Spaces:
Runtime error
Runtime error
Update run.py
Browse files
run.py
CHANGED
|
@@ -164,20 +164,20 @@ else:
|
|
| 164 |
import os
|
| 165 |
import requests
|
| 166 |
import subprocess
|
| 167 |
-
modelPath="/home/af/gguf/models/
|
| 168 |
if(os.path.exists(modelPath)==False):
|
| 169 |
#url="https://huggingface.co/TheBloke/WizardLM-13B-V1.2-GGUF/resolve/main/wizardlm-13b-v1.2.Q4_0.gguf"
|
| 170 |
-
|
| 171 |
#url="https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.2-GGUF/resolve/main/mistral-7b-instruct-v0.2.Q4_0.gguf?download=true"
|
| 172 |
-
url="https://huggingface.co/TheBloke/DiscoLM_German_7b_v1-GGUF/resolve/main/discolm_german_7b_v1.Q4_0.gguf?download=true"
|
| 173 |
response = requests.get(url)
|
| 174 |
-
with open("./
|
| 175 |
file.write(response.content)
|
| 176 |
print("Model downloaded")
|
| 177 |
-
modelPath="./
|
| 178 |
print(modelPath)
|
| 179 |
n="20"
|
| 180 |
-
if("
|
| 181 |
command = ["python3", "-m", "llama_cpp.server", "--model", modelPath, "--host", "0.0.0.0", "--port", "2600", "--n_threads", "8", "--n_gpu_layers", n]
|
| 182 |
subprocess.Popen(command)
|
| 183 |
print("Server ready!")
|
|
@@ -189,32 +189,28 @@ else:
|
|
| 189 |
def extend_prompt(message="", history=None, system=None, RAGAddon=None, system2=None, zeichenlimit=None,historylimit=4): #float("Inf")
|
| 190 |
startOfString=""
|
| 191 |
if zeichenlimit is None: zeichenlimit=1000000000 # :-)
|
| 192 |
-
template0=" [INST]{system}\n [/INST] </s>"
|
| 193 |
template1=" [INST] {message} [/INST]"
|
| 194 |
template2=" {response}</s>"
|
| 195 |
-
if("
|
| 196 |
template0="<start_of_turn>user{system}</end_of_turn>"
|
| 197 |
template1="<start_of_turn>user{message}</end_of_turn><start_of_turn>model"
|
| 198 |
template2="{response}</end_of_turn>"
|
| 199 |
-
if("
|
| 200 |
startOfString="<s>"
|
| 201 |
-
template0=" [INST]{system}\n [/INST] </s>"
|
| 202 |
template1=" [INST] {message} [/INST]"
|
| 203 |
template2=" {response}</s>"
|
| 204 |
if("Mistral-7B-Instruct" in modelPath): #https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2
|
| 205 |
startOfString="<s>"
|
| 206 |
-
template0="[INST]{system}\n [/INST]</s>"
|
| 207 |
template1="[INST] {message} [/INST]"
|
| 208 |
template2=" {response}</s>"
|
| 209 |
-
if("
|
| 210 |
template0="GPT4 Correct User: {system}<|end_of_turn|>GPT4 Correct Assistant: Okay.<|end_of_turn|>"
|
| 211 |
template1="GPT4 Correct User: {message}<|end_of_turn|>GPT4 Correct Assistant: "
|
| 212 |
template2="{response}<|end_of_turn|>"
|
| 213 |
-
if("SauerkrautLM-7b-HerO" in modelPath): #https://huggingface.co/VAGOsolutions/SauerkrautLM-7b-HerO
|
| 214 |
-
template0="<|im_start|>system\n{system}<|im_end|>\n"
|
| 215 |
-
template1="<|im_start|>user\n{message}<|im_end|>\n<|im_start|>assistant\n"
|
| 216 |
-
template2="{response}<|im_end|>\n"
|
| 217 |
-
if("discolm_german_7b" in modelPath): #https://huggingface.co/DiscoResearch/DiscoLM_German_7b_v1
|
| 218 |
template0="<|im_start|>system\n{system}<|im_end|>\n"
|
| 219 |
template1="<|im_start|>user\n{message}<|im_end|>\n<|im_start|>assistant\n"
|
| 220 |
template2="{response}<|im_end|>\n"
|
|
@@ -222,7 +218,7 @@ def extend_prompt(message="", history=None, system=None, RAGAddon=None, system2=
|
|
| 222 |
template0="{system} " #<s>
|
| 223 |
template1="USER: {message} ASSISTANT: "
|
| 224 |
template2="{response}</s>"
|
| 225 |
-
if("
|
| 226 |
template0="Instruct: {system}\nOutput: Okay.\n"
|
| 227 |
template1="Instruct: {message}\nOutput:"
|
| 228 |
template2="{response}\n"
|
|
|
|
| 164 |
import os
|
| 165 |
import requests
|
| 166 |
import subprocess
|
| 167 |
+
modelPath="/home/af/gguf/models/Discolm_german_7b_v1.Q4_0.gguf"
|
| 168 |
if(os.path.exists(modelPath)==False):
|
| 169 |
#url="https://huggingface.co/TheBloke/WizardLM-13B-V1.2-GGUF/resolve/main/wizardlm-13b-v1.2.Q4_0.gguf"
|
| 170 |
+
url="https://huggingface.co/TheBloke/Mixtral-8x7B-Instruct-v0.1-GGUF/resolve/main/mixtral-8x7b-instruct-v0.1.Q4_0.gguf?download=true"
|
| 171 |
#url="https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.2-GGUF/resolve/main/mistral-7b-instruct-v0.2.Q4_0.gguf?download=true"
|
| 172 |
+
#url="https://huggingface.co/TheBloke/DiscoLM_German_7b_v1-GGUF/resolve/main/discolm_german_7b_v1.Q4_0.gguf?download=true"
|
| 173 |
response = requests.get(url)
|
| 174 |
+
with open("./Mixtral-8x7b-instruct.gguf", mode="wb") as file:
|
| 175 |
file.write(response.content)
|
| 176 |
print("Model downloaded")
|
| 177 |
+
modelPath="./Mixtral-8x7b-instruct.gguf"
|
| 178 |
print(modelPath)
|
| 179 |
n="20"
|
| 180 |
+
if("Mixtral-8x7b-instruct" in modelPath): n="0" # mixtral seems to cause problems here...
|
| 181 |
command = ["python3", "-m", "llama_cpp.server", "--model", modelPath, "--host", "0.0.0.0", "--port", "2600", "--n_threads", "8", "--n_gpu_layers", n]
|
| 182 |
subprocess.Popen(command)
|
| 183 |
print("Server ready!")
|
|
|
|
| 189 |
def extend_prompt(message="", history=None, system=None, RAGAddon=None, system2=None, zeichenlimit=None,historylimit=4): #float("Inf")
|
| 190 |
startOfString=""
|
| 191 |
if zeichenlimit is None: zeichenlimit=1000000000 # :-)
|
| 192 |
+
template0=" [INST]{system}\n [/INST] </s>"
|
| 193 |
template1=" [INST] {message} [/INST]"
|
| 194 |
template2=" {response}</s>"
|
| 195 |
+
if("Gemma-" in modelPath): # https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1
|
| 196 |
template0="<start_of_turn>user{system}</end_of_turn>"
|
| 197 |
template1="<start_of_turn>user{message}</end_of_turn><start_of_turn>model"
|
| 198 |
template2="{response}</end_of_turn>"
|
| 199 |
+
if("Mixtral-8x7b-instruct" in modelPath): # https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1
|
| 200 |
startOfString="<s>"
|
| 201 |
+
template0=" [INST]{system}\n [/INST] </s>"
|
| 202 |
template1=" [INST] {message} [/INST]"
|
| 203 |
template2=" {response}</s>"
|
| 204 |
if("Mistral-7B-Instruct" in modelPath): #https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2
|
| 205 |
startOfString="<s>"
|
| 206 |
+
template0="[INST]{system}\n [/INST]</s>"
|
| 207 |
template1="[INST] {message} [/INST]"
|
| 208 |
template2=" {response}</s>"
|
| 209 |
+
if("Openchat-3.5" in modelPath): #https://huggingface.co/TheBloke/openchat-3.5-0106-GGUF
|
| 210 |
template0="GPT4 Correct User: {system}<|end_of_turn|>GPT4 Correct Assistant: Okay.<|end_of_turn|>"
|
| 211 |
template1="GPT4 Correct User: {message}<|end_of_turn|>GPT4 Correct Assistant: "
|
| 212 |
template2="{response}<|end_of_turn|>"
|
| 213 |
+
if(("Discolm_german_7b" in modelPath) or ("SauerkrautLM-7b-HerO" in modelPath)): #https://huggingface.co/VAGOsolutions/SauerkrautLM-7b-HerO
|
|
|
|
|
|
|
|
|
|
|
|
|
| 214 |
template0="<|im_start|>system\n{system}<|im_end|>\n"
|
| 215 |
template1="<|im_start|>user\n{message}<|im_end|>\n<|im_start|>assistant\n"
|
| 216 |
template2="{response}<|im_end|>\n"
|
|
|
|
| 218 |
template0="{system} " #<s>
|
| 219 |
template1="USER: {message} ASSISTANT: "
|
| 220 |
template2="{response}</s>"
|
| 221 |
+
if("Phi-2" in modelPath): #https://huggingface.co/TheBloke/phi-2-GGUF
|
| 222 |
template0="Instruct: {system}\nOutput: Okay.\n"
|
| 223 |
template1="Instruct: {message}\nOutput:"
|
| 224 |
template2="{response}\n"
|