OLMo-150M-as_fm3_tg_omi1_omi2-pkwCU0s1: step45000-unsharded-hf

pretrain/OLMo-150M-as_fm3_tg_omi1_omi2-pkwCU0s1/step45000-unsharded-hf/README.md

@@ -0,0 +1,17 @@
+---
+library_name: transformers
+tags: [olmo, echo-chamber, math, pretrain]
+---
+
+# OLMo-150M-as_fm3_tg_omi1_omi2-pkwCU0s1 / step 45000
+
+OLMo checkpoint (step 45000, 47.19B tokens)
+from a pretraining run that reproduces *Echo Chamber: RL Post-training Amplifies Behaviors Learned in
+Pretraining* (arXiv:2504.07912) with the authors' code. d_model 768, 12 layers,
+12 heads, mlp_ratio 8, weight tying False, Llama-2 tokenizer (32000).
+Mixture: see the run name (`as` Algebraic-Stack, `fm3` FineMath-3+, `tg` TinyGSM, `omi1`/`omi2`
+OpenMathInstruct-1/2, `mmqa` MetaMathQA, `<n>x` = n passes). Loss: cross entropy.
+
+Converted with `convert_olmo_to_hf_new.py --no_fix_eos_token_id` (so `eos_token_id` is 0 as in the
+authors' released checkpoints); the sibling `stepN-unsharded/` holds the native OLMo checkpoint with the
+optimizer state.

pretrain/OLMo-150M-as_fm3_tg_omi1_omi2-pkwCU0s1/step45000-unsharded-hf/config.json

@@ -0,0 +1,26 @@
+{
+  "architectures": [
+    "OlmoForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "clip_qkv": null,
+  "eos_token_id": 0,
+  "hidden_act": "silu",
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "max_position_embeddings": 2048,
+  "model_type": "olmo",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "num_key_value_heads": 12,
+  "pad_token_id": 1,
+  "rope_scaling": null,
+  "rope_theta": 10000.0,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float32",
+  "transformers_version": "4.50.0",
+  "use_cache": true,
+  "vocab_size": 32000
+}

pretrain/OLMo-150M-as_fm3_tg_omi1_omi2-pkwCU0s1/step45000-unsharded-hf/generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "eos_token_id": 0,
+  "pad_token_id": 1,
+  "transformers_version": "4.50.0"
+}

pretrain/OLMo-150M-as_fm3_tg_omi1_omi2-pkwCU0s1/step45000-unsharded-hf/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:944892fdc2226f0f0315a7c0ebd9eb0c3b8ce3f6c8252f70421240b1bd35a15b
+size 649602384

pretrain/OLMo-150M-as_fm3_tg_omi1_omi2-pkwCU0s1/step45000-unsharded-hf/special_tokens_map.json

@@ -0,0 +1,23 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

pretrain/OLMo-150M-as_fm3_tg_omi1_omi2-pkwCU0s1/step45000-unsharded-hf/tokenizer_config.json

@@ -0,0 +1,43 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "add_prefix_space": null,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "extra_special_tokens": {},
+  "legacy": false,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": null,
+  "padding_side": "right",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "LlamaTokenizerFast",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}