Hugging Face's logo

asim22
/
sub38-26

Safetensors
gpt_optimized
custom_code
Model card Files Community
asim22 commited on
Commit
eba0c20
·
verified ·
1 Parent(s): 5665395

Outer Step 2. Inner Step 540. Batch Size 272

Browse files
Files changed (2) hide show
  1. config.json +6 -6
  2. model.safetensors +1 -1
config.json CHANGED
@@ -268,18 +268,18 @@
268
  "AutoModelForCausalLM": "distributed/optimized-gpt2-500m--modeling_gpt_optimized.GPTOptim"
269
  },
270
  "block_list": [
271
- 5113652,
272
- 5113656,
273
- 5113661,
274
- 5113665,
275
- 5113668
276
  ],
277
  "block_size": 1024,
278
  "bos_token_id": 50256,
279
  "embd_pdrop": 0.1,
280
  "eos_token_id": 50256,
281
  "initializer_range": 0.02,
282
- "inner_step": 536,
283
  "inner_steps": 0,
284
  "last_allreduce_block": 5072405,
285
  "layer_norm_epsilon": 1e-05,
 
268
  "AutoModelForCausalLM": "distributed/optimized-gpt2-500m--modeling_gpt_optimized.GPTOptim"
269
  },
270
  "block_list": [
271
+ 5113685,
272
+ 5113689,
273
+ 5113693,
274
+ 5113696,
275
+ 5113700
276
  ],
277
  "block_size": 1024,
278
  "bos_token_id": 50256,
279
  "embd_pdrop": 0.1,
280
  "eos_token_id": 50256,
281
  "initializer_range": 0.02,
282
+ "inner_step": 540,
283
  "inner_steps": 0,
284
  "last_allreduce_block": 5072405,
285
  "layer_norm_epsilon": 1e-05,
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:79eed6c6d14a10df25f81a93c85018b99a5c2bd21b8e416e041ddb2a975228cf
3
  size 4040701744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:892608e5875d653511dadc3797041cfd4efa385704531b615738a0e17a6e9c4f
3
  size 4040701744