Training checkpoint 152000 - 1.6B tokens

README.md

@@ -18,11 +18,11 @@ pipeline_tag: text-generation
 
 ## Current Training Status
 
-- **Training Step**: 150,000
-- **Tokens Processed**: 1.54B tokens
-- **Current Loss**: 4.6410
+- **Training Step**: 152,000
+- **Tokens Processed**: 1.56B tokens
+- **Current Loss**: 4.6221
 - **Spike Rate**: 0.0503
-- **Learning Rate**: 1.97e-04
+- **Learning Rate**: 1.96e-04
 
 ## Model Architecture
 
@@ -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_150000.pt"
+    filename="checkpoint_152000.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**: 10.2% complete towards 15B tokens
+**Training Progress**: 10.4% complete towards 15B tokens

checkpoint_152000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18c34df0d3fec6c545436791a61915b6b2a9fd3ed9bc2fd271b1c494bb7e5252
+size 999026666

config.json

@@ -4,11 +4,11 @@
   "hidden_size": 768,
   "num_layers": 12,
   "max_seq_length": 1024,
-  "training_step": 150000,
-  "tokens_processed": 1536000000,
-  "loss": 4.6410125829420315,
-  "spike_rate": 0.05033871127453886,
-  "learning_rate": 0.00019653410997216126,
-  "epoch": 0.3072,
-  "progress_percent": 10.24002621446711
+  "training_step": 152000,
+  "tokens_processed": 1556480000,
+  "loss": 4.622095163003601,
+  "spike_rate": 0.050331375331145026,
+  "learning_rate": 0.00019641900038833723,
+  "epoch": 0.311296,
+  "progress_percent": 10.37655989732667
 }