Training checkpoint 84000 - 0.9B tokens

README.md

@@ -18,9 +18,9 @@ pipeline_tag: text-generation
 
 ## Current Training Status
 
-- **Training Step**: 82,000
-- **Tokens Processed**: 0.84B tokens
-- **Current Loss**: 4.6582
+- **Training Step**: 84,000
+- **Tokens Processed**: 0.86B tokens
+- **Current Loss**: 4.7287
 - **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_82000.pt"
+    filename="checkpoint_84000.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**: 5.6% complete towards 15B tokens
+**Training Progress**: 5.7% complete towards 15B tokens

checkpoint_84000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98771e730a17d68bbebd45cb7d4382889d527d9476f2ef0aab895b883ab5904f
+size 999026492

config.json

@@ -4,11 +4,11 @@
   "hidden_size": 768,
   "num_layers": 12,
   "max_seq_length": 1024,
-  "training_step": 82000,
-  "tokens_processed": 839680000,
-  "loss": 4.658208935955752,
-  "spike_rate": 0.050245297771789446,
-  "learning_rate": 0.00019933608108536793,
-  "epoch": 0.167936,
-  "progress_percent": 5.59788099724202
+  "training_step": 84000,
+  "tokens_processed": 860160000,
+  "loss": 4.728705899119385,
+  "spike_rate": 0.050229199544315054,
+  "learning_rate": 0.00019928479904231087,
+  "epoch": 0.172032,
+  "progress_percent": 5.7344146801015805
 }