Nano.Deep.Reasoner.11m | 20,000 new examples | step 1,441

.gitattributes

@@ -33,3 +33,4 @@ 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
+seen_examples.jsonl filter=lfs diff=lfs merge=lfs -text

README.md

@@ -62,6 +62,14 @@ and padded to exactly `1096` positions.
 
 No oversized example is intentionally split across separate training examples.
 
+## Example selection
+
+Starting from Session 3, new examples are selected via a deterministic
+shuffled scan of the dataset (seeded, reproducible across runs) rather than
+raw sequential order, to avoid category/source concentration within a
+session. Sessions 1-2 (40,000 examples) were selected sequentially before
+this correction and remain part of the trained corpus.
+
 ## Dataset
 
 `Plans11/Organized_PreTrain_1k_Context`
@@ -94,8 +102,8 @@ The tokenizer becomes immutable after its initial creation.
 
 ## Current progress
 
-- Unique examples trained: 40,000
-- Global optimizer steps: 816
+- Unique examples trained: 60,000
+- Global optimizer steps: 1,441
 - Last session: 20,000
 
 This is an experimental small language model and is not guaranteed to

config.json

@@ -30,6 +30,7 @@
     "context_length": 1096,
     "strict_max_example_tokens": 1095,
     "causal_next_token_prediction": true,
+    "shuffled_deterministic_scan": true,
     "reasoning_tokens": [
       "<|think|>",
       "<|thought|>",

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7f9dc4e6787f4d1bf45204fe2544a66799c37ad80d173f3cb7fc600adb02910f
+oid sha256:688ea1e3ca87fbd698c6998a9c4c4696ec0213408ef4dc2222488716230c1f74
 size 50821608

progress.json

@@ -5,12 +5,35 @@
   "context_length": 1096,
   "vocab_size": 4096,
   "parameter_count": 11229120,
-  "global_step": 816,
-  "unique_examples_completed": 40000,
+  "global_step": 1441,
+  "unique_examples_completed": 60000,
   "last_session_examples": 20000,
   "max_session_examples": 20000,
-  "timestamp": 1787134475.6632547,
+  "last_session_category_distribution": {
+    "Instruct": 5032,
+    "Code_Instruct": 13020,
+    "Think": 1514,
+    "Reasoning": 166,
+    "Thought": 208,
+    "Tool_Calling": 41,
+    "Chat": 19
+  },
+  "last_session_source_distribution": {
+    "Plans11/Organized_PreTrain_WIUAI_1.3M": 10392,
+    "Plans11/Organized_PreTrain_Cyber_Security_640k": 5154,
+    "Plans11/Organized_PreTrain_Instruct_366k": 2852,
+    "Plans11/Organized_PreTrain_Frontier_Traces_47k": 318,
+    "Plans11/Organized_PreTrain_Frontier_Traces_150k": 504,
+    "Plans11/Organized_Pretrain_HR_213K": 260,
+    "Plans11/Organized_PreTrain_Frontier_Traces_34K": 120,
+    "Plans11/Organized_PreTrain_Persona_GOD_Seed_52K": 167,
+    "Plans11/Organized_PreTrain_Frontier_Traces_33k": 52,
+    "Plans11/Organized_PreTrain_Agent_287k": 41,
+    "Plans11/Organized_PreTrain_Coding_239k": 140
+  },
+  "timestamp": 1787136664.0159044,
   "causal_next_token": true,
   "strict_example_context": true,
-  "tokenizer_immutable": true
+  "tokenizer_immutable": true,
+  "shuffled_deterministic_scan": true
 }

training_metadata.json

@@ -6,8 +6,8 @@
   "strict_example_limit": 1095,
   "vocab_size": 4096,
   "dataset": "Plans11/Organized_PreTrain_1k_Context",
-  "global_step": 816,
-  "unique_examples_trained": 40000,
+  "global_step": 1441,
+  "unique_examples_trained": 60000,
   "last_session_size": 20000,
   "learning_rate": 0.0003,
   "batch_size": 8,
@@ -22,5 +22,6 @@
   ],
   "causal_objective": "next_token_prediction",
   "tokenizer_immutable": true,
-  "content_hash_deduplication": true
+  "content_hash_deduplication": true,
+  "shuffled_deterministic_scan": true
 }

training_state.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5ce8251ce45a4da4d787297cc8173504f30550cf68462a16e9615fec20a88819
+oid sha256:254905c05bfdb9a9566abd56f9dcd30238de77f558b4020871ef0f0d6d64866c
 size 134840827