      "learning_rate": 3.777924554357096e-06,
+      "loss": 0.805419921875,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.23764,
+      "step": 29,
+      "tokens/total": 60817408,
+      "tokens/train_per_sec_per_gpu": 446.37,
+      "tokens/trainable": 37848236
+    },
+    {
+      "epoch": 0.7164179104477612,
+      "grad_norm": 0.81640625,
+      "learning_rate": 3.45704275117204e-06,
+      "loss": 0.8038330078125,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.23409,
+      "step": 30,
+      "tokens/total": 62857216,
+      "tokens/train_per_sec_per_gpu": 438.36,
+      "tokens/trainable": 39118856
+    },
+    {
+      "epoch": 0.7402985074626866,
+      "grad_norm": 0.3984375,
+      "learning_rate": 3.1487564587782306e-06,
+      "loss": 0.788330078125,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.19972,
+      "step": 31,
+      "tokens/total": 64954368,
+      "tokens/train_per_sec_per_gpu": 436.03,
+      "tokens/trainable": 40422428
+    },
+    {
+      "epoch": 0.764179104477612,
+      "grad_norm": 0.427734375,
+      "learning_rate": 2.854966364683872e-06,
+      "loss": 0.7547607421875,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.1271,
+      "step": 32,
+      "tokens/total": 67051520,
+      "tokens/train_per_sec_per_gpu": 428.1,
+      "tokens/trainable": 41732852
+    },
+    {
+      "epoch": 0.7880597014925373,
+      "grad_norm": 0.67578125,
+      "learning_rate": 2.577483782514174e-06,
+      "loss": 0.7708740234375,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.16165,
+      "step": 33,
+      "tokens/total": 69148672,
+      "tokens/train_per_sec_per_gpu": 455.57,
+      "tokens/trainable": 43032008
+    },
+    {
+      "epoch": 0.8119402985074626,
+      "grad_norm": 0.5546875,
+      "learning_rate": 2.3180194846605367e-06,
+      "loss": 0.76708984375,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.15349,
+      "step": 34,
+      "tokens/total": 71245824,
+      "tokens/train_per_sec_per_gpu": 471.74,
+      "tokens/trainable": 44341020
+    },
+    {
+      "epoch": 0.835820895522388,
+      "grad_norm": 0.408203125,
+      "learning_rate": 2.0781731547998614e-06,
+      "loss": 0.771240234375,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.16245,
+      "step": 35,
+      "tokens/total": 73342976,
+      "tokens/train_per_sec_per_gpu": 436.91,
+      "tokens/trainable": 45641400
+    },
+    {
+      "epoch": 0.8597014925373134,
+      "grad_norm": 0.427734375,
+      "learning_rate": 1.8594235253127373e-06,
+      "loss": 0.7926025390625,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.20914,
+      "step": 36,
+      "tokens/total": 75440128,
+      "tokens/train_per_sec_per_gpu": 431.34,
+      "tokens/trainable": 46950400
+    },
+    {
+      "epoch": 0.8835820895522388,
+      "grad_norm": 0.451171875,
+      "learning_rate": 1.6631192604065856e-06,
+      "loss": 0.7681884765625,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.15586,
+      "step": 37,
+      "tokens/total": 77537280,
+      "tokens/train_per_sec_per_gpu": 450.64,
+      "tokens/trainable": 48267888
+    },
+    {
+      "epoch": 0.9074626865671642,
+      "grad_norm": 0.4453125,
+      "learning_rate": 1.490470641152345e-06,
+      "loss": 0.77734375,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.17569,
+      "step": 38,
+      "tokens/total": 79634432,
+      "tokens/train_per_sec_per_gpu": 422.43,
+      "tokens/trainable": 49555608
+    },
+    {
+      "epoch": 0.9313432835820895,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.3425421036992098e-06,
+      "loss": 0.7789306640625,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.17914,
+      "step": 39,
+      "tokens/total": 81731584,
+      "tokens/train_per_sec_per_gpu": 422.98,
+      "tokens/trainable": 50858288
+    },
+    {
+      "epoch": 0.9552238805970149,
+      "grad_norm": 0.9609375,
+      "learning_rate": 1.2202456766718092e-06,
+      "loss": 0.759033203125,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.13621,
+      "step": 40,
+      "tokens/total": 83828736,
+      "tokens/train_per_sec_per_gpu": 454.1,
+      "tokens/trainable": 52180452
+    },
+    {
+      "epoch": 0.9791044776119403,
+      "grad_norm": 0.455078125,
+      "learning_rate": 1.1243353582104556e-06,
+      "loss": 0.7525634765625,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.12243,
+      "step": 41,
+      "tokens/total": 85925888,
+      "tokens/train_per_sec_per_gpu": 438.47,
+      "tokens/trainable": 53470404
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.44921875,
+      "learning_rate": 1.0554024673218808e-06,
+      "loss": 0.77685546875,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.17462,
+      "step": 42,
+      "tokens/total": 87760896,
+      "tokens/train_per_sec_per_gpu": 472.93,
+      "tokens/trainable": 54608264
+    },
+    {
+      "epoch": 1.0238805970149254,
+      "grad_norm": 0.447265625,
+      "learning_rate": 1.0138719982009243e-06,
+      "loss": 0.7520751953125,
+      "memory/device_reserved (GiB)": 89.53,
+      "memory/max_active (GiB)": 79.77,
+      "memory/max_allocated (GiB)": 79.77,
+      "ppl": 2.1214,
+      "step": 43,
+      "tokens/total": 89858048,
+      "tokens/train_per_sec_per_gpu": 434.64,
+      "tokens/trainable": 55922724
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 43,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 43,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.025374495315853e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}