audio
audioduration (s) 8.16
29.3
| sentence
stringclasses 6
values | start_time
stringclasses 6
values | end_time
stringclasses 6
values |
|---|---|---|---|
SPIN. Self play, fine tuning that improves LLMs. Tricksy, it's a form of fast-in for instance involving sparsity. Phi-2 is a model from Microsoft. Lightning-Attention-2 is an alternative to Flash-Attention. Mixtral-8x7B. This is a mixture of experts model. Solar 10.7B is a Mistral model with some extra layers added in. | 00:00:00.000 | 00:00:29.320 |
|
OpenChat is a fine tune of the Mistral model. Notux-8x7B-v1.0 This is a version of the Mixtral model, fine-tuned. Gemini Pro is Google's best model, or perhaps not as good as Gemini Ultra. Microsoft Phi-2, I've already mentioned. | 00:00:29.320 | 00:00:53.200 |
|
DeciLM is a high-speed 7B model, that's DeciLM-7B. Arena Elo is a means of comparing LLMs. MT Bench is another metric and MMLU is also. | 00:00:53.200 | 00:01:10.560 |
|
GPT-4 Turbo is a fast GPT-4 model by OpenAI. Mistral Medium is a mixture of experts but with larger experts than Mixtral-8x7B. Claude 1, Claude 2 or Claude 2.0 are the latest Claude models. Mixtral-8x7B-Instruct-v1, or rather, v0.1 that's Mixtral-8x7B- | 00:01:11.760 | 00:01:40.380 |
|
Instruct-v0.1. That's the latest mixture of experts. Yi-34B-Chat is a very strong fine tune of Llama. Claude-Instant-1 is one of the Claude models. Tulu-2-DPO-70B is a DPO fine tuned model by Allen AI. It's a fine tune of the Llama 2 model, the 70B. | 00:01:40.380 | 00:02:06.240 |
|
WizardLM-70B is also a fine tune of the Llama 70B model. | 00:02:06.240 | 00:02:14.400 |
README.md exists but content is empty.
Use the Edit dataset card button to edit it.
- Downloads last month
- 33