Spaces:
Runtime error
Runtime error
Miaoran000
commited on
Commit
•
0544709
1
Parent(s):
fd9d58a
update model_operations.py for new llms
Browse files- src/backend/model_operations.py +44 -16
src/backend/model_operations.py
CHANGED
@@ -164,7 +164,7 @@ class SummaryGenerator:
|
|
164 |
using_replicate_api = False
|
165 |
replicate_api_models = ['snowflake', 'llama-3.1-405b']
|
166 |
using_pipeline = False
|
167 |
-
pipeline_models = ['llama-3.1', 'phi-3-mini','falcon-7b', 'phi-3.5', 'mistral-nemo']
|
168 |
|
169 |
for replicate_api_model in replicate_api_models:
|
170 |
if replicate_api_model in self.model_id.lower():
|
@@ -222,6 +222,7 @@ class SummaryGenerator:
|
|
222 |
print(result)
|
223 |
return result
|
224 |
|
|
|
225 |
elif 'grok' in self.model_id.lower(): # xai
|
226 |
XAI_API_KEY = os.getenv("XAI_API_KEY")
|
227 |
client = OpenAI(
|
@@ -241,6 +242,7 @@ class SummaryGenerator:
|
|
241 |
print(result)
|
242 |
return result
|
243 |
|
|
|
244 |
elif 'gemini' in self.model_id.lower():
|
245 |
vertexai.init(project=os.getenv("GOOGLE_PROJECT_ID"), location="us-central1")
|
246 |
model = GenerativeModel(
|
@@ -249,7 +251,7 @@ class SummaryGenerator:
|
|
249 |
)
|
250 |
generation_config = {
|
251 |
"temperature": 0,
|
252 |
-
"max_output_tokens":
|
253 |
}
|
254 |
safety_settings = [
|
255 |
SafetySetting(
|
@@ -277,6 +279,8 @@ class SummaryGenerator:
|
|
277 |
result = response.text
|
278 |
print(result)
|
279 |
return result
|
|
|
|
|
280 |
elif using_replicate_api:
|
281 |
print("using replicate")
|
282 |
if 'snowflake' in self.model_id.lower():
|
@@ -306,6 +310,7 @@ class SummaryGenerator:
|
|
306 |
print(response)
|
307 |
return response
|
308 |
|
|
|
309 |
elif 'claude' in self.model_id.lower(): # using anthropic api
|
310 |
print('using Anthropic API')
|
311 |
client = anthropic.Anthropic()
|
@@ -331,6 +336,7 @@ class SummaryGenerator:
|
|
331 |
print(result)
|
332 |
return result
|
333 |
|
|
|
334 |
elif 'command-r' in self.model_id.lower() or 'aya-expanse' in self.model_id.lower():
|
335 |
co = cohere.ClientV2(os.getenv('COHERE_API_TOKEN'))
|
336 |
response = co.chat(
|
@@ -345,6 +351,7 @@ class SummaryGenerator:
|
|
345 |
print(result)
|
346 |
return result
|
347 |
|
|
|
348 |
elif 'mistral-large' in self.model_id.lower():
|
349 |
api_key = os.environ["MISTRAL_API_KEY"]
|
350 |
client = Mistral(api_key=api_key)
|
@@ -369,6 +376,7 @@ class SummaryGenerator:
|
|
369 |
print(result)
|
370 |
return result
|
371 |
|
|
|
372 |
elif 'deepseek' in self.model_id.lower():
|
373 |
client = OpenAI(api_key=os.getenv("DeepSeek_API_KEY"), base_url="https://api.deepseek.com")
|
374 |
response = client.chat.completions.create(
|
@@ -385,20 +393,21 @@ class SummaryGenerator:
|
|
385 |
print(result)
|
386 |
return result
|
387 |
|
388 |
-
# Using HF
|
389 |
elif self.local_model is None and self.local_pipeline is None:
|
390 |
if using_pipeline:
|
391 |
self.local_pipeline = pipeline(
|
392 |
"text-generation",
|
393 |
model=self.model_id,
|
394 |
tokenizer=AutoTokenizer.from_pretrained(self.model_id),
|
395 |
-
torch_dtype=torch.bfloat16 if 'llama-3.2' in self.model_id.lower() else "auto",
|
396 |
device_map="auto",
|
397 |
trust_remote_code=True
|
398 |
)
|
399 |
else:
|
400 |
if 'ragamuffin' in self.model_id.lower():
|
401 |
self.tokenizer = AutoTokenizer.from_pretrained(os.path.join('/home/miaoran', self.model_id))
|
|
|
402 |
else:
|
403 |
