Hugging Face's logo

tensor-tech
/
sn38-3

Safetensors
gpt_optimized
custom_code
Model card Files
xet
Community
tensor-tech commited on
Commit
d3d8cd0
·
verified ·
1 Parent(s): fbc4aca

Run 3. Outer Step 6. Inner Step 64.

Browse files
Files changed (2) hide show
  1. config.json +6 -6
  2. inner_optimizer.pt +1 -1
config.json CHANGED
@@ -268,18 +268,18 @@
268
  "AutoModelForCausalLM": "distributed/optimized-gpt2-500m--modeling_gpt_optimized.GPTOptim"
269
  },
270
  "block_list": [
271
- 5387147,
272
- 5387153,
273
- 5387158,
274
- 5387164,
275
- 5387169
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": 63,
283
  "inner_steps": 0,
284
  "last_allreduce_block": 5373127,
285
  "layer_norm_epsilon": 1e-05,
 
268
  "AutoModelForCausalLM": "distributed/optimized-gpt2-500m--modeling_gpt_optimized.GPTOptim"
269
  },
270
  "block_list": [
271
+ 5387196,
272
+ 5387201,
273
+ 5387207,
274
+ 5387212,
275
+ 5387217
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": 64,
283
  "inner_steps": 0,
284
  "last_allreduce_block": 5373127,
285
  "layer_norm_epsilon": 1e-05,
inner_optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:81e862dd0b9a4680c4897890046bbf3cec3dc3f3e9f45943828d1d644acf71d0
3
  size 8081782026
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c536dd06718bd4ce4d0d1bcba93a513bc53ad693c89ff175b3a8921f80168c15
3
  size 8081782026