Olmo anchor at common step 300

chat_template.jinja

@@ -0,0 +1,16 @@
+{%- set has_system = messages|selectattr('role', 'equalto', 'system')|list|length > 0 -%}{%- if not has_system -%}{{- '<|im_start|>system
+You are a helpful function-calling AI assistant. ' -}}{%- if tools is none -%}{{- 'You do not currently have access to any functions. <functions></functions><|im_end|>
+' -}}{%- else -%}{{- 'You are provided with function signatures within <functions></functions> XML tags. You may call one or more functions to assist with the user query. Output any function calls within <function_calls></function_calls> XML tags. Do not make assumptions about what values to plug into functions.' -}}{{- '<functions>' -}}{{- tools | tojson -}}{{- '</functions><|im_end|>
+' -}}{%- endif -%}{%- endif -%}{%- for message in messages -%}{%- if message['role'] == 'system' -%}{{- '<|im_start|>system
+' + message['content'] -}}{%- if tools is not none -%}{{- '<functions>' -}}{{- tools | tojson -}}{{- '</functions>' -}}{%- elif message.get('functions', none) is not none -%}{{- ' <functions>' + message['functions'] + '</functions>' -}}{%- endif -%}{{- '<|im_end|>
+' -}}{%- elif message['role'] == 'user' -%}{{- '<|im_start|>user
+' + message['content'] + '<|im_end|>
+' -}}{%- elif message['role'] == 'assistant' -%}{{- '<|im_start|>assistant
+' -}}{%- if message.get('content', none) is not none -%}{{- message['content'] -}}{%- endif -%}{%- if message.get('function_calls', none) is not none -%}{{- '<function_calls>' + message['function_calls'] + '</function_calls>' -}}{% elif message.get('tool_calls', none) is not none %}{{- '<function_calls>' -}}{%- for tool_call in message['tool_calls'] %}{%- if tool_call is mapping and tool_call.get('function', none) is not none %}{%- set args = tool_call['function']['arguments'] -%}{%- set ns = namespace(arguments_list=[]) -%}{%- for key, value in args.items() -%}{%- set ns.arguments_list = ns.arguments_list + [key ~ '=' ~ (value | tojson)] -%}{%- endfor -%}{%- set arguments = ns.arguments_list | join(', ') -%}{{- tool_call['function']['name'] + '(' + arguments + ')' -}}{%- if not loop.last -%}{{ '
+' }}{%- endif -%}{% else %}{{- tool_call -}}{%- endif %}{%- endfor %}{{- '</function_calls>' -}}{%- endif -%}{%- if not loop.last -%}{{- '<|im_end|>' + '
+' -}}{%- else -%}{{- eos_token -}}{%- endif -%}{%- elif message['role'] == 'environment' -%}{{- '<|im_start|>environment
+' + message['content'] + '<|im_end|>
+' -}}{%- elif message['role'] == 'tool' -%}{{- '<|im_start|>environment
+' + message['content'] + '<|im_end|>
+' -}}{%- endif -%}{%- if loop.last and add_generation_prompt -%}{{- '<|im_start|>assistant
+' -}}{%- endif -%}{%- endfor -%}

config.json

@@ -0,0 +1,75 @@
+{
+  "architectures": [
+    "Olmo3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": null,
+  "dtype": "bfloat16",
+  "eos_token_id": 100257,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 11008,
+  "layer_types": [
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 65536,
+  "model_type": "olmo3",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 32,
+  "pad_token_id": 100277,
+  "rms_norm_eps": 1e-06,
+  "rope_parameters": {
+    "full_attention": {
+      "attention_factor": 1.2079441541679836,
+      "beta_fast": 32,
+      "beta_slow": 1,
+      "factor": 8.0,
+      "original_max_position_embeddings": 8192,
+      "rope_theta": 500000,
+      "rope_type": "yarn"
+    },
+    "sliding_attention": {
+      "rope_theta": 500000.0,
+      "rope_type": "default"
+    }
+  },
+  "sliding_window": 4096,
+  "tie_word_embeddings": false,
+  "transformers_version": "5.17.0",
+  "use_cache": true,
+  "vocab_size": 100278
+}