self.tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-hf" if 'openelm' in self.model_id.lower() else self.model_id, trust_remote_code=True)
|
404 |
print("Tokenizer loaded")
|
@@ -420,7 +429,12 @@ class SummaryGenerator:
|
|
420 |
# self.local_model = AutoModelForCausalLM.from_pretrained(os.path.join('/home/miaoran', self.model_id),
|
421 |
# torch_dtype=torch.bfloat16, # forcing bfloat16 for now
|
422 |
# attn_implementation="flash_attention_2")
|
423 |
-
|
|
|
|
|
|
|
|
|
|
|
424 |
else:
|
425 |
self.local_model = AutoModelForCausalLM.from_pretrained(self.model_id, trust_remote_code=True, device_map="auto")#torch_dtype="auto"
|
426 |
# print(self.local_model.device)
|
@@ -435,7 +449,7 @@ class SummaryGenerator:
|
|
435 |
]
|
436 |
outputs = self.local_pipeline(
|
437 |
messages,
|
438 |
-
max_new_tokens=
|
439 |
# return_full_text=False,
|
440 |
do_sample=False
|
441 |
)
|
@@ -445,6 +459,8 @@ class SummaryGenerator:
|
|
445 |
|
446 |
elif self.local_model: # cannot call API. using local model / pipeline
|
447 |
print('Using local model')
|
|
|
|
|
448 |
if 'gemma' in self.model_id.lower() or 'mistral-7b' in self.model_id.lower():
|
449 |
messages=[
|
450 |
# gemma-1.1, mistral-7b does not accept system role
|
@@ -478,29 +494,41 @@ class SummaryGenerator:
|
|
478 |
{"role": "system", "content": system_prompt},
|
479 |
{"role": "user", "content": user_prompt}
|
480 |
]
|
481 |
-
prompt = self.tokenizer.apply_chat_template(messages,add_generation_prompt=True, tokenize=False)
|
482 |
-
|
483 |
-
#
|
484 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
485 |
if 'granite' in self.model_id.lower():
|
486 |
self.local_model.eval()
|
487 |
outputs = self.local_model.generate(**input_ids, max_new_tokens=250)
|
|
|
|
|
|
|
|
|
488 |
else:
|
489 |
with torch.no_grad():
|
490 |
outputs = self.local_model.generate(**input_ids, do_sample=True, max_new_tokens=250, temperature=0.01)#, pad_token_id=self.tokenizer.eos_token_id
|
491 |
if 'glm' in self.model_id.lower() or 'ragamuffin' in self.model_id.lower() or 'granite' in self.model_id.lower():
|
492 |
outputs = outputs[:, input_ids['input_ids'].shape[1]:]
|
493 |
-
elif 'qwen2-vl' in self.model_id.lower() or 'qwen2.5' in self.model_id.lower():
|
494 |
outputs = [
|
495 |
out_ids[len(in_ids) :] for in_ids, out_ids in zip(input_ids.input_ids, outputs)
|
496 |
]
|
497 |
-
|
|
|
498 |
if 'qwen2-vl' in self.model_id.lower():
|
499 |
result = self.processor.batch_decode(
|
500 |
outputs, skip_special_tokens=True, clean_up_tokenization_spaces=False
|
501 |
)[0]
|
502 |
-
|
503 |
-
|
504 |
else:
|
505 |
result = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
|
506 |
|
@@ -512,9 +540,9 @@ class SummaryGenerator:
|
|
512 |
result = result.split(messages[-1]['content'])[1].strip()
|
513 |
elif 'qwen2-vl' in self.model_id.lower() or 'qwen2.5' in self.model_id.lower():
|
514 |
pass
|
|
|
|
|
515 |
else:
|
516 |
-
# print(prompt)
|
517 |
-
# print('-'*50)
|
518 |
result = result.replace(prompt.strip(), '')
|
519 |
|
520 |
print(result)
|
|
|
164 |
using_replicate_api = False
|
165 |
replicate_api_models = ['snowflake', 'llama-3.1-405b']
|
166 |
using_pipeline = False
|
167 |
+
pipeline_models = ['llama-3.1', 'phi-3-mini','falcon-7b', 'phi-3.5', 'mistral-nemo', 'llama-3.3']
|
168 |
|
169 |
for replicate_api_model in replicate_api_models:
|
170 |
if replicate_api_model in self.model_id.lower():
|
|
|
