Training checkpoint 100000 - 1.0B tokens

README.md

@@ -18,9 +18,9 @@ pipeline_tag: text-generation
 
 ## Current Training Status
 
-- **Training Step**: 98,000
-- **Tokens Processed**: 1.00B tokens
-- **Current Loss**: 4.6930
+- **Training Step**: 100,000
+- **Tokens Processed**: 1.02B tokens
+- **Current Loss**: 4.6722
 - **Spike Rate**: 0.0502
 - **Learning Rate**: 1.99e-04
 
@@ -54,7 +54,7 @@ This represents ongoing training of the first large-scale spiking neural network
 from huggingface_hub import hf_hub_download
 checkpoint = hf_hub_download(
     repo_id="rootxhacker/piking-llm-5b-3epochs-exp",
-    filename="checkpoint_98000.pt"
+    filename="checkpoint_100000.pt"
 )
 
 # Load with custom spiking model code
@@ -65,4 +65,4 @@ checkpoint = hf_hub_download(
 
 **🔬 This is live research in progress! Check back for updates as training continues.**
 
-**Training Progress**: 6.7% complete towards 15B tokens
+**Training Progress**: 6.8% complete towards 15B tokens

checkpoint_100000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bae099d358b752cda7f6a42f8f9a95d67cea887ae0ed78dd54deab26597e0d36
+size 999026666

config.json

@@ -4,11 +4,11 @@
   "hidden_size": 768,
   "num_layers": 12,
   "max_seq_length": 1024,
-  "training_step": 98000,
-  "tokens_processed": 1003520000,
-  "loss": 4.692955921361772,
-  "spike_rate": 0.05024591377919592,
-  "learning_rate": 0.00019887267140425137,
-  "epoch": 0.200704,
-  "progress_percent": 6.690150460118511
+  "training_step": 100000,
+  "tokens_processed": 1024000000,
+  "loss": 4.672218005677345,
+  "spike_rate": 0.050236698603096354,
+  "learning_rate": 0.00019880621622818317,
+  "epoch": 0.2048,
+  "progress_percent": 6.826684142978073
 }