generation_config.json

@@ -0,0 +1,3 @@
+{
+  "transformers_version": "5.17.0"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a18bc2a462fa1564b76fd7cc5b5efa596ffabfcda556bdda76e7019e227539f2
+size 14596063960

run_manifest.json

@@ -0,0 +1,6 @@
+{
+ "arm": "olmo-compassion-web-s3407",
+ "exported_checkpoint": "checkpoint-300",
+ "export_rule": "Olmo common step 300 (2026-10-16): no step had every arm passing GATE_OLMO; 300 was the most even (TAC 32-39/52 completed for all arms)",
+ "recipe": "legacy part one, RoPE-guarded"
+}

tokenizer_config.json

@@ -0,0 +1,19 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "is_local": true,
+  "local_files_only": false,
+  "max_length": 8192,
+  "model_max_length": 65536,
+  "pad_token": "<|pad|>",
+  "padding_side": "right",
+  "stride": 0,
+  "tokenizer_class": "GPT2Tokenizer",
+  "truncation_side": "right",
+  "truncation_strategy": "longest_first",
+  "unk_token": "<|endoftext|>"
+}

trainer_state.json

@@ -0,0 +1,511 @@
+{
+  "best_global_step": 300,
+  "best_metric": 0.31705835461616516,
+  "best_model_checkpoint": "/workspace/anchor/olmo-compassion-web-s3407/train/checkpoint-300",
+  "epoch": 0.32076984763432237,
+  "eval_steps": 50,
+  "global_step": 300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.005346164127238706,
+      "grad_norm": 1.1581324338912964,
+      "learning_rate": 1.7021276595744682e-05,
+      "loss": 1.1219,
+      "step": 5
+    },
+    {
+      "epoch": 0.010692328254477412,
+      "grad_norm": 1.3560162782669067,
+      "learning_rate": 3.829787234042553e-05,
+      "loss": 0.9691,
+      "step": 10
+    },
+    {
+      "epoch": 0.01603849238171612,
+      "grad_norm": 0.800166130065918,
+      "learning_rate": 5.9574468085106384e-05,
+      "loss": 0.8709,
+      "step": 15
+    },
+    {
+      "epoch": 0.021384656508954825,
+      "grad_norm": 1.2073516845703125,
+      "learning_rate": 8.085106382978723e-05,
+      "loss": 0.6689,
+      "step": 20
+    },
+    {
+      "epoch": 0.02673082063619353,
+      "grad_norm": 1.3612271547317505,
+      "learning_rate": 0.00010212765957446809,
+      "loss": 0.6009,
+      "step": 25
+    },
+    {
+      "epoch": 0.03207698476343224,
+      "grad_norm": 1.600343108177185,
+      "learning_rate": 0.00012340425531914893,
+      "loss": 0.7026,
+      "step": 30
+    },
+    {
+      "epoch": 0.03742314889067094,
+      "grad_norm": 1.3995120525360107,
+      "learning_rate": 0.0001446808510638298,
+      "loss": 0.5802,
+      "step": 35
+    },
+    {
+      "epoch": 0.04276931301790965,
+      "grad_norm": 1.3475377559661865,
+      "learning_rate": 0.00016595744680851065,
+      "loss": 0.5142,
+      "step": 40
+    },
+    {
+      "epoch": 0.048115477145148355,
+      "grad_norm": 1.0339361429214478,
+      "learning_rate": 0.0001872340425531915,
+      "loss": 0.5379,
+      "step": 45
+    },
+    {
+      "epoch": 0.05346164127238706,
+      "grad_norm": 0.9642431735992432,
+      "learning_rate": 0.00019999750239122662,
+      "loss": 0.5008,
+      "step": 50
+    },
+    {
+      "epoch": 0.05346164127238706,
+      "eval_loss": 0.4235961437225342,
+      "eval_runtime": 199.0514,
+      "eval_samples_per_second": 1.929,
+      "eval_steps_per_second": 0.965,
+      "step": 50
+    },
+    {
+      "epoch": 0.05880780539962577,
+      "grad_norm": 1.158988118171692,
+      "learning_rate": 0.0001999694057253083,
+      "loss": 0.6179,
+      "step": 55
+    },
+    {
