Training checkpoint 60000 - 0.6B tokens

README.md

@@ -18,9 +18,9 @@ pipeline_tag: text-generation
 
 ## Current Training Status
 
-- **Training Step**: 58,000
-- **Tokens Processed**: 0.59B tokens
-- **Current Loss**: 4.7185
+- **Training Step**: 60,000
+- **Tokens Processed**: 0.61B tokens
+- **Current Loss**: 4.6101
 - **Spike Rate**: 0.0501
 - **Learning Rate**: 2.00e-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_58000.pt"
+    filename="checkpoint_60000.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**: 4.0% complete towards 15B tokens
+**Training Progress**: 4.1% complete towards 15B tokens

checkpoint_60000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc7afa99c8cb8bf25a5914f56bce31ac538b241bd96a76f773a3223ef79833b0
+size 999026492

config.json

@@ -4,11 +4,11 @@
   "hidden_size": 768,
   "num_layers": 12,
   "max_seq_length": 1024,
-  "training_step": 58000,
-  "tokens_processed": 593920000,
-  "loss": 4.718483947219073,
-  "spike_rate": 0.05012724474420012,
-  "learning_rate": 0.00019980290846532584,
-  "epoch": 0.118784,
-  "progress_percent": 3.9594768029272824
+  "training_step": 60000,
+  "tokens_processed": 614400000,
+  "loss": 4.610068006355983,
+  "spike_rate": 0.050140802952125564,
+  "learning_rate": 0.00019977449816433704,
+  "epoch": 0.12288,
+  "progress_percent": 4.096010485786843
 }