222 |
print(result)
|
223 |
return result
|
224 |
|
225 |
+
# Using Grok API
|
226 |
elif 'grok' in self.model_id.lower(): # xai
|
227 |
XAI_API_KEY = os.getenv("XAI_API_KEY")
|
228 |
client = OpenAI(
|
|
|
242 |
print(result)
|
243 |
return result
|
244 |
|
245 |
+
# Using Vertex AI API for Gemini models
|
246 |
elif 'gemini' in self.model_id.lower():
|
247 |
vertexai.init(project=os.getenv("GOOGLE_PROJECT_ID"), location="us-central1")
|
248 |
model = GenerativeModel(
|
|
|
251 |
)
|
252 |
generation_config = {
|
253 |
"temperature": 0,
|
254 |
+
"max_output_tokens": 500
|
255 |
}
|
256 |
safety_settings = [
|
257 |
SafetySetting(
|
|
|
279 |
result = response.text
|
280 |
print(result)
|
281 |
return result
|
282 |
+
|
283 |
+
# Using Replicate API
|
284 |
elif using_replicate_api:
|
285 |
print("using replicate")
|
286 |
if 'snowflake' in self.model_id.lower():
|
|
|
310 |
print(response)
|
311 |
return response
|
312 |
|
313 |
+
# Using Anthropic API for Claude models
|
314 |
elif 'claude' in self.model_id.lower(): # using anthropic api
|
315 |
print('using Anthropic API')
|
316 |
client = anthropic.Anthropic()
|
|
|
336 |
print(result)
|
337 |
return result
|
338 |
|
339 |
+
# Using Cohere API
|
340 |
elif 'command-r' in self.model_id.lower() or 'aya-expanse' in self.model_id.lower():
|
341 |
co = cohere.ClientV2(os.getenv('COHERE_API_TOKEN'))
|
342 |
response = co.chat(
|
|
|
351 |
print(result)
|
352 |
return result
|
353 |
|
354 |
+
# Using MistralAI API
|
355 |
elif 'mistral-large' in self.model_id.lower():
|
356 |
api_key = os.environ["MISTRAL_API_KEY"]
|
357 |
client = Mistral(api_key=api_key)
|
|
|
376 |
print(result)
|
377 |
return result
|
378 |
|
379 |
+
# Using Deepseek API
|
380 |
elif 'deepseek' in self.model_id.lower():
|
381 |
client = OpenAI(api_key=os.getenv("DeepSeek_API_KEY"), base_url="https://api.deepseek.com")
|
382 |
response = client.chat.completions.create(
|
|
|
393 |
print(result)
|
394 |
return result
|
395 |
|
396 |
+
# Using HF pipeline or local checkpoints
|
397 |
elif self.local_model is None and self.local_pipeline is None:
|
398 |
if using_pipeline:
|
399 |
self.local_pipeline = pipeline(
|
400 |
"text-generation",
|
401 |
model=self.model_id,
|
402 |
tokenizer=AutoTokenizer.from_pretrained(self.model_id),
|
403 |
+
torch_dtype=torch.bfloat16 if 'llama-3.2' in self.model_id.lower() or 'llama-3.3' in self.model_id.lower() else "auto",
|
404 |
device_map="auto",
|
405 |
trust_remote_code=True
|
406 |
)
|
407 |
else:
|
408 |
if 'ragamuffin' in self.model_id.lower():
|
409 |
self.tokenizer = AutoTokenizer.from_pretrained(os.path.join('/home/miaoran', self.model_id))
|
410 |
+
|
411 |
else:
|
412 |
self.tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-hf" if 'openelm' in self.model_id.lower() else self.model_id, trust_remote_code=True)
|
413 |
print("Tokenizer loaded")
|
|
|
429 |
# self.local_model = AutoModelForCausalLM.from_pretrained(os.path.join('/home/miaoran', self.model_id),
|
430 |
# torch_dtype=torch.bfloat16, # forcing bfloat16 for now
|
431 |
# attn_implementation="flash_attention_2")
|
432 |
+
elif 'olmo' in self.model_id.lower():
|
433 |
+
self.local_model = AutoModelForCausalLM.from_pretrained(self.model_id)#torch_dtype="auto"
|
434 |
+
|
435 |
+
elif 'qwq-' in self.model_id.lower():
|
436 |
+
self.local_model = AutoModelForCausalLM.from_pretrained(self.model_id, torch_dtype="auto", device_map="auto")
