ParisNeo
no thinking , no answer .. only DONE
Yesterday, i installed 2 lollms on 2 different machine and they worked well. today i wanted to delete and install again. After install , it works but no thinking, no answer just DONE
[INFO][2025-02-06 09:32:21] Selecting personality
- Selecting active personality 0 ...[INFO][2025-02-06 09:32:21] ok Selected lollms [INFO][2025-02-06 09:32:21] Saving configuration INFO: ::1:52670 - "POST /select_personality HTTP/1.1" 200 OK INFO: ::1:52670 - "POST /get_config HTTP/1.1" 200 OK INFO: ::1:52671 - "POST /get_config HTTP/1.1" 200 OK Listing all personalitiesINFO: ::1:52671 - "GET /list_bindings HTTP/1.1" 200 OK INFO: ::1:52671 - "GET /get_available_models HTTP/1.1" 200 OK Listing modelsok INFO: ::1:52671 - "GET /list_models HTTP/1.1" 200 OK Getting active model ok INFO: ::1:52671 - "GET /get_active_model HTTP/1.1" 200 OK INFO: ::1:52671 - "POST /get_config HTTP/1.1" 200 OK INFO: ::1:52671 - "POST /get_personality_languages_list HTTP/1.1" 200 OK INFO: ::1:52671 - "POST /get_personality_language HTTP/1.1" 200 OK INFO: ::1:52671 - "GET /is_rt_on HTTP/1.1" 200 OK OK INFO: ::1:52670 - "GET /get_all_personalities HTTP/1.1" 200 OK Loading discussion for client QPkiRKA1wyFMagQbAAAN ... ok INFO: ::1:52671 - "GET /is_rt_on HTTP/1.1" 200 OK INFO: ::1:52670 - "POST /get_discussion_files_list HTTP/1.1" 200 OK INFO: ::1:52672 - "POST /get_discussion_files_list HTTP/1.1" 200 OK INFO: ::1:52673 - "GET /get_generation_status HTTP/1.1" 200 OK Starting message generation by lollms [INFO][2025-02-06 09:32:34] Text generation requested by client: QPkiRKA1wyFMagQbAAAN [INFO][2025-02-06 09:32:34] Started generation task [INFO][2025-02-06 09:32:34] Received message : who are you (3) [INFO][2025-02-06 09:32:34] prompt has 1192 tokens [INFO][2025-02-06 09:32:34] warmup for generating up to 2898 tokens Llama.generate: 1 prefix-match hit, remaining 415 prompt tokens to eval llama_perf_context_print: load time = 1079.49 ms llama_perf_context_print: prompt eval time = 0.00 ms / 415 tokens ( 0.00 ms per token, inf tokens per second) llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second) llama_perf_context_print: total time = 7141.27 ms / 416 tokens [INFO][2025-02-06 09:32:42] Finished executing the generation
[INFO][2025-02-06 09:32:42] ## Done Generation ##
[INFO][2025-02-06 09:32:42] ╔══════════════════════════════════════════════════╗ [INFO][2025-02-06 09:32:42] ║ Done ║ [INFO][2025-02-06 09:32:42] ╚══════════════════════════════════════════════════╝ Llama.generate: 1 prefix-match hit, remaining 32 prompt tokens to eval INFO: ::1:52686 - "POST /get_discussion_files_list HTTP/1.1" 200 OK llama_perf_context_print: load time = 1079.49 ms llama_perf_context_print: prompt eval time = 0.00 ms / 32 tokens ( 0.00 ms per token, inf tokens per second) llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second) llama_perf_context_print: total time = 1120.33 ms / 33 tokens [INFO][2025-02-06 09:32:43]
Hi there,
Some models are rigid in terms of compliance to template, so I think with these, you would need to go to settings tab and in main configuration you can find a sub card named model template. Try selecting another template (not the lollms one) and it should work. If the model template does not exist in the list, make sure you take a look in the model card and manually set the template. Given the messages you are getting, nothing is wrong, it is just the model generating an end of sentence token.
I see that you are using a fusion between O1 deepseekR1 and qwen. This can happen as fine tuning may render the output model very template sensitive.
I will add an option to automatically extract the template format from the model itself, it became possible with the new versions of the binding.
Sorry for the inconveniance. I'm building this for free, alone and without any funds or external help.
Thanks your great job. And about "Sorry for the inconveniance. I'm building this for free, alone and without any funds or external help." i know what it means very well i am a delphi developer more then 30 years.. was alone any fund and no external help.. So, i would like to help you if i can but after 1-2 months.