Tongjilibo
commited on
Commit
•
af50f21
1
Parent(s):
d73ebe7
chatglm中增加generation_config
Browse files- chatglm-6b-int4/bert4torch_config.json +2 -1
- chatglm-6b-int8/bert4torch_config.json +2 -1
- chatglm-6b-v0.1.0/bert4torch_config.json +2 -1
- chatglm-6b/bert4torch_config.json +2 -1
- chatglm2-6b-32k/bert4torch_config.json +2 -1
- chatglm2-6b-int4/bert4torch_config.json +2 -1
- chatglm2-6b/bert4torch_config.json +2 -1
- chatglm3-6b-32k/bert4torch_config.json +2 -1
- chatglm3-6b/bert4torch_config.json +2 -1
- internlm-chat-7b/bert4torch_config.json +2 -1
chatglm-6b-int4/bert4torch_config.json
CHANGED
@@ -21,5 +21,6 @@
|
|
21 |
"tie_emb_prj_weight": false,
|
22 |
"quantization_bit": 4,
|
23 |
"quantization_method": "cpm_kernels",
|
24 |
-
"target_modules": ["q", "k", "v", "o", "intermediateDense", "outputDense"]
|
|
|
25 |
}
|
|
|
21 |
"tie_emb_prj_weight": false,
|
22 |
"quantization_bit": 4,
|
23 |
"quantization_method": "cpm_kernels",
|
24 |
+
"target_modules": ["q", "k", "v", "o", "intermediateDense", "outputDense"],
|
25 |
+
"generation_config": {"max_length": 2048}
|
26 |
}
|
chatglm-6b-int8/bert4torch_config.json
CHANGED
@@ -21,5 +21,6 @@
|
|
21 |
"quantization_bit": 8,
|
22 |
"quantization_method": "cpm_kernels",
|
23 |
"target_modules": ["q", "k", "v", "o", "intermediateDense", "outputDense"],
|
24 |
-
"tie_emb_prj_weight": false
|
|
|
25 |
}
|
|
|
21 |
"quantization_bit": 8,
|
22 |
"quantization_method": "cpm_kernels",
|
23 |
"target_modules": ["q", "k", "v", "o", "intermediateDense", "outputDense"],
|
24 |
+
"tie_emb_prj_weight": false,
|
25 |
+
"generation_config": {"max_length": 2048}
|
26 |
}
|
chatglm-6b-v0.1.0/bert4torch_config.json
CHANGED
@@ -18,5 +18,6 @@
|
|
18 |
"segment_vocab_size": 0,
|
19 |
"skip_init": true,
|
20 |
"rope_rank": "updown",
|
21 |
-
"tie_emb_prj_weight": false
|
|
|
22 |
}
|
|
|
18 |
"segment_vocab_size": 0,
|
19 |
"skip_init": true,
|
20 |
"rope_rank": "updown",
|
21 |
+
"tie_emb_prj_weight": false,
|
22 |
+
"generation_config": {"max_length": 2048}
|
23 |
}
|
chatglm-6b/bert4torch_config.json
CHANGED
@@ -18,5 +18,6 @@
|
|
18 |
"segment_vocab_size": 0,
|
19 |
"skip_init": true,
|
20 |
"rope_rank": "updown",
|
21 |
-
"tie_emb_prj_weight": false
|
|
|
22 |
}
|
|
|
18 |
"segment_vocab_size": 0,
|
19 |
"skip_init": true,
|
20 |
"rope_rank": "updown",
|
21 |
+
"tie_emb_prj_weight": false,
|
22 |
+
"generation_config": {"max_length": 2048}
|
23 |
}
|
chatglm2-6b-32k/bert4torch_config.json
CHANGED
@@ -18,5 +18,6 @@
|
|
18 |
"rope_rank": "adjacent",
|
19 |
"rope_ratio": 16,
|
20 |
"position_encoding_2d": true,
|
21 |
-
"flash_attention": true
|
|
|
22 |
}
|
|
|
18 |
"rope_rank": "adjacent",
|
19 |
"rope_ratio": 16,
|
20 |
"position_encoding_2d": true,
|
21 |
+
"flash_attention": true,
|
22 |
+
"generation_config": {"max_length": 32768}
|
23 |
}
|
chatglm2-6b-int4/bert4torch_config.json
CHANGED
@@ -20,5 +20,6 @@
|
|
20 |
"flash_attention": true,
|
21 |
"quantization_bit": 4,
|
22 |
"quantization_method": "cpm_kernels",
|
23 |
-
"target_modules": ["q", "k", "v", "o", "intermediateDense", "outputDense"]
|
|
|
24 |
}
|
|
|
20 |
"flash_attention": true,
|
21 |
"quantization_bit": 4,
|
22 |
"quantization_method": "cpm_kernels",
|
23 |
+
"target_modules": ["q", "k", "v", "o", "intermediateDense", "outputDense"],
|
24 |
+
"generation_config": {"max_length": 32768}
|
25 |
}
|
chatglm2-6b/bert4torch_config.json
CHANGED
@@ -17,5 +17,6 @@
|
|
17 |
"rmsnorm": true,
|
18 |
"rope_rank": "adjacent",
|
19 |
"position_encoding_2d": true,
|
20 |
-
"flash_attention": true
|
|
|
21 |
}
|
|
|
17 |
"rmsnorm": true,
|
18 |
"rope_rank": "adjacent",
|
19 |
"position_encoding_2d": true,
|
20 |
+
"flash_attention": true,
|
21 |
+
"generation_config": {"max_length": 32768}
|
22 |
}
|
chatglm3-6b-32k/bert4torch_config.json
CHANGED
@@ -18,5 +18,6 @@
|
|
18 |
"rope_rank": "adjacent",
|
19 |
"ntk_alpha": 44.24632364389211,
|
20 |
"position_encoding_2d": true,
|
21 |
-
"flash_attention": true
|
|
|
22 |
}
|
|
|
18 |
"rope_rank": "adjacent",
|
19 |
"ntk_alpha": 44.24632364389211,
|
20 |
"position_encoding_2d": true,
|
21 |
+
"flash_attention": true,
|
22 |
+
"generation_config": {"end_id": [2, 64795, 64797], "max_length": 32768}
|
23 |
}
|
chatglm3-6b/bert4torch_config.json
CHANGED
@@ -17,5 +17,6 @@
|
|
17 |
"rmsnorm": true,
|
18 |
"rope_rank": "adjacent",
|
19 |
"position_encoding_2d": true,
|
20 |
-
"flash_attention": true
|
|
|
21 |
}
|
|
|
17 |
"rmsnorm": true,
|
18 |
"rope_rank": "adjacent",
|
19 |
"position_encoding_2d": true,
|
20 |
+
"flash_attention": true,
|
21 |
+
"generation_config": {"end_id": [2, 64795, 64797], "max_length": 8192}
|
22 |
}
|
internlm-chat-7b/bert4torch_config.json
CHANGED
@@ -12,5 +12,6 @@
|
|
12 |
"skip_init": true,
|
13 |
"rope_rank": "updown",
|
14 |
"torch_dtype": "float16",
|
15 |
-
"tie_word_embeddings": false
|
|
|
16 |
}
|
|
|
12 |
"skip_init": true,
|
13 |
"rope_rank": "updown",
|
14 |
"torch_dtype": "float16",
|
15 |
+
"tie_word_embeddings": false,
|
16 |
+
"generation_config": {"tokenizer_config": {"skip_special_tokens": true}, "end_id": [2, 103028]}
|
17 |
}
|