ExternalVendorCode/lollms
1
0
Fork 0
mirror of https://github.com/ParisNeo/lollms.git synced 2024-12-18 20:27:58 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

added generate image and enhanced context handling

Browse Source
This commit is contained in:
Saifeddine ALOUI 2024-08-06 15:44:58 +02:00
parent 3ddec9895d
commit 716b9fa86c
1 changed files with 10 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
lollms/server/endpoints/lollms_generator.py
View File

@ -125,9 +125,11 @@ async def lollms_generate(request: LollmsGenerateRequest):
prompt = request.prompt prompt = request.prompt
if elf_server.config.debug: if elf_server.config.debug:
ASCIIColors.yellow(prompt) ASCIIColors.yellow(prompt)
n_predict = request.n_predict if request.n_predict>0 else 1024 tokens = elf_server.model.tokenize(prompt)
n_tokens = len(tokens)
ASCIIColors.yellow(f"Prompt input size {n_tokens}")
n_predict = min(min(elf_server.config.ctx_size-n_tokens-1,elf_server.config.max_n_predict), request.n_predict) if request.n_predict>0 else min(elf_server.config.ctx_size-n_tokens-1,elf_server.config.max_n_predict)
stream = request.stream stream = request.stream
prompt_tokens = len(elf_server.binding.tokenize(prompt))
if elf_server.binding is not None: if elf_server.binding is not None:
if stream: if stream:
new_output={"new_values":[]} new_output={"new_values":[]}
@ -200,6 +202,7 @@ async def lollms_generate(request: LollmsGenerateRequest):
return True return True
elf_server.binding.generate( elf_server.binding.generate(
prompt, prompt,
n_predict, n_predict,
@ -207,6 +210,7 @@ async def lollms_generate(request: LollmsGenerateRequest):
temperature=request.temperature or elf_server.config.temperature temperature=request.temperature or elf_server.config.temperature
) )
completion_tokens = len(elf_server.binding.tokenize(reception_manager.reception_buffer)) completion_tokens = len(elf_server.binding.tokenize(reception_manager.reception_buffer))
ASCIIColors.yellow(f"Generated: {completion_tokens} tokens")
return PlainTextResponse(reception_manager.reception_buffer) return PlainTextResponse(reception_manager.reception_buffer)
else: else:
return None return None
@ -265,7 +269,10 @@ async def lollms_generate_with_images(request: LollmsGenerateRequest):
reception_manager=RECEPTION_MANAGER() reception_manager=RECEPTION_MANAGER()
prompt = request.prompt prompt = request.prompt
encoded_images = request.images encoded_images = request.images
n_predict = request.n_predict if request.n_predict>0 else 1024 tokens = elf_server.model.tokenize(prompt)
n_tokens = len(tokens)
ASCIIColors.yellow(f"Prompt input size {n_tokens}")
n_predict = min(min(elf_server.config.ctx_size-n_tokens-1,elf_server.config.max_n_predict), request.n_predict) if request.n_predict>0 else min(elf_server.config.ctx_size-n_tokens-1,elf_server.config.max_n_predict)
stream = request.stream stream = request.stream
prompt_tokens = len(elf_server.binding.tokenize(prompt)) prompt_tokens = len(elf_server.binding.tokenize(prompt))
if elf_server.binding is not None: if elf_server.binding is not None: