Spaces:
Running
on
T4
Running
on
T4
sparkleman
commited on
Commit
·
96d9268
1
Parent(s):
8dbcd8c
UPDATE
Browse files- config.production.yaml +22 -22
config.production.yaml
CHANGED
|
@@ -4,12 +4,10 @@ STRATEGY: "cuda fp16"
|
|
| 4 |
RWKV_CUDA_ON: True
|
| 5 |
CHUNK_LEN: 256
|
| 6 |
MODELS:
|
| 7 |
-
- SERVICE_NAME: "
|
| 8 |
-
DOWNLOAD_MODEL_FILE_NAME: "
|
| 9 |
-
DOWNLOAD_MODEL_REPO_ID: "BlinkDL/
|
| 10 |
REASONING: True
|
| 11 |
-
DEFAULT_CHAT: True
|
| 12 |
-
DEFAULT_REASONING: True
|
| 13 |
DEFAULT_SAMPLER:
|
| 14 |
max_tokens: 4096
|
| 15 |
temperature: 1.0
|
|
@@ -21,21 +19,6 @@ MODELS:
|
|
| 21 |
- "\n\n"
|
| 22 |
stop_tokens:
|
| 23 |
- 0
|
| 24 |
-
# - SERVICE_NAME: "RWKV7-G1-0.4B-75%trained-20250317-ctx4k"
|
| 25 |
-
# DOWNLOAD_MODEL_FILE_NAME: "RWKV7-G1-0.4B-75%trained-20250317-ctx4k.pth"
|
| 26 |
-
# DOWNLOAD_MODEL_REPO_ID: "BlinkDL/temp-latest-training-models"
|
| 27 |
-
# REASONING: True
|
| 28 |
-
# DEFAULT_SAMPLER:
|
| 29 |
-
# max_tokens: 4096
|
| 30 |
-
# temperature: 1.0
|
| 31 |
-
# top_p: 0.3
|
| 32 |
-
# presence_penalty: 0.5
|
| 33 |
-
# count_penalty: 0.5
|
| 34 |
-
# penalty_decay: 0.996
|
| 35 |
-
# stop:
|
| 36 |
-
# - "\n\n"
|
| 37 |
-
# stop_tokens:
|
| 38 |
-
# - 0
|
| 39 |
- SERVICE_NAME: "RWKV7-G1-1.5B-32%trained-20250319-ctx4k"
|
| 40 |
DOWNLOAD_MODEL_FILE_NAME: "RWKV7-G1-1.5B-32%trained-20250319-ctx4k.pth"
|
| 41 |
DOWNLOAD_MODEL_REPO_ID: "BlinkDL/temp-latest-training-models"
|
|
@@ -51,8 +34,8 @@ MODELS:
|
|
| 51 |
- "\n\n"
|
| 52 |
stop_tokens:
|
| 53 |
- 0
|
| 54 |
-
- SERVICE_NAME: "RWKV7-G1-
|
| 55 |
-
DOWNLOAD_MODEL_FILE_NAME: "RWKV7-G1-
|
| 56 |
DOWNLOAD_MODEL_REPO_ID: "BlinkDL/temp-latest-training-models"
|
| 57 |
REASONING: True
|
| 58 |
DEFAULT_SAMPLER:
|
|
@@ -66,3 +49,20 @@ MODELS:
|
|
| 66 |
- "\n\n"
|
| 67 |
stop_tokens:
|
| 68 |
- 0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
RWKV_CUDA_ON: True
|
| 5 |
CHUNK_LEN: 256
|
| 6 |
MODELS:
|
| 7 |
+
- SERVICE_NAME: "RWKV7-G1-2.9B-16%trained-20250313-ctx4k"
|
| 8 |
+
DOWNLOAD_MODEL_FILE_NAME: "RWKV7-G1-2.9B-16%trained-20250313-ctx4k.pth"
|
| 9 |
+
DOWNLOAD_MODEL_REPO_ID: "BlinkDL/temp-latest-training-models"
|
| 10 |
REASONING: True
|
|
|
|
|
|
|
| 11 |
DEFAULT_SAMPLER:
|
| 12 |
max_tokens: 4096
|
| 13 |
temperature: 1.0
|
|
|
|
| 19 |
- "\n\n"
|
| 20 |
stop_tokens:
|
| 21 |
- 0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
- SERVICE_NAME: "RWKV7-G1-1.5B-32%trained-20250319-ctx4k"
|
| 23 |
DOWNLOAD_MODEL_FILE_NAME: "RWKV7-G1-1.5B-32%trained-20250319-ctx4k.pth"
|
| 24 |
DOWNLOAD_MODEL_REPO_ID: "BlinkDL/temp-latest-training-models"
|
|
|
|
| 34 |
- "\n\n"
|
| 35 |
stop_tokens:
|
| 36 |
- 0
|
| 37 |
+
- SERVICE_NAME: "RWKV7-G1-0.4B-75%trained-20250317-ctx4k"
|
| 38 |
+
DOWNLOAD_MODEL_FILE_NAME: "RWKV7-G1-0.4B-75%trained-20250317-ctx4k.pth"
|
| 39 |
DOWNLOAD_MODEL_REPO_ID: "BlinkDL/temp-latest-training-models"
|
| 40 |
REASONING: True
|
| 41 |
DEFAULT_SAMPLER:
|
|
|
|
| 49 |
- "\n\n"
|
| 50 |
stop_tokens:
|
| 51 |
- 0
|
| 52 |
+
- SERVICE_NAME: "rwkv7-g1-0.1b-20250307-ctx4096"
|
| 53 |
+
DOWNLOAD_MODEL_FILE_NAME: "rwkv7-g1-0.1b-20250307-ctx4096.pth"
|
| 54 |
+
DOWNLOAD_MODEL_REPO_ID: "BlinkDL/rwkv7-g1"
|
| 55 |
+
REASONING: True
|
| 56 |
+
DEFAULT_CHAT: True
|
| 57 |
+
DEFAULT_REASONING: True
|
| 58 |
+
DEFAULT_SAMPLER:
|
| 59 |
+
max_tokens: 4096
|
| 60 |
+
temperature: 1.0
|
| 61 |
+
top_p: 0.3
|
| 62 |
+
presence_penalty: 0.5
|
| 63 |
+
count_penalty: 0.5
|
| 64 |
+
penalty_decay: 0.996
|
| 65 |
+
stop:
|
| 66 |
+
- "\n\n"
|
| 67 |
+
stop_tokens:
|
| 68 |
+
- 0
|