Training checkpoint 24000 - 0.2B tokens

README.md

@@ -18,11 +18,11 @@ pipeline_tag: text-generation
 
 ## Current Training Status
 
-- **Training Step**: 22,000
-- **Tokens Processed**: 0.23B tokens
-- **Current Loss**: 4.6581
-- **Spike Rate**: 0.0128
-- **Learning Rate**: 1.72e-04
+- **Training Step**: 24,000
+- **Tokens Processed**: 0.25B tokens
+- **Current Loss**: 4.7107
+- **Spike Rate**: 0.0134
+- **Learning Rate**: 1.85e-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_22000.pt"
+    filename="checkpoint_24000.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**: 1.5% complete towards 15B tokens
+**Training Progress**: 1.6% complete towards 15B tokens

checkpoint_24000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77d7b10c161cdce049b5ff5d11e24a9f2d197fb81da98791f6a86f7937b30a01
+size 999026492

config.json

@@ -4,11 +4,11 @@
   "hidden_size": 768,
   "num_layers": 12,
   "max_seq_length": 1024,
-  "training_step": 22000,
-  "tokens_processed": 225280000,
-  "loss": 4.65813939004963,
-  "spike_rate": 0.012800171157018493,
-  "learning_rate": 0.0001720923089817558,
-  "epoch": 0.045056,
-  "progress_percent": 1.501870511455176
+  "training_step": 24000,
+  "tokens_processed": 245760000,
+  "loss": 4.710656542452263,
+  "spike_rate": 0.013364357058069197,
+  "learning_rate": 0.00018493535648803967,
+  "epoch": 0.049152,
+  "progress_percent": 1.6384041943147374
 }