+      "epoch": 0.06415396952686447,
+      "grad_norm": 1.2102830410003662,
+      "learning_rate": 0.00019991009918336413,
+      "loss": 0.4687,
+      "step": 60
+    },
+    {
+      "epoch": 0.06950013365410318,
+      "grad_norm": 0.8248657584190369,
+      "learning_rate": 0.00019981960128055687,
+      "loss": 0.3937,
+      "step": 65
+    },
+    {
+      "epoch": 0.07484629778134189,
+      "grad_norm": 0.9328722953796387,
+      "learning_rate": 0.0001996979402698136,
+      "loss": 0.5076,
+      "step": 70
+    },
+    {
+      "epoch": 0.08019246190858059,
+      "grad_norm": 0.6152486801147461,
+      "learning_rate": 0.00019954515413300516,
+      "loss": 0.4881,
+      "step": 75
+    },
+    {
+      "epoch": 0.0855386260358193,
+      "grad_norm": 0.6562164425849915,
+      "learning_rate": 0.00019936129056908865,
+      "loss": 0.6226,
+      "step": 80
+    },
+    {
+      "epoch": 0.090884790163058,
+      "grad_norm": 0.6681051254272461,
+      "learning_rate": 0.00019914640697921585,
+      "loss": 0.4718,
+      "step": 85
+    },
+    {
+      "epoch": 0.09623095429029671,
+      "grad_norm": 0.5437894463539124,
+      "learning_rate": 0.00019890057044881306,
+      "loss": 0.5527,
+      "step": 90
+    },
+    {
+      "epoch": 0.10157711841753542,
+      "grad_norm": 0.657391369342804,
+      "learning_rate": 0.0001986238577266374,
+      "loss": 0.4103,
+      "step": 95
+    },
+    {
+      "epoch": 0.10692328254477412,
+      "grad_norm": 0.6284145712852478,
+      "learning_rate": 0.00019831635520081625,
+      "loss": 0.4827,
+      "step": 100
+    },
+    {
+      "epoch": 0.10692328254477412,
+      "eval_loss": 0.34383365511894226,
+      "eval_runtime": 69.5524,
+      "eval_samples_per_second": 5.521,
+      "eval_steps_per_second": 2.761,
+      "step": 100
+    },
+    {
+      "epoch": 0.11226944667201283,
+      "grad_norm": 0.638647198677063,
+      "learning_rate": 0.00019797815887187726,
+      "loss": 0.4555,
+      "step": 105
+    },
+    {
+      "epoch": 0.11761561079925154,
+      "grad_norm": 0.46918949484825134,
+      "learning_rate": 0.00019760937432277786,
+      "loss": 0.4608,
+      "step": 110
+    },
+    {
+      "epoch": 0.12296177492649024,
+      "grad_norm": 0.5412453413009644,
+      "learning_rate": 0.0001972101166859425,
+      "loss": 0.5782,
+      "step": 115
+    },
+    {
+      "epoch": 0.12830793905372895,
+      "grad_norm": 0.46240514516830444,
+      "learning_rate": 0.0001967805106073192,
+      "loss": 0.3468,
+      "step": 120
+    },
+    {
+      "epoch": 0.13365410318096765,
+      "grad_norm": 0.5159165263175964,
+      "learning_rate": 0.00019632069020746572,
+      "loss": 0.4631,
+      "step": 125
+    },
+    {
+      "epoch": 0.13900026730820636,
+      "grad_norm": 0.638958752155304,
+      "learning_rate": 0.00019583079903967785,
+      "loss": 0.5907,
+      "step": 130
+    },
+    {
+      "epoch": 0.14434643143544507,
+      "grad_norm": 0.6933673620223999,
+      "learning_rate": 0.000195310990045173,
+      "loss": 0.5108,
+      "step": 135
+    },
+    {
+      "epoch": 0.14969259556268377,
+      "grad_norm": 0.45193880796432495,
+      "learning_rate": 0.0001947614255053428,
+      "loss": 0.3193,
+      "step": 140
+    },
+    {
+      "epoch": 0.15503875968992248,
+      "grad_norm": 0.51473468542099,
+      "learning_rate": 0.00019418227699108956,
+      "loss": 0.4773,
+      "step": 145
+    },
+    {
+      "epoch": 0.16038492381716118,
