HIT-SCIR/ELMoForManyLangs

German model doesn't have the cnn_50_100_512_4096_sample.json

agemagician opened this issue · 8 comments

agemagician commented

Hello,

I downloaded the German model and I tried to configure the config.json, but the zip file for the German model doesn't have the cnn_50_100_512_4096_sample.json.

Could you please upload the missing file ?

agemagician commented

I also checked English and Arabic models and I Cann't find the "cnn_50_100_512_4096_sample.json" file.

Did I missed something or it is really missing ?

agemagician commented

Ok, I think you need to make it a little bit clear in the readme file.

For anyone having the same issue, you need to set the path to the configs directory of this rep location in your local computer.
OR
You can download this file again from the configs directory of this rep to your model directory.

  • 👍2
Oneplus commented

readme was updated, thanks!

jiaok99 commented

useful

Mleader2 commented

thank's

Fireboyar commented

hello,i just want to know how to get the cnn_50_100_512_4096_sample.json

mustfkeskin commented

hello,i just want to know how to get the cnn_50_100_512_4096_sample.json

{
"encoder": {
"name": "elmo",
"projection_dim": 512,
"cell_clip": 3,
"proj_clip": 3,
"dim": 4096,
"n_layers": 2
},

"token_embedder": {
	"name": "cnn",
	"activation": "relu",
	"filters": [[1, 32], [2, 32], [3, 64], [4, 128], [5, 256], [6, 512], [7, 1024]],
	"n_highway": 2, 
	"word_dim": 100,
	"char_dim": 50,
	"max_characters_per_token": 50 		
},

"classifier": {
	"name": "sampled_softmax",
	"n_samples": 8192
},
"dropout": 0.1

}

  • 👍4
fooSynaptic commented

someone which cannot find the cnn model should check the model is located in the root dir of this rep.