Training checkpoint 58000 - 0.6B tokens

README.md

@@ -18,9 +18,9 @@ pipeline_tag: text-generation
 
 ## Current Training Status
 
-- **Training Step**: 56,000
-- **Tokens Processed**: 0.57B tokens
-- **Current Loss**: 4.6716
+- **Training Step**: 58,000
+- **Tokens Processed**: 0.59B tokens
+- **Current Loss**: 4.7185
 - **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_56000.pt"
+    filename="checkpoint_58000.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**: 3.8% complete towards 15B tokens
+**Training Progress**: 4.0% complete towards 15B tokens

checkpoint_58000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29ec8b776b969741ae92e53596cb8116b1ee258a231dbb24c9c108fa4fe6a33c
+size 999026492

config.json

@@ -4,11 +4,11 @@
   "hidden_size": 768,
   "num_layers": 12,
   "max_seq_length": 1024,
-  "training_step": 56000,
-  "tokens_processed": 573440000,
-  "loss": 4.671550419552317,
-  "spike_rate": 0.05013260768717433,
-  "learning_rate": 0.0001998294083789053,
-  "epoch": 0.114688,
-  "progress_percent": 3.8229431200677206
+  "training_step": 58000,
+  "tokens_processed": 593920000,
+  "loss": 4.718483947219073,
+  "spike_rate": 0.05012724474420012,
+  "learning_rate": 0.00019980290846532584,
+  "epoch": 0.118784,
+  "progress_percent": 3.9594768029272824
 }