+      "grad_norm": 0.506137490272522,
+      "learning_rate": 0.00019357372530926322,
+      "loss": 0.5489,
+      "step": 150
+    },
+    {
+      "epoch": 0.16038492381716118,
+      "eval_loss": 0.333650141954422,
+      "eval_runtime": 70.467,
+      "eval_samples_per_second": 5.449,
+      "eval_steps_per_second": 2.725,
+      "step": 150
+    },
+    {
+      "epoch": 0.1657310879443999,
+      "grad_norm": 0.5713365077972412,
+      "learning_rate": 0.00019293596044621416,
+      "loss": 0.4885,
+      "step": 155
+    },
+    {
+      "epoch": 0.1710772520716386,
+      "grad_norm": 0.6154384016990662,
+      "learning_rate": 0.00019226918150848068,
+      "loss": 0.4649,
+      "step": 160
+    },
+    {
+      "epoch": 0.1764234161988773,
+      "grad_norm": 0.7966128587722778,
+      "learning_rate": 0.00019157359666062903,
+      "loss": 0.4658,
+      "step": 165
+    },
+    {
+      "epoch": 0.181769580326116,
+      "grad_norm": 0.5122365355491638,
+      "learning_rate": 0.00019084942306026568,
+      "loss": 0.4831,
+      "step": 170
+    },
+    {
+      "epoch": 0.18711574445335472,
+      "grad_norm": 0.558390200138092,
+      "learning_rate": 0.0001900968867902419,
+      "loss": 0.5583,
+      "step": 175
+    },
+    {
+      "epoch": 0.19246190858059342,
+      "grad_norm": 0.5675086975097656,
+      "learning_rate": 0.00018931622278807212,
+      "loss": 0.578,
+      "step": 180
+    },
+    {
+      "epoch": 0.19780807270783213,
+      "grad_norm": 0.6163905262947083,
+      "learning_rate": 0.00018850767477258753,
+      "loss": 0.2651,
+      "step": 185
+    },
+    {
+      "epoch": 0.20315423683507083,
+      "grad_norm": 0.5942338705062866,
+      "learning_rate": 0.00018767149516784874,
+      "loss": 0.4397,
+      "step": 190
+    },
+    {
+      "epoch": 0.20850040096230954,
+      "grad_norm": 0.5109464526176453,
+      "learning_rate": 0.00018680794502434018,
+      "loss": 0.4011,
+      "step": 195
+    },
+    {
+      "epoch": 0.21384656508954825,
+      "grad_norm": 0.5306816101074219,
+      "learning_rate": 0.00018591729393747187,
+      "loss": 0.4759,
+      "step": 200
+    },
+    {
+      "epoch": 0.21384656508954825,
+      "eval_loss": 0.32967355847358704,
+      "eval_runtime": 70.2659,
+      "eval_samples_per_second": 5.465,
+      "eval_steps_per_second": 2.732,
+      "step": 200
+    },
+    {
+      "epoch": 0.21919272921678695,
+      "grad_norm": 0.5127652287483215,
+      "learning_rate": 0.0001849998199634131,
+      "loss": 0.5282,
+      "step": 205
+    },
+    {
+      "epoch": 0.22453889334402566,
+      "grad_norm": 0.5448350310325623,
+      "learning_rate": 0.00018405580953228493,
+      "loss": 0.4201,
+      "step": 210
+    },
+    {
+      "epoch": 0.22988505747126436,
+      "grad_norm": 0.42060723900794983,
+      "learning_rate": 0.00018308555735873844,
+      "loss": 0.3607,
+      "step": 215
+    },
+    {
+      "epoch": 0.23523122159850307,
+      "grad_norm": 0.6876929998397827,
+      "learning_rate": 0.0001820893663499464,
+      "loss": 0.5565,
+      "step": 220
+    },
+    {
+      "epoch": 0.24057738572574178,
+      "grad_norm": 0.48969799280166626,
+      "learning_rate": 0.00018106754751103735,
+      "loss": 0.4648,
+      "step": 225
+    },
+    {
+      "epoch": 0.24592354985298048,
+      "grad_norm": 0.4601738452911377,
+      "learning_rate": 0.00018002041984800174,
+      "loss": 0.4168,
+      "step": 230
