abhinavkulkarni
commited on
Commit
•
6981fd8
1
Parent(s):
caf93be
Upload MPTForCausalLM
Browse files- config.json +4 -4
- generation_config.json +1 -1
config.json
CHANGED
@@ -6,7 +6,7 @@
|
|
6 |
"attn_config": {
|
7 |
"alibi": true,
|
8 |
"alibi_bias_max": 8,
|
9 |
-
"attn_impl": "
|
10 |
"attn_pdrop": 0,
|
11 |
"attn_type": "multihead_attention",
|
12 |
"attn_uses_sequence_id": false,
|
@@ -16,8 +16,8 @@
|
|
16 |
"softmax_scale": null
|
17 |
},
|
18 |
"auto_map": {
|
19 |
-
"AutoConfig": "configuration_mpt.MPTConfig",
|
20 |
-
"AutoModelForCausalLM": "modeling_mpt.MPTForCausalLM"
|
21 |
},
|
22 |
"d_model": 4096,
|
23 |
"emb_pdrop": 0,
|
@@ -46,7 +46,7 @@
|
|
46 |
"resid_pdrop": 0,
|
47 |
"tokenizer_name": "sam-mosaic/gpt-neox-20b-chatml",
|
48 |
"torch_dtype": "float16",
|
49 |
-
"transformers_version": "4.
|
50 |
"use_cache": false,
|
51 |
"verbose": 0,
|
52 |
"vocab_size": 50432
|
|
|
6 |
"attn_config": {
|
7 |
"alibi": true,
|
8 |
"alibi_bias_max": 8,
|
9 |
+
"attn_impl": "triton",
|
10 |
"attn_pdrop": 0,
|
11 |
"attn_type": "multihead_attention",
|
12 |
"attn_uses_sequence_id": false,
|
|
|
16 |
"softmax_scale": null
|
17 |
},
|
18 |
"auto_map": {
|
19 |
+
"AutoConfig": "mosaicml/mpt-7b-chat--configuration_mpt.MPTConfig",
|
20 |
+
"AutoModelForCausalLM": "mosaicml/mpt-7b-chat--modeling_mpt.MPTForCausalLM"
|
21 |
},
|
22 |
"d_model": 4096,
|
23 |
"emb_pdrop": 0,
|
|
|
46 |
"resid_pdrop": 0,
|
47 |
"tokenizer_name": "sam-mosaic/gpt-neox-20b-chatml",
|
48 |
"torch_dtype": "float16",
|
49 |
+
"transformers_version": "4.30.0.dev0",
|
50 |
"use_cache": false,
|
51 |
"verbose": 0,
|
52 |
"vocab_size": 50432
|
generation_config.json
CHANGED
@@ -1,5 +1,5 @@
|
|
1 |
{
|
2 |
"_from_model_config": true,
|
3 |
-
"transformers_version": "4.
|
4 |
"use_cache": false
|
5 |
}
|
|
|
1 |
{
|
2 |
"_from_model_config": true,
|
3 |
+
"transformers_version": "4.30.0.dev0",
|
4 |
"use_cache": false
|
5 |
}
|