|
437 |
+
|
438 |
else:
|
439 |
self.local_model = AutoModelForCausalLM.from_pretrained(self.model_id, trust_remote_code=True, device_map="auto")#torch_dtype="auto"
|
440 |
# print(self.local_model.device)
|
|
|
449 |
]
|
450 |
outputs = self.local_pipeline(
|
451 |
messages,
|
452 |
+
max_new_tokens=256,
|
453 |
# return_full_text=False,
|
454 |
do_sample=False
|
455 |
)
|
|
|
459 |
|
460 |
elif self.local_model: # cannot call API. using local model / pipeline
|
461 |
print('Using local model')
|
462 |
+
|
463 |
+
# Set appropriate prompt based on model document
|
464 |
if 'gemma' in self.model_id.lower() or 'mistral-7b' in self.model_id.lower():
|
465 |
messages=[
|
466 |
# gemma-1.1, mistral-7b does not accept system role
|
|
|
494 |
{"role": "system", "content": system_prompt},
|
495 |
{"role": "user", "content": user_prompt}
|
496 |
]
|
497 |
+
prompt = self.tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)
|
498 |
+
|
499 |
+
# Tokenize inputs
|
500 |
+
if 'olmo' in self.model_id.lower():
|
501 |
+
input_ids = self.tokenizer([prompt], return_tensors='pt', return_token_type_ids=False)#.to(self.device)
|
502 |
+
elif 'qwq' in self.model_id.lower():
|
503 |
+
input_ids = self.tokenizer([prompt], return_tensors="pt").to(self.device)
|
504 |
+
else:
|
505 |
+
input_ids = self.tokenizer(prompt, return_tensors="pt").to(self.device)
|
506 |
+
|
507 |
+
# Generate outputs
|
508 |
if 'granite' in self.model_id.lower():
|
509 |
self.local_model.eval()
|
510 |
outputs = self.local_model.generate(**input_ids, max_new_tokens=250)
|
511 |
+
elif 'olmo' in self.model_id.lower():
|
512 |
+
outputs = self.local_model.generate(**input_ids, max_new_tokens=250, do_sample=True, temperature=0.01)#top_k=50, top_p=0.95)
|
513 |
+
elif 'qwq' in self.model_id.lower():
|
514 |
+
outputs = self.local_model.generate(**input_ids, max_new_tokens=512, do_sample=True, temperature=0.01)
|
515 |
else:
|
516 |
with torch.no_grad():
|
517 |
outputs = self.local_model.generate(**input_ids, do_sample=True, max_new_tokens=250, temperature=0.01)#, pad_token_id=self.tokenizer.eos_token_id
|
518 |
if 'glm' in self.model_id.lower() or 'ragamuffin' in self.model_id.lower() or 'granite' in self.model_id.lower():
|
519 |
outputs = outputs[:, input_ids['input_ids'].shape[1]:]
|
520 |
+
elif 'qwen2-vl' in self.model_id.lower() or 'qwen2.5' in self.model_id.lower() or 'qwq-' in self.model_id.lower():
|
521 |
outputs = [
|
522 |
out_ids[len(in_ids) :] for in_ids, out_ids in zip(input_ids.input_ids, outputs)
|
523 |
]
|
524 |
+
|
525 |
+
# Decode outputs
|
526 |
if 'qwen2-vl' in self.model_id.lower():
|
527 |
result = self.processor.batch_decode(
|
528 |
outputs, skip_special_tokens=True, clean_up_tokenization_spaces=False
|
529 |
)[0]
|
530 |
+
elif 'olmo' in self.model_id.lower() or 'qwq' in self.model_id.lower():
|
531 |
+
result = self.tokenizer.batch_decode(outputs, skip_special_tokens=True)[0]
|
532 |
else:
|
533 |
result = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
|
534 |
|
|
|
540 |
result = result.split(messages[-1]['content'])[1].strip()
|
541 |
elif 'qwen2-vl' in self.model_id.lower() or 'qwen2.5' in self.model_id.lower():
|
542 |
pass
|
543 |
+
elif 'olmo' in self.model_id.lower():
|
544 |
+
result = result.split("<|assistant|>\n")[-1]
|
545 |
else:
|
|
|
|
|
546 |
result = result.replace(prompt.strip(), '')
|
547 |
|
548 |
print(result)
|