+    },
+    {
+      "epoch": 0.2512697139802192,
+      "grad_norm": 0.47914230823516846,
+      "learning_rate": 0.00017894831026809985,
+      "loss": 0.4555,
+      "step": 235
+    },
+    {
+      "epoch": 0.2566158781074579,
+      "grad_norm": 0.592524528503418,
+      "learning_rate": 0.00017785155347780352,
+      "loss": 0.4027,
+      "step": 240
+    },
+    {
+      "epoch": 0.2619620422346966,
+      "grad_norm": 0.6497352719306946,
+      "learning_rate": 0.00017673049187830252,
+      "loss": 0.4129,
+      "step": 245
+    },
+    {
+      "epoch": 0.2673082063619353,
+      "grad_norm": 0.6578811407089233,
+      "learning_rate": 0.00017558547545860868,
+      "loss": 0.5317,
+      "step": 250
+    },
+    {
+      "epoch": 0.2673082063619353,
+      "eval_loss": 0.32128238677978516,
+      "eval_runtime": 71.2754,
+      "eval_samples_per_second": 5.388,
+      "eval_steps_per_second": 2.694,
+      "step": 250
+    },
+    {
+      "epoch": 0.272654370489174,
+      "grad_norm": 0.5883632898330688,
+      "learning_rate": 0.0001744168616862918,
+      "loss": 0.3787,
+      "step": 255
+    },
+    {
+      "epoch": 0.2780005346164127,
+      "grad_norm": 0.6391058564186096,
+      "learning_rate": 0.00017322501539587988,
+      "loss": 0.4548,
+      "step": 260
+    },
+    {
+      "epoch": 0.2833466987436514,
+      "grad_norm": 0.5929397940635681,
+      "learning_rate": 0.00017201030867496005,
+      "loss": 0.4696,
+      "step": 265
+    },
+    {
+      "epoch": 0.28869286287089013,
+      "grad_norm": 0.41368502378463745,
+      "learning_rate": 0.00017077312074801498,
+      "loss": 0.3731,
+      "step": 270
+    },
+    {
+      "epoch": 0.29403902699812884,
+      "grad_norm": 0.5383670926094055,
+      "learning_rate": 0.00016951383785803115,
+      "loss": 0.2873,
+      "step": 275
+    },
+    {
+      "epoch": 0.29938519112536754,
+      "grad_norm": 0.8195973634719849,
+      "learning_rate": 0.00016823285314591593,
+      "loss": 0.5282,
+      "step": 280
+    },
+    {
+      "epoch": 0.30473135525260625,
+      "grad_norm": 0.4255969822406769,
+      "learning_rate": 0.00016693056652776126,
+      "loss": 0.4153,
+      "step": 285
+    },
+    {
+      "epoch": 0.31007751937984496,
+      "grad_norm": 0.6933779716491699,
+      "learning_rate": 0.00016560738456999232,
+      "loss": 0.5884,
+      "step": 290
+    },
+    {
+      "epoch": 0.31542368350708366,
+      "grad_norm": 0.5111712217330933,
+      "learning_rate": 0.00016426372036243945,
+      "loss": 0.4062,
+      "step": 295
+    },
+    {
+      "epoch": 0.32076984763432237,
+      "grad_norm": 0.5927426815032959,
+      "learning_rate": 0.00016289999338937427,
+      "loss": 0.4782,
+      "step": 300
+    },
+    {
+      "epoch": 0.32076984763432237,
+      "eval_loss": 0.31705835461616516,
+      "eval_runtime": 70.7238,
+      "eval_samples_per_second": 5.43,
+      "eval_steps_per_second": 2.715,
+      "step": 300
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 936,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 100,
+  "stateful_callbacks": {
+    "EarlyStoppingCallback": {
+      "args": {
+        "early_stopping_patience": 2,
+        "early_stopping_threshold": 0.001
+      },
+      "attributes": {
+        "early_stopping_patience_counter": 0
+      }
+    },
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.165295431271383e+17,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}