Training checkpoint 184000 - 1.9B tokens

README.md

@@ -18,11 +18,11 @@ pipeline_tag: text-generation
 
 ## Current Training Status
 
-- **Training Step**: 182,000
-- **Tokens Processed**: 1.86B tokens
-- **Current Loss**: 4.7050
+- **Training Step**: 184,000
+- **Tokens Processed**: 1.88B tokens
+- **Current Loss**: 4.5155
 - **Spike Rate**: 0.0504
-- **Learning Rate**: 1.94e-04
+- **Learning Rate**: 9.64e-06
 
 ## 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_182000.pt"
+    filename="checkpoint_184000.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**: 12.4% complete towards 15B tokens
+**Training Progress**: 12.6% complete towards 15B tokens

checkpoint_184000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed0691238b5c0f1be6bbb1d23639fb81f26d5d6a361718a35f1541ab67a61d2e
+size 999026730

config.json

@@ -4,11 +4,11 @@
   "hidden_size": 768,
   "num_layers": 12,
   "max_seq_length": 1024,
-  "training_step": 182000,
-  "tokens_processed": 1863680000,
-  "loss": 4.704965196601672,
-  "spike_rate": 0.0504144437587055,
-  "learning_rate": 0.00019447220794276915,
-  "epoch": 0.372736,
-  "progress_percent": 12.424565140220093
+  "training_step": 184000,
+  "tokens_processed": 1884160000,
+  "loss": 4.5154644428033786,
+  "spike_rate": 0.050425242981711645,
+  "learning_rate": 9.644922584877327e-06,
+  "epoch": 0.376832,
+  "progress_percent": 12.561066666666667
 }